New results

pythia-14m-seed1/step0/EleutherAI__pythia-14m-seed1/results_2024-07-05T13-24-18.246759.json.bak

@@ -0,0 +1,2606 @@
+{
+  "results": {
+    "blimp": {
+      "acc,none": 0.5175074626865671,
+      "acc_stderr,none": 0.001882002716920592,
+      "alias": "blimp"
+    },
+    "blimp_adjunct_island": {
+      "acc,none": 0.56,
+      "acc_stderr,none": 0.015704987954361718,
+      "alias": " - blimp_adjunct_island"
+    },
+    "blimp_anaphor_gender_agreement": {
+      "acc,none": 0.558,
+      "acc_stderr,none": 0.01571250721186415,
+      "alias": " - blimp_anaphor_gender_agreement"
+    },
+    "blimp_anaphor_number_agreement": {
+      "acc,none": 0.539,
+      "acc_stderr,none": 0.015771104201283047,
+      "alias": " - blimp_anaphor_number_agreement"
+    },
+    "blimp_animate_subject_passive": {
+      "acc,none": 0.593,
+      "acc_stderr,none": 0.015543249100255474,
+      "alias": " - blimp_animate_subject_passive"
+    },
+    "blimp_animate_subject_trans": {
+      "acc,none": 0.813,
+      "acc_stderr,none": 0.012336254828074168,
+      "alias": " - blimp_animate_subject_trans"
+    },
+    "blimp_causative": {
+      "acc,none": 0.457,
+      "acc_stderr,none": 0.0157606915901365,
+      "alias": " - blimp_causative"
+    },
+    "blimp_complex_NP_island": {
+      "acc,none": 0.509,
+      "acc_stderr,none": 0.0158167369950053,
+      "alias": " - blimp_complex_NP_island"
+    },
+    "blimp_coordinate_structure_constraint_complex_left_branch": {
+      "acc,none": 0.545,
+      "acc_stderr,none": 0.015755101498347232,
+      "alias": " - blimp_coordinate_structure_constraint_complex_left_branch"
+    },
+    "blimp_coordinate_structure_constraint_object_extraction": {
+      "acc,none": 0.531,
+      "acc_stderr,none": 0.015788865959538965,
+      "alias": " - blimp_coordinate_structure_constraint_object_extraction"
+    },
+    "blimp_determiner_noun_agreement_1": {
+      "acc,none": 0.538,
+      "acc_stderr,none": 0.015773547629015002,
+      "alias": " - blimp_determiner_noun_agreement_1"
+    },
+    "blimp_determiner_noun_agreement_2": {
+      "acc,none": 0.53,
+      "acc_stderr,none": 0.015790799515836725,
+      "alias": " - blimp_determiner_noun_agreement_2"
+    },
+    "blimp_determiner_noun_agreement_irregular_1": {
+      "acc,none": 0.498,
+      "acc_stderr,none": 0.01581917337430266,
+      "alias": " - blimp_determiner_noun_agreement_irregular_1"
+    },
+    "blimp_determiner_noun_agreement_irregular_2": {
+      "acc,none": 0.471,
+      "acc_stderr,none": 0.015792669451628764,
+      "alias": " - blimp_determiner_noun_agreement_irregular_2"
+    },
+    "blimp_determiner_noun_agreement_with_adj_2": {
+      "acc,none": 0.538,
+      "acc_stderr,none": 0.015773547629015002,
+      "alias": " - blimp_determiner_noun_agreement_with_adj_2"
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": {
+      "acc,none": 0.54,
+      "acc_stderr,none": 0.0157685969143944,
+      "alias": " - blimp_determiner_noun_agreement_with_adj_irregular_1"
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": {
+      "acc,none": 0.525,
+      "acc_stderr,none": 0.015799513429996023,
+      "alias": " - blimp_determiner_noun_agreement_with_adj_irregular_2"
+    },
+    "blimp_determiner_noun_agreement_with_adjective_1": {
+      "acc,none": 0.499,
+      "acc_stderr,none": 0.015819268290576817,
+      "alias": " - blimp_determiner_noun_agreement_with_adjective_1"
+    },
+    "blimp_distractor_agreement_relational_noun": {
+      "acc,none": 0.477,
+      "acc_stderr,none": 0.015802554246726087,
+      "alias": " - blimp_distractor_agreement_relational_noun"
+    },
+    "blimp_distractor_agreement_relative_clause": {
+      "acc,none": 0.511,
+      "acc_stderr,none": 0.01581547119529257,
+      "alias": " - blimp_distractor_agreement_relative_clause"
+    },
+    "blimp_drop_argument": {
+      "acc,none": 0.663,
+      "acc_stderr,none": 0.01495508791865346,
+      "alias": " - blimp_drop_argument"
+    },
+    "blimp_ellipsis_n_bar_1": {
+      "acc,none": 0.431,
+      "acc_stderr,none": 0.01566794448817344,
+      "alias": " - blimp_ellipsis_n_bar_1"
+    },
+    "blimp_ellipsis_n_bar_2": {
+      "acc,none": 0.327,
+      "acc_stderr,none": 0.014842213153411162,
+      "alias": " - blimp_ellipsis_n_bar_2"
+    },
+    "blimp_existential_there_object_raising": {
+      "acc,none": 0.69,
+      "acc_stderr,none": 0.014632638658632766,
+      "alias": " - blimp_existential_there_object_raising"
+    },
+    "blimp_existential_there_quantifiers_1": {
+      "acc,none": 0.483,
+      "acc_stderr,none": 0.015810153729833274,
+      "alias": " - blimp_existential_there_quantifiers_1"
+    },
+    "blimp_existential_there_quantifiers_2": {
+      "acc,none": 0.833,
+      "acc_stderr,none": 0.011800434324644593,
+      "alias": " - blimp_existential_there_quantifiers_2"
+    },
+    "blimp_existential_there_subject_raising": {
+      "acc,none": 0.576,
+      "acc_stderr,none": 0.015635487471405186,
+      "alias": " - blimp_existential_there_subject_raising"
+    },
+    "blimp_expletive_it_object_raising": {
+      "acc,none": 0.589,
+      "acc_stderr,none": 0.01556667341859915,
+      "alias": " - blimp_expletive_it_object_raising"
+    },
+    "blimp_inchoative": {
+      "acc,none": 0.421,
+      "acc_stderr,none": 0.015620595475301285,
+      "alias": " - blimp_inchoative"
+    },
+    "blimp_intransitive": {
+      "acc,none": 0.554,
+      "acc_stderr,none": 0.01572677116675039,
+      "alias": " - blimp_intransitive"
+    },
+    "blimp_irregular_past_participle_adjectives": {
+      "acc,none": 0.533,
+      "acc_stderr,none": 0.015784807891138876,
+      "alias": " - blimp_irregular_past_participle_adjectives"
+    },
+    "blimp_irregular_past_participle_verbs": {
+      "acc,none": 0.391,
+      "acc_stderr,none": 0.015438826294681775,
+      "alias": " - blimp_irregular_past_participle_verbs"
+    },
+    "blimp_irregular_plural_subject_verb_agreement_1": {
+      "acc,none": 0.528,
+      "acc_stderr,none": 0.015794475789511517,
+      "alias": " - blimp_irregular_plural_subject_verb_agreement_1"
+    },
+    "blimp_irregular_plural_subject_verb_agreement_2": {
+      "acc,none": 0.536,
+      "acc_stderr,none": 0.015778243024904673,
+      "alias": " - blimp_irregular_plural_subject_verb_agreement_2"
+    },
+    "blimp_left_branch_island_echo_question": {
+      "acc,none": 0.61,
+      "acc_stderr,none": 0.01543172505386673,
+      "alias": " - blimp_left_branch_island_echo_question"
+    },
+    "blimp_left_branch_island_simple_question": {
+      "acc,none": 0.516,
+      "acc_stderr,none": 0.015811198373114985,
+      "alias": " - blimp_left_branch_island_simple_question"
+    },
+    "blimp_matrix_question_npi_licensor_present": {
+      "acc,none": 0.596,
+      "acc_stderr,none": 0.015524980677122437,
+      "alias": " - blimp_matrix_question_npi_licensor_present"
+    },
+    "blimp_npi_present_1": {
+      "acc,none": 0.566,
+      "acc_stderr,none": 0.01568087656637504,
+      "alias": " - blimp_npi_present_1"
+    },
+    "blimp_npi_present_2": {
+      "acc,none": 0.586,
+      "acc_stderr,none": 0.01558354410417755,
+      "alias": " - blimp_npi_present_2"
+    },
+    "blimp_only_npi_licensor_present": {
+      "acc,none": 0.209,
+      "acc_stderr,none": 0.012864077288499339,
+      "alias": " - blimp_only_npi_licensor_present"
+    },
+    "blimp_only_npi_scope": {
+      "acc,none": 0.603,
+      "acc_stderr,none": 0.015480007449307982,
+      "alias": " - blimp_only_npi_scope"
+    },
+    "blimp_passive_1": {
+      "acc,none": 0.637,
+      "acc_stderr,none": 0.01521389044467136,
+      "alias": " - blimp_passive_1"
+    },
+    "blimp_passive_2": {
+      "acc,none": 0.652,
+      "acc_stderr,none": 0.015070604603768328,
+      "alias": " - blimp_passive_2"
+    },
+    "blimp_principle_A_c_command": {
+      "acc,none": 0.629,
+      "acc_stderr,none": 0.015283736211823096,
+      "alias": " - blimp_principle_A_c_command"
+    },
+    "blimp_principle_A_case_1": {
+      "acc,none": 0.31,
+      "acc_stderr,none": 0.014632638658632765,
+      "alias": " - blimp_principle_A_case_1"
+    },
+    "blimp_principle_A_case_2": {
+      "acc,none": 0.508,
+      "acc_stderr,none": 0.015817274929209084,
+      "alias": " - blimp_principle_A_case_2"
+    },
+    "blimp_principle_A_domain_1": {
+      "acc,none": 0.605,
+      "acc_stderr,none": 0.015466551464829328,
+      "alias": " - blimp_principle_A_domain_1"
+    },
+    "blimp_principle_A_domain_2": {
+      "acc,none": 0.517,
+      "acc_stderr,none": 0.015810153729833274,
+      "alias": " - blimp_principle_A_domain_2"
+    },
+    "blimp_principle_A_domain_3": {
+      "acc,none": 0.505,
+      "acc_stderr,none": 0.015818508944436743,
+      "alias": " - blimp_principle_A_domain_3"
+    },
+    "blimp_principle_A_reconstruction": {
+      "acc,none": 0.493,
+      "acc_stderr,none": 0.01581774956184353,
+      "alias": " - blimp_principle_A_reconstruction"
+    },
+    "blimp_regular_plural_subject_verb_agreement_1": {
+      "acc,none": 0.442,
+      "acc_stderr,none": 0.015712507211864152,
+      "alias": " - blimp_regular_plural_subject_verb_agreement_1"
+    },
+    "blimp_regular_plural_subject_verb_agreement_2": {
+      "acc,none": 0.513,
+      "acc_stderr,none": 0.015813952101896636,
+      "alias": " - blimp_regular_plural_subject_verb_agreement_2"
+    },
+    "blimp_sentential_negation_npi_licensor_present": {
+      "acc,none": 0.655,
+      "acc_stderr,none": 0.015039986742055365,
+      "alias": " - blimp_sentential_negation_npi_licensor_present"
+    },
+    "blimp_sentential_negation_npi_scope": {
+      "acc,none": 0.398,
+      "acc_stderr,none": 0.015486634102859016,
+      "alias": " - blimp_sentential_negation_npi_scope"
+    },
+    "blimp_sentential_subject_island": {
+      "acc,none": 0.51,
+      "acc_stderr,none": 0.015816135752773183,
+      "alias": " - blimp_sentential_subject_island"
+    },
+    "blimp_superlative_quantifiers_1": {
+      "acc,none": 0.348,
+      "acc_stderr,none": 0.015070604603768328,
+      "alias": " - blimp_superlative_quantifiers_1"
+    },
+    "blimp_superlative_quantifiers_2": {
+      "acc,none": 0.371,
+      "acc_stderr,none": 0.015283736211823096,
+      "alias": " - blimp_superlative_quantifiers_2"
+    },
+    "blimp_tough_vs_raising_1": {
+      "acc,none": 0.446,
+      "acc_stderr,none": 0.01572677116675039,
+      "alias": " - blimp_tough_vs_raising_1"
+    },
+    "blimp_tough_vs_raising_2": {
+      "acc,none": 0.566,
+      "acc_stderr,none": 0.01568087656637504,
+      "alias": " - blimp_tough_vs_raising_2"
+    },
+    "blimp_transitive": {
+      "acc,none": 0.515,
+      "acc_stderr,none": 0.01581217964181488,
+      "alias": " - blimp_transitive"
+    },
+    "blimp_wh_island": {
+      "acc,none": 0.432,
+      "acc_stderr,none": 0.015672320237336324,
+      "alias": " - blimp_wh_island"
+    },
+    "blimp_wh_questions_object_gap": {
+      "acc,none": 0.364,
+      "acc_stderr,none": 0.015222868840522005,
+      "alias": " - blimp_wh_questions_object_gap"
+    },
+    "blimp_wh_questions_subject_gap": {
+      "acc,none": 0.373,
+      "acc_stderr,none": 0.015300493622922927,
+      "alias": " - blimp_wh_questions_subject_gap"
+    },
+    "blimp_wh_questions_subject_gap_long_distance": {
+      "acc,none": 0.391,
+      "acc_stderr,none": 0.015438826294681775,
+      "alias": " - blimp_wh_questions_subject_gap_long_distance"
+    },
+    "blimp_wh_vs_that_no_gap": {
+      "acc,none": 0.317,
+      "acc_stderr,none": 0.014721675438880174,
+      "alias": " - blimp_wh_vs_that_no_gap"
+    },
+    "blimp_wh_vs_that_no_gap_long_distance": {
+      "acc,none": 0.346,
+      "acc_stderr,none": 0.01505026612756434,
+      "alias": " - blimp_wh_vs_that_no_gap_long_distance"
+    },
+    "blimp_wh_vs_that_with_gap": {
+      "acc,none": 0.687,
+      "acc_stderr,none": 0.014671272822977775,
+      "alias": " - blimp_wh_vs_that_with_gap"
+    },
+    "blimp_wh_vs_that_with_gap_long_distance": {
+      "acc,none": 0.67,
+      "acc_stderr,none": 0.014876872027456623,
+      "alias": " - blimp_wh_vs_that_with_gap_long_distance"
+    }
+  },
+  "groups": {
+    "blimp": {
+      "acc,none": 0.5175074626865671,
+      "acc_stderr,none": 0.001882002716920592,
+      "alias": "blimp"
+    }
+  },
+  "group_subtasks": {
+    "blimp": [
+      "blimp_causative",
+      "blimp_determiner_noun_agreement_irregular_2",
+      "blimp_passive_1",
+      "blimp_wh_questions_subject_gap",
+      "blimp_irregular_plural_subject_verb_agreement_1",
+      "blimp_wh_vs_that_no_gap",
+      "blimp_complex_NP_island",
+      "blimp_principle_A_reconstruction",
+      "blimp_anaphor_number_agreement",
+      "blimp_principle_A_domain_3",
+      "blimp_anaphor_gender_agreement",
+      "blimp_irregular_past_participle_adjectives",
+      "blimp_regular_plural_subject_verb_agreement_1",
+      "blimp_sentential_subject_island",
+      "blimp_left_branch_island_simple_question",
+      "blimp_distractor_agreement_relative_clause",
+      "blimp_irregular_plural_subject_verb_agreement_2",
+      "blimp_left_branch_island_echo_question",
+      "blimp_wh_vs_that_no_gap_long_distance",
+      "blimp_ellipsis_n_bar_1",
+      "blimp_determiner_noun_agreement_1",
+      "blimp_superlative_quantifiers_2",
+      "blimp_ellipsis_n_bar_2",
+      "blimp_adjunct_island",
+      "blimp_expletive_it_object_raising",
+      "blimp_principle_A_c_command",
+      "blimp_principle_A_domain_1",
+      "blimp_superlative_quantifiers_1",
+      "blimp_only_npi_scope",
+      "blimp_tough_vs_raising_1",
+      "blimp_transitive",
+      "blimp_intransitive",
+      "blimp_wh_questions_object_gap",
+      "blimp_only_npi_licensor_present",
+      "blimp_wh_vs_that_with_gap",
+      "blimp_existential_there_quantifiers_1",
+      "blimp_coordinate_structure_constraint_complex_left_branch",
+      "blimp_principle_A_domain_2",
+      "blimp_determiner_noun_agreement_with_adj_2",
+      "blimp_wh_island",
+      "blimp_existential_there_quantifiers_2",
+      "blimp_existential_there_object_raising",
+      "blimp_determiner_noun_agreement_with_adj_irregular_1",
+      "blimp_npi_present_2",
+      "blimp_animate_subject_passive",
+      "blimp_wh_questions_subject_gap_long_distance",
+      "blimp_existential_there_subject_raising",
+      "blimp_principle_A_case_2",
+      "blimp_principle_A_case_1",
+      "blimp_wh_vs_that_with_gap_long_distance",
+      "blimp_determiner_noun_agreement_irregular_1",
+      "blimp_drop_argument",
+      "blimp_sentential_negation_npi_scope",
+      "blimp_coordinate_structure_constraint_object_extraction",
+      "blimp_determiner_noun_agreement_2",
+      "blimp_irregular_past_participle_verbs",
+      "blimp_tough_vs_raising_2",
+      "blimp_sentential_negation_npi_licensor_present",
+      "blimp_distractor_agreement_relational_noun",
+      "blimp_passive_2",
+      "blimp_determiner_noun_agreement_with_adjective_1",
+      "blimp_inchoative",
+      "blimp_animate_subject_trans",
+      "blimp_regular_plural_subject_verb_agreement_2",
+      "blimp_npi_present_1",
+      "blimp_determiner_noun_agreement_with_adj_irregular_2",
+      "blimp_matrix_question_npi_licensor_present"
+    ]
+  },
+  "configs": {
+    "blimp_adjunct_island": {
+      "task": "blimp_adjunct_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "adjunct_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_anaphor_gender_agreement": {
+      "task": "blimp_anaphor_gender_agreement",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "anaphor_gender_agreement",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_anaphor_number_agreement": {
+      "task": "blimp_anaphor_number_agreement",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "anaphor_number_agreement",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_animate_subject_passive": {
+      "task": "blimp_animate_subject_passive",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "animate_subject_passive",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_animate_subject_trans": {
+      "task": "blimp_animate_subject_trans",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "animate_subject_trans",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_causative": {
+      "task": "blimp_causative",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "causative",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_complex_NP_island": {
+      "task": "blimp_complex_NP_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "complex_NP_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_coordinate_structure_constraint_complex_left_branch": {
+      "task": "blimp_coordinate_structure_constraint_complex_left_branch",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "coordinate_structure_constraint_complex_left_branch",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_coordinate_structure_constraint_object_extraction": {
+      "task": "blimp_coordinate_structure_constraint_object_extraction",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "coordinate_structure_constraint_object_extraction",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_1": {
+      "task": "blimp_determiner_noun_agreement_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_2": {
+      "task": "blimp_determiner_noun_agreement_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_irregular_1": {
+      "task": "blimp_determiner_noun_agreement_irregular_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_irregular_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_irregular_2": {
+      "task": "blimp_determiner_noun_agreement_irregular_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_irregular_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adj_2": {
+      "task": "blimp_determiner_noun_agreement_with_adj_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adj_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": {
+      "task": "blimp_determiner_noun_agreement_with_adj_irregular_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adj_irregular_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": {
+      "task": "blimp_determiner_noun_agreement_with_adj_irregular_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adj_irregular_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adjective_1": {
+      "task": "blimp_determiner_noun_agreement_with_adjective_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adjective_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_distractor_agreement_relational_noun": {
+      "task": "blimp_distractor_agreement_relational_noun",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "distractor_agreement_relational_noun",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_distractor_agreement_relative_clause": {
+      "task": "blimp_distractor_agreement_relative_clause",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "distractor_agreement_relative_clause",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_drop_argument": {
+      "task": "blimp_drop_argument",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "drop_argument",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_ellipsis_n_bar_1": {
+      "task": "blimp_ellipsis_n_bar_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "ellipsis_n_bar_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_ellipsis_n_bar_2": {
+      "task": "blimp_ellipsis_n_bar_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "ellipsis_n_bar_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_object_raising": {
+      "task": "blimp_existential_there_object_raising",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_object_raising",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_quantifiers_1": {
+      "task": "blimp_existential_there_quantifiers_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_quantifiers_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_quantifiers_2": {
+      "task": "blimp_existential_there_quantifiers_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_quantifiers_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_subject_raising": {
+      "task": "blimp_existential_there_subject_raising",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_subject_raising",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_expletive_it_object_raising": {
+      "task": "blimp_expletive_it_object_raising",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "expletive_it_object_raising",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_inchoative": {
+      "task": "blimp_inchoative",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "inchoative",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_intransitive": {
+      "task": "blimp_intransitive",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "intransitive",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_past_participle_adjectives": {
+      "task": "blimp_irregular_past_participle_adjectives",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_past_participle_adjectives",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_past_participle_verbs": {
+      "task": "blimp_irregular_past_participle_verbs",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_past_participle_verbs",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_plural_subject_verb_agreement_1": {
+      "task": "blimp_irregular_plural_subject_verb_agreement_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_plural_subject_verb_agreement_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_plural_subject_verb_agreement_2": {
+      "task": "blimp_irregular_plural_subject_verb_agreement_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_plural_subject_verb_agreement_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_left_branch_island_echo_question": {
+      "task": "blimp_left_branch_island_echo_question",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "left_branch_island_echo_question",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_left_branch_island_simple_question": {
+      "task": "blimp_left_branch_island_simple_question",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "left_branch_island_simple_question",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_matrix_question_npi_licensor_present": {
+      "task": "blimp_matrix_question_npi_licensor_present",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "matrix_question_npi_licensor_present",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_npi_present_1": {
+      "task": "blimp_npi_present_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "npi_present_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_npi_present_2": {
+      "task": "blimp_npi_present_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "npi_present_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_only_npi_licensor_present": {
+      "task": "blimp_only_npi_licensor_present",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "only_npi_licensor_present",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_only_npi_scope": {
+      "task": "blimp_only_npi_scope",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "only_npi_scope",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_passive_1": {
+      "task": "blimp_passive_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "passive_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_passive_2": {
+      "task": "blimp_passive_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "passive_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_c_command": {
+      "task": "blimp_principle_A_c_command",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_c_command",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_case_1": {
+      "task": "blimp_principle_A_case_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_case_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_case_2": {
+      "task": "blimp_principle_A_case_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_case_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_domain_1": {
+      "task": "blimp_principle_A_domain_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_domain_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_domain_2": {
+      "task": "blimp_principle_A_domain_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_domain_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_domain_3": {
+      "task": "blimp_principle_A_domain_3",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_domain_3",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_reconstruction": {
+      "task": "blimp_principle_A_reconstruction",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_reconstruction",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_regular_plural_subject_verb_agreement_1": {
+      "task": "blimp_regular_plural_subject_verb_agreement_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "regular_plural_subject_verb_agreement_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_regular_plural_subject_verb_agreement_2": {
+      "task": "blimp_regular_plural_subject_verb_agreement_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "regular_plural_subject_verb_agreement_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_sentential_negation_npi_licensor_present": {
+      "task": "blimp_sentential_negation_npi_licensor_present",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "sentential_negation_npi_licensor_present",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_sentential_negation_npi_scope": {
+      "task": "blimp_sentential_negation_npi_scope",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "sentential_negation_npi_scope",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_sentential_subject_island": {
+      "task": "blimp_sentential_subject_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "sentential_subject_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_superlative_quantifiers_1": {
+      "task": "blimp_superlative_quantifiers_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "superlative_quantifiers_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_superlative_quantifiers_2": {
+      "task": "blimp_superlative_quantifiers_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "superlative_quantifiers_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_tough_vs_raising_1": {
+      "task": "blimp_tough_vs_raising_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "tough_vs_raising_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_tough_vs_raising_2": {
+      "task": "blimp_tough_vs_raising_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "tough_vs_raising_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_transitive": {
+      "task": "blimp_transitive",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "transitive",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_island": {
+      "task": "blimp_wh_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_questions_object_gap": {
+      "task": "blimp_wh_questions_object_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_questions_object_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_questions_subject_gap": {
+      "task": "blimp_wh_questions_subject_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_questions_subject_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_questions_subject_gap_long_distance": {
+      "task": "blimp_wh_questions_subject_gap_long_distance",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_questions_subject_gap_long_distance",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_no_gap": {
+      "task": "blimp_wh_vs_that_no_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_no_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_no_gap_long_distance": {
+      "task": "blimp_wh_vs_that_no_gap_long_distance",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_no_gap_long_distance",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_with_gap": {
+      "task": "blimp_wh_vs_that_with_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_with_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_with_gap_long_distance": {
+      "task": "blimp_wh_vs_that_with_gap_long_distance",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_with_gap_long_distance",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "blimp_adjunct_island": 1.0,
+    "blimp_anaphor_gender_agreement": 1.0,
+    "blimp_anaphor_number_agreement": 1.0,
+    "blimp_animate_subject_passive": 1.0,
+    "blimp_animate_subject_trans": 1.0,
+    "blimp_causative": 1.0,
+    "blimp_complex_NP_island": 1.0,
+    "blimp_coordinate_structure_constraint_complex_left_branch": 1.0,
+    "blimp_coordinate_structure_constraint_object_extraction": 1.0,
+    "blimp_determiner_noun_agreement_1": 1.0,
+    "blimp_determiner_noun_agreement_2": 1.0,
+    "blimp_determiner_noun_agreement_irregular_1": 1.0,
+    "blimp_determiner_noun_agreement_irregular_2": 1.0,
+    "blimp_determiner_noun_agreement_with_adj_2": 1.0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": 1.0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": 1.0,
+    "blimp_determiner_noun_agreement_with_adjective_1": 1.0,
+    "blimp_distractor_agreement_relational_noun": 1.0,
+    "blimp_distractor_agreement_relative_clause": 1.0,
+    "blimp_drop_argument": 1.0,
+    "blimp_ellipsis_n_bar_1": 1.0,
+    "blimp_ellipsis_n_bar_2": 1.0,
+    "blimp_existential_there_object_raising": 1.0,
+    "blimp_existential_there_quantifiers_1": 1.0,
+    "blimp_existential_there_quantifiers_2": 1.0,
+    "blimp_existential_there_subject_raising": 1.0,
+    "blimp_expletive_it_object_raising": 1.0,
+    "blimp_inchoative": 1.0,
+    "blimp_intransitive": 1.0,
+    "blimp_irregular_past_participle_adjectives": 1.0,
+    "blimp_irregular_past_participle_verbs": 1.0,
+    "blimp_irregular_plural_subject_verb_agreement_1": 1.0,
+    "blimp_irregular_plural_subject_verb_agreement_2": 1.0,
+    "blimp_left_branch_island_echo_question": 1.0,
+    "blimp_left_branch_island_simple_question": 1.0,
+    "blimp_matrix_question_npi_licensor_present": 1.0,
+    "blimp_npi_present_1": 1.0,
+    "blimp_npi_present_2": 1.0,
+    "blimp_only_npi_licensor_present": 1.0,
+    "blimp_only_npi_scope": 1.0,
+    "blimp_passive_1": 1.0,
+    "blimp_passive_2": 1.0,
+    "blimp_principle_A_c_command": 1.0,
+    "blimp_principle_A_case_1": 1.0,
+    "blimp_principle_A_case_2": 1.0,
+    "blimp_principle_A_domain_1": 1.0,
+    "blimp_principle_A_domain_2": 1.0,
+    "blimp_principle_A_domain_3": 1.0,
+    "blimp_principle_A_reconstruction": 1.0,
+    "blimp_regular_plural_subject_verb_agreement_1": 1.0,
+    "blimp_regular_plural_subject_verb_agreement_2": 1.0,
+    "blimp_sentential_negation_npi_licensor_present": 1.0,
+    "blimp_sentential_negation_npi_scope": 1.0,
+    "blimp_sentential_subject_island": 1.0,
+    "blimp_superlative_quantifiers_1": 1.0,
+    "blimp_superlative_quantifiers_2": 1.0,
+    "blimp_tough_vs_raising_1": 1.0,
+    "blimp_tough_vs_raising_2": 1.0,
+    "blimp_transitive": 1.0,
+    "blimp_wh_island": 1.0,
+    "blimp_wh_questions_object_gap": 1.0,
+    "blimp_wh_questions_subject_gap": 1.0,
+    "blimp_wh_questions_subject_gap_long_distance": 1.0,
+    "blimp_wh_vs_that_no_gap": 1.0,
+    "blimp_wh_vs_that_no_gap_long_distance": 1.0,
+    "blimp_wh_vs_that_with_gap": 1.0,
+    "blimp_wh_vs_that_with_gap_long_distance": 1.0
+  },
+  "n-shot": {
+    "blimp": 0,
+    "blimp_adjunct_island": 0,
+    "blimp_anaphor_gender_agreement": 0,
+    "blimp_anaphor_number_agreement": 0,
+    "blimp_animate_subject_passive": 0,
+    "blimp_animate_subject_trans": 0,
+    "blimp_causative": 0,
+    "blimp_complex_NP_island": 0,
+    "blimp_coordinate_structure_constraint_complex_left_branch": 0,
+    "blimp_coordinate_structure_constraint_object_extraction": 0,
+    "blimp_determiner_noun_agreement_1": 0,
+    "blimp_determiner_noun_agreement_2": 0,
+    "blimp_determiner_noun_agreement_irregular_1": 0,
+    "blimp_determiner_noun_agreement_irregular_2": 0,
+    "blimp_determiner_noun_agreement_with_adj_2": 0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": 0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": 0,
+    "blimp_determiner_noun_agreement_with_adjective_1": 0,
+    "blimp_distractor_agreement_relational_noun": 0,
+    "blimp_distractor_agreement_relative_clause": 0,
+    "blimp_drop_argument": 0,
+    "blimp_ellipsis_n_bar_1": 0,
+    "blimp_ellipsis_n_bar_2": 0,
+    "blimp_existential_there_object_raising": 0,
+    "blimp_existential_there_quantifiers_1": 0,
+    "blimp_existential_there_quantifiers_2": 0,
+    "blimp_existential_there_subject_raising": 0,
+    "blimp_expletive_it_object_raising": 0,
+    "blimp_inchoative": 0,
+    "blimp_intransitive": 0,
+    "blimp_irregular_past_participle_adjectives": 0,
+    "blimp_irregular_past_participle_verbs": 0,
+    "blimp_irregular_plural_subject_verb_agreement_1": 0,
+    "blimp_irregular_plural_subject_verb_agreement_2": 0,
+    "blimp_left_branch_island_echo_question": 0,
+    "blimp_left_branch_island_simple_question": 0,
+    "blimp_matrix_question_npi_licensor_present": 0,
+    "blimp_npi_present_1": 0,
+    "blimp_npi_present_2": 0,
+    "blimp_only_npi_licensor_present": 0,
+    "blimp_only_npi_scope": 0,
+    "blimp_passive_1": 0,
+    "blimp_passive_2": 0,
+    "blimp_principle_A_c_command": 0,
+    "blimp_principle_A_case_1": 0,
+    "blimp_principle_A_case_2": 0,
+    "blimp_principle_A_domain_1": 0,
+    "blimp_principle_A_domain_2": 0,
+    "blimp_principle_A_domain_3": 0,
+    "blimp_principle_A_reconstruction": 0,
+    "blimp_regular_plural_subject_verb_agreement_1": 0,
+    "blimp_regular_plural_subject_verb_agreement_2": 0,
+    "blimp_sentential_negation_npi_licensor_present": 0,
+    "blimp_sentential_negation_npi_scope": 0,
+    "blimp_sentential_subject_island": 0,
+    "blimp_superlative_quantifiers_1": 0,
+    "blimp_superlative_quantifiers_2": 0,
+    "blimp_tough_vs_raising_1": 0,
+    "blimp_tough_vs_raising_2": 0,
+    "blimp_transitive": 0,
+    "blimp_wh_island": 0,
+    "blimp_wh_questions_object_gap": 0,
+    "blimp_wh_questions_subject_gap": 0,
+    "blimp_wh_questions_subject_gap_long_distance": 0,
+    "blimp_wh_vs_that_no_gap": 0,
+    "blimp_wh_vs_that_no_gap_long_distance": 0,
+    "blimp_wh_vs_that_with_gap": 0,
+    "blimp_wh_vs_that_with_gap_long_distance": 0
+  },
+  "n-samples": {
+    "blimp_causative": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_irregular_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_passive_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_questions_subject_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_plural_subject_verb_agreement_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_no_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_complex_NP_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_reconstruction": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_anaphor_number_agreement": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_domain_3": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_anaphor_gender_agreement": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_past_participle_adjectives": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_regular_plural_subject_verb_agreement_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_sentential_subject_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_left_branch_island_simple_question": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_distractor_agreement_relative_clause": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_plural_subject_verb_agreement_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_left_branch_island_echo_question": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_no_gap_long_distance": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_ellipsis_n_bar_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_superlative_quantifiers_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_ellipsis_n_bar_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_adjunct_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_expletive_it_object_raising": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_c_command": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_domain_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_superlative_quantifiers_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_only_npi_scope": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_tough_vs_raising_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_transitive": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_intransitive": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_questions_object_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_only_npi_licensor_present": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_with_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_quantifiers_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_coordinate_structure_constraint_complex_left_branch": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_domain_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adj_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_quantifiers_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_object_raising": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_npi_present_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_animate_subject_passive": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_questions_subject_gap_long_distance": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_subject_raising": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_case_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_case_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_with_gap_long_distance": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_irregular_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_drop_argument": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_sentential_negation_npi_scope": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_coordinate_structure_constraint_object_extraction": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_past_participle_verbs": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_tough_vs_raising_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_sentential_negation_npi_licensor_present": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_distractor_agreement_relational_noun": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_passive_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adjective_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_inchoative": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_animate_subject_trans": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_regular_plural_subject_verb_agreement_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_npi_present_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_matrix_question_npi_licensor_present": {
+      "original": 1000,
+      "effective": 1000
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step0",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step0",
+    "model_sha": "322fbe766b2a11f41f91b60d93bb282da3b1d236",
+    "batch_size": "1024",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1720210700.1683106,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.114.2.el7.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.54.15\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              5000.11\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 6716623.672507783,
+  "end_time": 6717037.494629572,
+  "total_evaluation_time_seconds": "413.8221217887476"
+}

pythia-14m-seed1/step0/EleutherAI__pythia-14m-seed1/results_2024-07-11T13-45-02.042989.json.bak

@@ -0,0 +1,2606 @@
+{
+  "results": {
+    "blimp": {
+      "acc,none": 0.5176865671641792,
+      "acc_stderr,none": 0.0018818593441844549,
+      "alias": "blimp"
+    },
+    "blimp_adjunct_island": {
+      "acc,none": 0.561,
+      "acc_stderr,none": 0.015701131345400736,
+      "alias": " - blimp_adjunct_island"
+    },
+    "blimp_anaphor_gender_agreement": {
+      "acc,none": 0.559,
+      "acc_stderr,none": 0.015708779894242766,
+      "alias": " - blimp_anaphor_gender_agreement"
+    },
+    "blimp_anaphor_number_agreement": {
+      "acc,none": 0.541,
+      "acc_stderr,none": 0.015766025737882314,
+      "alias": " - blimp_anaphor_number_agreement"
+    },
+    "blimp_animate_subject_passive": {
+      "acc,none": 0.593,
+      "acc_stderr,none": 0.015543249100255474,
+      "alias": " - blimp_animate_subject_passive"
+    },
+    "blimp_animate_subject_trans": {
+      "acc,none": 0.813,
+      "acc_stderr,none": 0.012336254828074168,
+      "alias": " - blimp_animate_subject_trans"
+    },
+    "blimp_causative": {
+      "acc,none": 0.456,
+      "acc_stderr,none": 0.01575792855397918,
+      "alias": " - blimp_causative"
+    },
+    "blimp_complex_NP_island": {
+      "acc,none": 0.509,
+      "acc_stderr,none": 0.0158167369950053,
+      "alias": " - blimp_complex_NP_island"
+    },
+    "blimp_coordinate_structure_constraint_complex_left_branch": {
+      "acc,none": 0.547,
+      "acc_stderr,none": 0.015749255189977687,
+      "alias": " - blimp_coordinate_structure_constraint_complex_left_branch"
+    },
+    "blimp_coordinate_structure_constraint_object_extraction": {
+      "acc,none": 0.531,
+      "acc_stderr,none": 0.015788865959538965,
+      "alias": " - blimp_coordinate_structure_constraint_object_extraction"
+    },
+    "blimp_determiner_noun_agreement_1": {
+      "acc,none": 0.539,
+      "acc_stderr,none": 0.015771104201283047,
+      "alias": " - blimp_determiner_noun_agreement_1"
+    },
+    "blimp_determiner_noun_agreement_2": {
+      "acc,none": 0.531,
+      "acc_stderr,none": 0.015788865959538965,
+      "alias": " - blimp_determiner_noun_agreement_2"
+    },
+    "blimp_determiner_noun_agreement_irregular_1": {
+      "acc,none": 0.498,
+      "acc_stderr,none": 0.01581917337430266,
+      "alias": " - blimp_determiner_noun_agreement_irregular_1"
+    },
+    "blimp_determiner_noun_agreement_irregular_2": {
+      "acc,none": 0.474,
+      "acc_stderr,none": 0.015797897758042797,
+      "alias": " - blimp_determiner_noun_agreement_irregular_2"
+    },
+    "blimp_determiner_noun_agreement_with_adj_2": {
+      "acc,none": 0.54,
+      "acc_stderr,none": 0.0157685969143944,
+      "alias": " - blimp_determiner_noun_agreement_with_adj_2"
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": {
+      "acc,none": 0.541,
+      "acc_stderr,none": 0.015766025737882314,
+      "alias": " - blimp_determiner_noun_agreement_with_adj_irregular_1"
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": {
+      "acc,none": 0.522,
+      "acc_stderr,none": 0.01580397942816194,
+      "alias": " - blimp_determiner_noun_agreement_with_adj_irregular_2"
+    },
+    "blimp_determiner_noun_agreement_with_adjective_1": {
+      "acc,none": 0.499,
+      "acc_stderr,none": 0.015819268290576817,
+      "alias": " - blimp_determiner_noun_agreement_with_adjective_1"
+    },
+    "blimp_distractor_agreement_relational_noun": {
+      "acc,none": 0.478,
+      "acc_stderr,none": 0.01580397942816194,
+      "alias": " - blimp_distractor_agreement_relational_noun"
+    },
+    "blimp_distractor_agreement_relative_clause": {
+      "acc,none": 0.509,
+      "acc_stderr,none": 0.0158167369950053,
+      "alias": " - blimp_distractor_agreement_relative_clause"
+    },
+    "blimp_drop_argument": {
+      "acc,none": 0.663,
+      "acc_stderr,none": 0.01495508791865346,
+      "alias": " - blimp_drop_argument"
+    },
+    "blimp_ellipsis_n_bar_1": {
+      "acc,none": 0.431,
+      "acc_stderr,none": 0.01566794448817344,
+      "alias": " - blimp_ellipsis_n_bar_1"
+    },
+    "blimp_ellipsis_n_bar_2": {
+      "acc,none": 0.327,
+      "acc_stderr,none": 0.014842213153411162,
+      "alias": " - blimp_ellipsis_n_bar_2"
+    },
+    "blimp_existential_there_object_raising": {
+      "acc,none": 0.691,
+      "acc_stderr,none": 0.014619600977206347,
+      "alias": " - blimp_existential_there_object_raising"
+    },
+    "blimp_existential_there_quantifiers_1": {
+      "acc,none": 0.485,
+      "acc_stderr,none": 0.01581217964181488,
+      "alias": " - blimp_existential_there_quantifiers_1"
+    },
+    "blimp_existential_there_quantifiers_2": {
+      "acc,none": 0.833,
+      "acc_stderr,none": 0.011800434324644593,
+      "alias": " - blimp_existential_there_quantifiers_2"
+    },
+    "blimp_existential_there_subject_raising": {
+      "acc,none": 0.575,
+      "acc_stderr,none": 0.01564032031704017,
+      "alias": " - blimp_existential_there_subject_raising"
+    },
+    "blimp_expletive_it_object_raising": {
+      "acc,none": 0.59,
+      "acc_stderr,none": 0.015560917136921659,
+      "alias": " - blimp_expletive_it_object_raising"
+    },
+    "blimp_inchoative": {
+      "acc,none": 0.423,
+      "acc_stderr,none": 0.015630589090476255,
+      "alias": " - blimp_inchoative"
+    },
+    "blimp_intransitive": {
+      "acc,none": 0.555,
+      "acc_stderr,none": 0.015723301886761007,
+      "alias": " - blimp_intransitive"
+    },
+    "blimp_irregular_past_participle_adjectives": {
+      "acc,none": 0.531,
+      "acc_stderr,none": 0.015788865959538965,
+      "alias": " - blimp_irregular_past_participle_adjectives"
+    },
+    "blimp_irregular_past_participle_verbs": {
+      "acc,none": 0.391,
+      "acc_stderr,none": 0.015438826294681775,
+      "alias": " - blimp_irregular_past_participle_verbs"
+    },
+    "blimp_irregular_plural_subject_verb_agreement_1": {
+      "acc,none": 0.527,
+      "acc_stderr,none": 0.01579621855130273,
+      "alias": " - blimp_irregular_plural_subject_verb_agreement_1"
+    },
+    "blimp_irregular_plural_subject_verb_agreement_2": {
+      "acc,none": 0.535,
+      "acc_stderr,none": 0.015780495050030086,
+      "alias": " - blimp_irregular_plural_subject_verb_agreement_2"
+    },
+    "blimp_left_branch_island_echo_question": {
+      "acc,none": 0.609,
+      "acc_stderr,none": 0.015438826294681775,
+      "alias": " - blimp_left_branch_island_echo_question"
+    },
+    "blimp_left_branch_island_simple_question": {
+      "acc,none": 0.514,
+      "acc_stderr,none": 0.01581309754773093,
+      "alias": " - blimp_left_branch_island_simple_question"
+    },
+    "blimp_matrix_question_npi_licensor_present": {
+      "acc,none": 0.596,
+      "acc_stderr,none": 0.015524980677122437,
+      "alias": " - blimp_matrix_question_npi_licensor_present"
+    },
+    "blimp_npi_present_1": {
+      "acc,none": 0.567,
+      "acc_stderr,none": 0.01567663091218125,
+      "alias": " - blimp_npi_present_1"
+    },
+    "blimp_npi_present_2": {
+      "acc,none": 0.587,
+      "acc_stderr,none": 0.015577986829936457,
+      "alias": " - blimp_npi_present_2"
+    },
+    "blimp_only_npi_licensor_present": {
+      "acc,none": 0.209,
+      "acc_stderr,none": 0.012864077288499339,
+      "alias": " - blimp_only_npi_licensor_present"
+    },
+    "blimp_only_npi_scope": {
+      "acc,none": 0.601,
+      "acc_stderr,none": 0.015493193313163012,
+      "alias": " - blimp_only_npi_scope"
+    },
+    "blimp_passive_1": {
+      "acc,none": 0.637,
+      "acc_stderr,none": 0.01521389044467136,
+      "alias": " - blimp_passive_1"
+    },
+    "blimp_passive_2": {
+      "acc,none": 0.651,
+      "acc_stderr,none": 0.015080663991563052,
+      "alias": " - blimp_passive_2"
+    },
+    "blimp_principle_A_c_command": {
+      "acc,none": 0.629,
+      "acc_stderr,none": 0.015283736211823096,
+      "alias": " - blimp_principle_A_c_command"
+    },
+    "blimp_principle_A_case_1": {
+      "acc,none": 0.308,
+      "acc_stderr,none": 0.01460648312734278,
+      "alias": " - blimp_principle_A_case_1"
+    },
+    "blimp_principle_A_case_2": {
+      "acc,none": 0.509,
+      "acc_stderr,none": 0.0158167369950053,
+      "alias": " - blimp_principle_A_case_2"
+    },
+    "blimp_principle_A_domain_1": {
+      "acc,none": 0.605,
+      "acc_stderr,none": 0.015466551464829328,
+      "alias": " - blimp_principle_A_domain_1"
+    },
+    "blimp_principle_A_domain_2": {
+      "acc,none": 0.518,
+      "acc_stderr,none": 0.01580904569940659,
+      "alias": " - blimp_principle_A_domain_2"
+    },
+    "blimp_principle_A_domain_3": {
+      "acc,none": 0.504,
+      "acc_stderr,none": 0.01581879370351084,
+      "alias": " - blimp_principle_A_domain_3"
+    },
+    "blimp_principle_A_reconstruction": {
+      "acc,none": 0.494,
+      "acc_stderr,none": 0.015818160898606836,
+      "alias": " - blimp_principle_A_reconstruction"
+    },
+    "blimp_regular_plural_subject_verb_agreement_1": {
+      "acc,none": 0.441,
+      "acc_stderr,none": 0.015708779894242766,
+      "alias": " - blimp_regular_plural_subject_verb_agreement_1"
+    },
+    "blimp_regular_plural_subject_verb_agreement_2": {
+      "acc,none": 0.515,
+      "acc_stderr,none": 0.01581217964181488,
+      "alias": " - blimp_regular_plural_subject_verb_agreement_2"
+    },
+    "blimp_sentential_negation_npi_licensor_present": {
+      "acc,none": 0.657,
+      "acc_stderr,none": 0.015019206922357005,
+      "alias": " - blimp_sentential_negation_npi_licensor_present"
+    },
+    "blimp_sentential_negation_npi_scope": {
+      "acc,none": 0.398,
+      "acc_stderr,none": 0.015486634102859016,
+      "alias": " - blimp_sentential_negation_npi_scope"
+    },
+    "blimp_sentential_subject_island": {
+      "acc,none": 0.511,
+      "acc_stderr,none": 0.01581547119529257,
+      "alias": " - blimp_sentential_subject_island"
+    },
+    "blimp_superlative_quantifiers_1": {
+      "acc,none": 0.349,
+      "acc_stderr,none": 0.015080663991563052,
+      "alias": " - blimp_superlative_quantifiers_1"
+    },
+    "blimp_superlative_quantifiers_2": {
+      "acc,none": 0.371,
+      "acc_stderr,none": 0.015283736211823096,
+      "alias": " - blimp_superlative_quantifiers_2"
+    },
+    "blimp_tough_vs_raising_1": {
+      "acc,none": 0.445,
+      "acc_stderr,none": 0.015723301886761007,
+      "alias": " - blimp_tough_vs_raising_1"
+    },
+    "blimp_tough_vs_raising_2": {
+      "acc,none": 0.566,
+      "acc_stderr,none": 0.01568087656637504,
+      "alias": " - blimp_tough_vs_raising_2"
+    },
+    "blimp_transitive": {
+      "acc,none": 0.515,
+      "acc_stderr,none": 0.01581217964181488,
+      "alias": " - blimp_transitive"
+    },
+    "blimp_wh_island": {
+      "acc,none": 0.433,
+      "acc_stderr,none": 0.01567663091218125,
+      "alias": " - blimp_wh_island"
+    },
+    "blimp_wh_questions_object_gap": {
+      "acc,none": 0.364,
+      "acc_stderr,none": 0.015222868840522005,
+      "alias": " - blimp_wh_questions_object_gap"
+    },
+    "blimp_wh_questions_subject_gap": {
+      "acc,none": 0.373,
+      "acc_stderr,none": 0.015300493622922927,
+      "alias": " - blimp_wh_questions_subject_gap"
+    },
+    "blimp_wh_questions_subject_gap_long_distance": {
+      "acc,none": 0.389,
+      "acc_stderr,none": 0.015424555647308514,
+      "alias": " - blimp_wh_questions_subject_gap_long_distance"
+    },
+    "blimp_wh_vs_that_no_gap": {
+      "acc,none": 0.315,
+      "acc_stderr,none": 0.014696631960792617,
+      "alias": " - blimp_wh_vs_that_no_gap"
+    },
+    "blimp_wh_vs_that_no_gap_long_distance": {
+      "acc,none": 0.348,
+      "acc_stderr,none": 0.015070604603768328,
+      "alias": " - blimp_wh_vs_that_no_gap_long_distance"
+    },
+    "blimp_wh_vs_that_with_gap": {
+      "acc,none": 0.688,
+      "acc_stderr,none": 0.014658474370509057,
+      "alias": " - blimp_wh_vs_that_with_gap"
+    },
+    "blimp_wh_vs_that_with_gap_long_distance": {
+      "acc,none": 0.671,
+      "acc_stderr,none": 0.014865395385928355,
+      "alias": " - blimp_wh_vs_that_with_gap_long_distance"
+    }
+  },
+  "groups": {
+    "blimp": {
+      "acc,none": 0.5176865671641792,
+      "acc_stderr,none": 0.0018818593441844549,
+      "alias": "blimp"
+    }
+  },
+  "group_subtasks": {
+    "blimp": [
+      "blimp_causative",
+      "blimp_determiner_noun_agreement_irregular_2",
+      "blimp_passive_1",
+      "blimp_wh_questions_subject_gap",
+      "blimp_irregular_plural_subject_verb_agreement_1",
+      "blimp_wh_vs_that_no_gap",
+      "blimp_complex_NP_island",
+      "blimp_principle_A_reconstruction",
+      "blimp_anaphor_number_agreement",
+      "blimp_principle_A_domain_3",
+      "blimp_anaphor_gender_agreement",
+      "blimp_irregular_past_participle_adjectives",
+      "blimp_regular_plural_subject_verb_agreement_1",
+      "blimp_sentential_subject_island",
+      "blimp_left_branch_island_simple_question",
+      "blimp_distractor_agreement_relative_clause",
+      "blimp_irregular_plural_subject_verb_agreement_2",
+      "blimp_left_branch_island_echo_question",
+      "blimp_wh_vs_that_no_gap_long_distance",
+      "blimp_ellipsis_n_bar_1",
+      "blimp_determiner_noun_agreement_1",
+      "blimp_superlative_quantifiers_2",
+      "blimp_ellipsis_n_bar_2",
+      "blimp_adjunct_island",
+      "blimp_expletive_it_object_raising",
+      "blimp_principle_A_c_command",
+      "blimp_principle_A_domain_1",
+      "blimp_superlative_quantifiers_1",
+      "blimp_only_npi_scope",
+      "blimp_tough_vs_raising_1",
+      "blimp_transitive",
+      "blimp_intransitive",
+      "blimp_wh_questions_object_gap",
+      "blimp_only_npi_licensor_present",
+      "blimp_wh_vs_that_with_gap",
+      "blimp_existential_there_quantifiers_1",
+      "blimp_coordinate_structure_constraint_complex_left_branch",
+      "blimp_principle_A_domain_2",
+      "blimp_determiner_noun_agreement_with_adj_2",
+      "blimp_wh_island",
+      "blimp_existential_there_quantifiers_2",
+      "blimp_existential_there_object_raising",
+      "blimp_determiner_noun_agreement_with_adj_irregular_1",
+      "blimp_npi_present_2",
+      "blimp_animate_subject_passive",
+      "blimp_wh_questions_subject_gap_long_distance",
+      "blimp_existential_there_subject_raising",
+      "blimp_principle_A_case_2",
+      "blimp_principle_A_case_1",
+      "blimp_wh_vs_that_with_gap_long_distance",
+      "blimp_determiner_noun_agreement_irregular_1",
+      "blimp_drop_argument",
+      "blimp_sentential_negation_npi_scope",
+      "blimp_coordinate_structure_constraint_object_extraction",
+      "blimp_determiner_noun_agreement_2",
+      "blimp_irregular_past_participle_verbs",
+      "blimp_tough_vs_raising_2",
+      "blimp_sentential_negation_npi_licensor_present",
+      "blimp_distractor_agreement_relational_noun",
+      "blimp_passive_2",
+      "blimp_determiner_noun_agreement_with_adjective_1",
+      "blimp_inchoative",
+      "blimp_animate_subject_trans",
+      "blimp_regular_plural_subject_verb_agreement_2",
+      "blimp_npi_present_1",
+      "blimp_determiner_noun_agreement_with_adj_irregular_2",
+      "blimp_matrix_question_npi_licensor_present"
+    ]
+  },
+  "configs": {
+    "blimp_adjunct_island": {
+      "task": "blimp_adjunct_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "adjunct_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_anaphor_gender_agreement": {
+      "task": "blimp_anaphor_gender_agreement",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "anaphor_gender_agreement",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_anaphor_number_agreement": {
+      "task": "blimp_anaphor_number_agreement",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "anaphor_number_agreement",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_animate_subject_passive": {
+      "task": "blimp_animate_subject_passive",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "animate_subject_passive",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_animate_subject_trans": {
+      "task": "blimp_animate_subject_trans",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "animate_subject_trans",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_causative": {
+      "task": "blimp_causative",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "causative",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_complex_NP_island": {
+      "task": "blimp_complex_NP_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "complex_NP_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_coordinate_structure_constraint_complex_left_branch": {
+      "task": "blimp_coordinate_structure_constraint_complex_left_branch",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "coordinate_structure_constraint_complex_left_branch",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_coordinate_structure_constraint_object_extraction": {
+      "task": "blimp_coordinate_structure_constraint_object_extraction",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "coordinate_structure_constraint_object_extraction",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_1": {
+      "task": "blimp_determiner_noun_agreement_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_2": {
+      "task": "blimp_determiner_noun_agreement_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_irregular_1": {
+      "task": "blimp_determiner_noun_agreement_irregular_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_irregular_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_irregular_2": {
+      "task": "blimp_determiner_noun_agreement_irregular_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_irregular_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adj_2": {
+      "task": "blimp_determiner_noun_agreement_with_adj_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adj_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": {
+      "task": "blimp_determiner_noun_agreement_with_adj_irregular_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adj_irregular_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": {
+      "task": "blimp_determiner_noun_agreement_with_adj_irregular_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adj_irregular_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_determiner_noun_agreement_with_adjective_1": {
+      "task": "blimp_determiner_noun_agreement_with_adjective_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "determiner_noun_agreement_with_adjective_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_distractor_agreement_relational_noun": {
+      "task": "blimp_distractor_agreement_relational_noun",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "distractor_agreement_relational_noun",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_distractor_agreement_relative_clause": {
+      "task": "blimp_distractor_agreement_relative_clause",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "distractor_agreement_relative_clause",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_drop_argument": {
+      "task": "blimp_drop_argument",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "drop_argument",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_ellipsis_n_bar_1": {
+      "task": "blimp_ellipsis_n_bar_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "ellipsis_n_bar_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_ellipsis_n_bar_2": {
+      "task": "blimp_ellipsis_n_bar_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "ellipsis_n_bar_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_object_raising": {
+      "task": "blimp_existential_there_object_raising",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_object_raising",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_quantifiers_1": {
+      "task": "blimp_existential_there_quantifiers_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_quantifiers_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_quantifiers_2": {
+      "task": "blimp_existential_there_quantifiers_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_quantifiers_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_existential_there_subject_raising": {
+      "task": "blimp_existential_there_subject_raising",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "existential_there_subject_raising",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_expletive_it_object_raising": {
+      "task": "blimp_expletive_it_object_raising",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "expletive_it_object_raising",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_inchoative": {
+      "task": "blimp_inchoative",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "inchoative",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_intransitive": {
+      "task": "blimp_intransitive",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "intransitive",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_past_participle_adjectives": {
+      "task": "blimp_irregular_past_participle_adjectives",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_past_participle_adjectives",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_past_participle_verbs": {
+      "task": "blimp_irregular_past_participle_verbs",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_past_participle_verbs",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_plural_subject_verb_agreement_1": {
+      "task": "blimp_irregular_plural_subject_verb_agreement_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_plural_subject_verb_agreement_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_irregular_plural_subject_verb_agreement_2": {
+      "task": "blimp_irregular_plural_subject_verb_agreement_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "irregular_plural_subject_verb_agreement_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_left_branch_island_echo_question": {
+      "task": "blimp_left_branch_island_echo_question",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "left_branch_island_echo_question",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_left_branch_island_simple_question": {
+      "task": "blimp_left_branch_island_simple_question",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "left_branch_island_simple_question",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_matrix_question_npi_licensor_present": {
+      "task": "blimp_matrix_question_npi_licensor_present",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "matrix_question_npi_licensor_present",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_npi_present_1": {
+      "task": "blimp_npi_present_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "npi_present_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_npi_present_2": {
+      "task": "blimp_npi_present_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "npi_present_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_only_npi_licensor_present": {
+      "task": "blimp_only_npi_licensor_present",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "only_npi_licensor_present",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_only_npi_scope": {
+      "task": "blimp_only_npi_scope",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "only_npi_scope",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_passive_1": {
+      "task": "blimp_passive_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "passive_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_passive_2": {
+      "task": "blimp_passive_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "passive_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_c_command": {
+      "task": "blimp_principle_A_c_command",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_c_command",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_case_1": {
+      "task": "blimp_principle_A_case_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_case_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_case_2": {
+      "task": "blimp_principle_A_case_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_case_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_domain_1": {
+      "task": "blimp_principle_A_domain_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_domain_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_domain_2": {
+      "task": "blimp_principle_A_domain_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_domain_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_domain_3": {
+      "task": "blimp_principle_A_domain_3",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_domain_3",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_principle_A_reconstruction": {
+      "task": "blimp_principle_A_reconstruction",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "principle_A_reconstruction",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_regular_plural_subject_verb_agreement_1": {
+      "task": "blimp_regular_plural_subject_verb_agreement_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "regular_plural_subject_verb_agreement_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_regular_plural_subject_verb_agreement_2": {
+      "task": "blimp_regular_plural_subject_verb_agreement_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "regular_plural_subject_verb_agreement_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_sentential_negation_npi_licensor_present": {
+      "task": "blimp_sentential_negation_npi_licensor_present",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "sentential_negation_npi_licensor_present",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_sentential_negation_npi_scope": {
+      "task": "blimp_sentential_negation_npi_scope",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "sentential_negation_npi_scope",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_sentential_subject_island": {
+      "task": "blimp_sentential_subject_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "sentential_subject_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_superlative_quantifiers_1": {
+      "task": "blimp_superlative_quantifiers_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "superlative_quantifiers_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_superlative_quantifiers_2": {
+      "task": "blimp_superlative_quantifiers_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "superlative_quantifiers_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_tough_vs_raising_1": {
+      "task": "blimp_tough_vs_raising_1",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "tough_vs_raising_1",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_tough_vs_raising_2": {
+      "task": "blimp_tough_vs_raising_2",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "tough_vs_raising_2",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_transitive": {
+      "task": "blimp_transitive",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "transitive",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_island": {
+      "task": "blimp_wh_island",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_island",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_questions_object_gap": {
+      "task": "blimp_wh_questions_object_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_questions_object_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_questions_subject_gap": {
+      "task": "blimp_wh_questions_subject_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_questions_subject_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_questions_subject_gap_long_distance": {
+      "task": "blimp_wh_questions_subject_gap_long_distance",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_questions_subject_gap_long_distance",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_no_gap": {
+      "task": "blimp_wh_vs_that_no_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_no_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_no_gap_long_distance": {
+      "task": "blimp_wh_vs_that_no_gap_long_distance",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_no_gap_long_distance",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_with_gap": {
+      "task": "blimp_wh_vs_that_with_gap",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_with_gap",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "blimp_wh_vs_that_with_gap_long_distance": {
+      "task": "blimp_wh_vs_that_with_gap_long_distance",
+      "group": "blimp",
+      "dataset_path": "blimp",
+      "dataset_name": "wh_vs_that_with_gap_long_distance",
+      "validation_split": "train",
+      "doc_to_text": "",
+      "doc_to_target": 0,
+      "doc_to_choice": "{{[sentence_good, sentence_bad]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "acc"
+        }
+      ],
+      "output_type": "multiple_choice",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{sentence_good}} {{sentence_bad}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "blimp_adjunct_island": 1.0,
+    "blimp_anaphor_gender_agreement": 1.0,
+    "blimp_anaphor_number_agreement": 1.0,
+    "blimp_animate_subject_passive": 1.0,
+    "blimp_animate_subject_trans": 1.0,
+    "blimp_causative": 1.0,
+    "blimp_complex_NP_island": 1.0,
+    "blimp_coordinate_structure_constraint_complex_left_branch": 1.0,
+    "blimp_coordinate_structure_constraint_object_extraction": 1.0,
+    "blimp_determiner_noun_agreement_1": 1.0,
+    "blimp_determiner_noun_agreement_2": 1.0,
+    "blimp_determiner_noun_agreement_irregular_1": 1.0,
+    "blimp_determiner_noun_agreement_irregular_2": 1.0,
+    "blimp_determiner_noun_agreement_with_adj_2": 1.0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": 1.0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": 1.0,
+    "blimp_determiner_noun_agreement_with_adjective_1": 1.0,
+    "blimp_distractor_agreement_relational_noun": 1.0,
+    "blimp_distractor_agreement_relative_clause": 1.0,
+    "blimp_drop_argument": 1.0,
+    "blimp_ellipsis_n_bar_1": 1.0,
+    "blimp_ellipsis_n_bar_2": 1.0,
+    "blimp_existential_there_object_raising": 1.0,
+    "blimp_existential_there_quantifiers_1": 1.0,
+    "blimp_existential_there_quantifiers_2": 1.0,
+    "blimp_existential_there_subject_raising": 1.0,
+    "blimp_expletive_it_object_raising": 1.0,
+    "blimp_inchoative": 1.0,
+    "blimp_intransitive": 1.0,
+    "blimp_irregular_past_participle_adjectives": 1.0,
+    "blimp_irregular_past_participle_verbs": 1.0,
+    "blimp_irregular_plural_subject_verb_agreement_1": 1.0,
+    "blimp_irregular_plural_subject_verb_agreement_2": 1.0,
+    "blimp_left_branch_island_echo_question": 1.0,
+    "blimp_left_branch_island_simple_question": 1.0,
+    "blimp_matrix_question_npi_licensor_present": 1.0,
+    "blimp_npi_present_1": 1.0,
+    "blimp_npi_present_2": 1.0,
+    "blimp_only_npi_licensor_present": 1.0,
+    "blimp_only_npi_scope": 1.0,
+    "blimp_passive_1": 1.0,
+    "blimp_passive_2": 1.0,
+    "blimp_principle_A_c_command": 1.0,
+    "blimp_principle_A_case_1": 1.0,
+    "blimp_principle_A_case_2": 1.0,
+    "blimp_principle_A_domain_1": 1.0,
+    "blimp_principle_A_domain_2": 1.0,
+    "blimp_principle_A_domain_3": 1.0,
+    "blimp_principle_A_reconstruction": 1.0,
+    "blimp_regular_plural_subject_verb_agreement_1": 1.0,
+    "blimp_regular_plural_subject_verb_agreement_2": 1.0,
+    "blimp_sentential_negation_npi_licensor_present": 1.0,
+    "blimp_sentential_negation_npi_scope": 1.0,
+    "blimp_sentential_subject_island": 1.0,
+    "blimp_superlative_quantifiers_1": 1.0,
+    "blimp_superlative_quantifiers_2": 1.0,
+    "blimp_tough_vs_raising_1": 1.0,
+    "blimp_tough_vs_raising_2": 1.0,
+    "blimp_transitive": 1.0,
+    "blimp_wh_island": 1.0,
+    "blimp_wh_questions_object_gap": 1.0,
+    "blimp_wh_questions_subject_gap": 1.0,
+    "blimp_wh_questions_subject_gap_long_distance": 1.0,
+    "blimp_wh_vs_that_no_gap": 1.0,
+    "blimp_wh_vs_that_no_gap_long_distance": 1.0,
+    "blimp_wh_vs_that_with_gap": 1.0,
+    "blimp_wh_vs_that_with_gap_long_distance": 1.0
+  },
+  "n-shot": {
+    "blimp": 0,
+    "blimp_adjunct_island": 0,
+    "blimp_anaphor_gender_agreement": 0,
+    "blimp_anaphor_number_agreement": 0,
+    "blimp_animate_subject_passive": 0,
+    "blimp_animate_subject_trans": 0,
+    "blimp_causative": 0,
+    "blimp_complex_NP_island": 0,
+    "blimp_coordinate_structure_constraint_complex_left_branch": 0,
+    "blimp_coordinate_structure_constraint_object_extraction": 0,
+    "blimp_determiner_noun_agreement_1": 0,
+    "blimp_determiner_noun_agreement_2": 0,
+    "blimp_determiner_noun_agreement_irregular_1": 0,
+    "blimp_determiner_noun_agreement_irregular_2": 0,
+    "blimp_determiner_noun_agreement_with_adj_2": 0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": 0,
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": 0,
+    "blimp_determiner_noun_agreement_with_adjective_1": 0,
+    "blimp_distractor_agreement_relational_noun": 0,
+    "blimp_distractor_agreement_relative_clause": 0,
+    "blimp_drop_argument": 0,
+    "blimp_ellipsis_n_bar_1": 0,
+    "blimp_ellipsis_n_bar_2": 0,
+    "blimp_existential_there_object_raising": 0,
+    "blimp_existential_there_quantifiers_1": 0,
+    "blimp_existential_there_quantifiers_2": 0,
+    "blimp_existential_there_subject_raising": 0,
+    "blimp_expletive_it_object_raising": 0,
+    "blimp_inchoative": 0,
+    "blimp_intransitive": 0,
+    "blimp_irregular_past_participle_adjectives": 0,
+    "blimp_irregular_past_participle_verbs": 0,
+    "blimp_irregular_plural_subject_verb_agreement_1": 0,
+    "blimp_irregular_plural_subject_verb_agreement_2": 0,
+    "blimp_left_branch_island_echo_question": 0,
+    "blimp_left_branch_island_simple_question": 0,
+    "blimp_matrix_question_npi_licensor_present": 0,
+    "blimp_npi_present_1": 0,
+    "blimp_npi_present_2": 0,
+    "blimp_only_npi_licensor_present": 0,
+    "blimp_only_npi_scope": 0,
+    "blimp_passive_1": 0,
+    "blimp_passive_2": 0,
+    "blimp_principle_A_c_command": 0,
+    "blimp_principle_A_case_1": 0,
+    "blimp_principle_A_case_2": 0,
+    "blimp_principle_A_domain_1": 0,
+    "blimp_principle_A_domain_2": 0,
+    "blimp_principle_A_domain_3": 0,
+    "blimp_principle_A_reconstruction": 0,
+    "blimp_regular_plural_subject_verb_agreement_1": 0,
+    "blimp_regular_plural_subject_verb_agreement_2": 0,
+    "blimp_sentential_negation_npi_licensor_present": 0,
+    "blimp_sentential_negation_npi_scope": 0,
+    "blimp_sentential_subject_island": 0,
+    "blimp_superlative_quantifiers_1": 0,
+    "blimp_superlative_quantifiers_2": 0,
+    "blimp_tough_vs_raising_1": 0,
+    "blimp_tough_vs_raising_2": 0,
+    "blimp_transitive": 0,
+    "blimp_wh_island": 0,
+    "blimp_wh_questions_object_gap": 0,
+    "blimp_wh_questions_subject_gap": 0,
+    "blimp_wh_questions_subject_gap_long_distance": 0,
+    "blimp_wh_vs_that_no_gap": 0,
+    "blimp_wh_vs_that_no_gap_long_distance": 0,
+    "blimp_wh_vs_that_with_gap": 0,
+    "blimp_wh_vs_that_with_gap_long_distance": 0
+  },
+  "n-samples": {
+    "blimp_causative": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_irregular_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_passive_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_questions_subject_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_plural_subject_verb_agreement_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_no_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_complex_NP_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_reconstruction": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_anaphor_number_agreement": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_domain_3": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_anaphor_gender_agreement": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_past_participle_adjectives": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_regular_plural_subject_verb_agreement_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_sentential_subject_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_left_branch_island_simple_question": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_distractor_agreement_relative_clause": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_plural_subject_verb_agreement_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_left_branch_island_echo_question": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_no_gap_long_distance": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_ellipsis_n_bar_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_superlative_quantifiers_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_ellipsis_n_bar_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_adjunct_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_expletive_it_object_raising": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_c_command": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_domain_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_superlative_quantifiers_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_only_npi_scope": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_tough_vs_raising_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_transitive": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_intransitive": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_questions_object_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_only_npi_licensor_present": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_with_gap": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_quantifiers_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_coordinate_structure_constraint_complex_left_branch": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_domain_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adj_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_island": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_quantifiers_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_object_raising": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_npi_present_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_animate_subject_passive": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_questions_subject_gap_long_distance": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_existential_there_subject_raising": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_case_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_principle_A_case_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_wh_vs_that_with_gap_long_distance": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_irregular_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_drop_argument": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_sentential_negation_npi_scope": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_coordinate_structure_constraint_object_extraction": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_irregular_past_participle_verbs": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_tough_vs_raising_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_sentential_negation_npi_licensor_present": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_distractor_agreement_relational_noun": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_passive_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adjective_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_inchoative": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_animate_subject_trans": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_regular_plural_subject_verb_agreement_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_npi_present_1": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_determiner_noun_agreement_with_adj_irregular_2": {
+      "original": 1000,
+      "effective": 1000
+    },
+    "blimp_matrix_question_npi_licensor_present": {
+      "original": 1000,
+      "effective": 1000
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step0,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step0",
+    "model_sha": "322fbe766b2a11f41f91b60d93bb282da3b1d236",
+    "batch_size": "1024",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1720730399.9945903,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: NVIDIA GeForce RTX 2080 Ti\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.91\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 695866.155545324,
+  "end_time": 696205.334018645,
+  "total_evaluation_time_seconds": "339.17847332102247"
+}

pythia-14m-seed1/step0/EleutherAI__pythia-14m-seed1/results_2024-08-12T05-34-26.878314.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 3577630.9945796956,
+      "perplexity_stderr,none": 346513.2535490424,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step0,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step0",
+    "model_sha": "322fbe766b2a11f41f91b60d93bb282da3b1d236",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723465836.4283125,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 586431.450237646,
+  "end_time": 586679.083148604,
+  "total_evaluation_time_seconds": "247.6329109580256"
+}

pythia-14m-seed1/step1/EleutherAI__pythia-14m-seed1/results_2024-08-12T05-38-36.798754.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 3577630.9945796956,
+      "perplexity_stderr,none": 346513.2535490424,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step1,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step1",
+    "model_sha": "391744899a133ba3b62507c3f5620e7844c74165",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723466089.8152428,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 586687.9045798,
+  "end_time": 586929.005673856,
+  "total_evaluation_time_seconds": "241.10109405603725"
+}

pythia-14m-seed1/step1000/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-20-43.708459.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 165785.28222773902,
+      "perplexity_stderr,none": 10148.605380657666,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step1000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step1000",
+    "model_sha": "e05524a5318c9889faa5114d3e8344527b02f7b3",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723468619.3948085,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 589217.172539005,
+  "end_time": 589455.915093445,
+  "total_evaluation_time_seconds": "238.74255443993025"
+}

pythia-14m-seed1/step10000/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-58-13.399858.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1706.8033769078256,
+      "perplexity_stderr,none": 80.98451307839903,
+      "acc,none": 0.09392586842615952,
+      "acc_stderr,none": 0.004064307401732146,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step10000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step10000",
+    "model_sha": "b65054c9d5be56a001a5d7cd7962bbd3b59e390b",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723470865.489443,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 591464.034879045,
+  "end_time": 591705.606917171,
+  "total_evaluation_time_seconds": "241.57203812606167"
+}

pythia-14m-seed1/step11000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-02-19.860338.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1825.5672976484855,
+      "perplexity_stderr,none": 83.87738455478605,
+      "acc,none": 0.08519309140306618,
+      "acc_stderr,none": 0.0038893674884810277,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step11000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step11000",
+    "model_sha": "5462937e3a2bc81a7bd55afba6d97055035f039a",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723471114.3282776,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 591711.589334535,
+  "end_time": 591952.06748634,
+  "total_evaluation_time_seconds": "240.47815180500038"
+}

pythia-14m-seed1/step12000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-06-26.034288.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1198.372281061769,
+      "perplexity_stderr,none": 55.44234602924562,
+      "acc,none": 0.11236173103046769,
+      "acc_stderr,none": 0.004399864402517477,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step12000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step12000",
+    "model_sha": "663bc4740e5033272b4d161b300f407388097236",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723471360.4058855,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 591958.125466808,
+  "end_time": 592198.241769348,
+  "total_evaluation_time_seconds": "240.11630254006013"
+}

pythia-14m-seed1/step128/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-08-24.350787.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 2161625.6376982476,
+      "perplexity_stderr,none": 190829.10781531825,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step128,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step128",
+    "model_sha": "fb3be0378c4bc5340bdc57314e644fbc83684898",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723467880.07867,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 588478.458262118,
+  "end_time": 588716.556904603,
+  "total_evaluation_time_seconds": "238.09864248498343"
+}

pythia-14m-seed1/step13000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-10-32.725397.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1435.0260911008309,
+      "perplexity_stderr,none": 66.65710190159811,
+      "acc,none": 0.10362895400737435,
+      "acc_stderr,none": 0.0042461618103990795,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step13000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step13000",
+    "model_sha": "279ac27ac96a15deb320d70e1110506ab7df3bb2",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723471605.7894518,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 592204.226383305,
+  "end_time": 592444.932036739,
+  "total_evaluation_time_seconds": "240.70565343392082"
+}

pythia-14m-seed1/step14000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-14-36.870500.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1330.1246672723364,
+      "perplexity_stderr,none": 61.00034368912197,
+      "acc,none": 0.09547836211915389,
+      "acc_stderr,none": 0.004094246940967937,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step14000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step14000",
+    "model_sha": "0c01936f9f66d64dfb5e0dfd8fe0a381c11315a5",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723471852.879908,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 592451.015213616,
+  "end_time": 592689.078115924,
+  "total_evaluation_time_seconds": "238.06290230806917"
+}

pythia-14m-seed1/step15000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-18-44.185809.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1300.454681174635,
+      "perplexity_stderr,none": 59.02578571631813,
+      "acc,none": 0.09023869590529789,
+      "acc_stderr,none": 0.003991831568038306,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step15000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step15000",
+    "model_sha": "9e47921509ab8f1200a0f21a5fdf6e3e8c32f7b2",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723472096.5834947,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 592694.964826637,
+  "end_time": 592936.392382781,
+  "total_evaluation_time_seconds": "241.4275561440736"
+}

pythia-14m-seed1/step16/EleutherAI__pythia-14m-seed1/results_2024-08-12T05-55-59.794548.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 3544256.031498556,
+      "perplexity_stderr,none": 342456.26642841654,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step16,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step16",
+    "model_sha": "3d5ad2743d7f32872bab8a63a3214757f5bafcac",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723467133.2361438,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 587726.05894772,
+  "end_time": 587972.00096739,
+  "total_evaluation_time_seconds": "245.94201967003755"
+}

pythia-14m-seed1/step16000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-22-52.656419.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1508.4200129111887,
+      "perplexity_stderr,none": 68.33885135504194,
+      "acc,none": 0.08771589365418203,
+      "acc_stderr,none": 0.003941089280335152,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step16000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step16000",
+    "model_sha": "f27f8309a5c14737f752871df8a3eaa59a3be527",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723472347.0784972,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 592942.377583877,
+  "end_time": 593184.863495992,
+  "total_evaluation_time_seconds": "242.48591211508028"
+}

pythia-14m-seed1/step17000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-27-02.544847.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1362.146262073091,
+      "perplexity_stderr,none": 64.59719270698977,
+      "acc,none": 0.09800116437026975,
+      "acc_stderr,none": 0.004142196325892838,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step17000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step17000",
+    "model_sha": "75f0e0ae07beaad37bae461145a4dbcc86e24692",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723472595.4651008,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 593190.810466927,
+  "end_time": 593434.751863186,
+  "total_evaluation_time_seconds": "243.94139625900425"
+}

pythia-14m-seed1/step18000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-31-07.685214.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1309.751280652761,
+      "perplexity_stderr,none": 59.10103664748906,
+      "acc,none": 0.09101494275179507,
+      "acc_stderr,none": 0.004007253297246434,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step18000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step18000",
+    "model_sha": "d6252bb5c699724f157757503e1c9425fe461a27",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723472842.273937,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 593440.757307799,
+  "end_time": 593679.892769297,
+  "total_evaluation_time_seconds": "239.13546149805188"
+}

pythia-14m-seed1/step19000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-35-11.859545.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1434.1712206097682,
+      "perplexity_stderr,none": 65.33748062140734,
+      "acc,none": 0.0894624490588007,
+      "acc_stderr,none": 0.003976320615111653,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step19000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step19000",
+    "model_sha": "e1ccd24dbac39b0dcdb8f41873d8a46c1026b22b",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723473087.5077255,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 593685.951817292,
+  "end_time": 593924.066392196,
+  "total_evaluation_time_seconds": "238.11457490397152"
+}

pythia-14m-seed1/step2/EleutherAI__pythia-14m-seed1/results_2024-08-12T05-42-57.791663.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 3577414.0382851013,
+      "perplexity_stderr,none": 346490.1079123521,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step2,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step2",
+    "model_sha": "a31976a6b857a128a1766c6d762c82422bf86b9a",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723466346.273168,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 586937.061340584,
+  "end_time": 587189.964603872,
+  "total_evaluation_time_seconds": "252.90326328808442"
+}

pythia-14m-seed1/step2000/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-25-00.074484.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 6442.748239816954,
+      "perplexity_stderr,none": 263.150168034604,
+      "acc,none": 0.02736270133902581,
+      "acc_stderr,none": 0.0022728303384483696,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step2000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step2000",
+    "model_sha": "e6f18bd42439e84072c51f5ed271662d04d022fc",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723468874.3639698,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 589463.891974103,
+  "end_time": 589712.281572735,
+  "total_evaluation_time_seconds": "248.38959863199852"
+}

pythia-14m-seed1/step20000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-39-21.895446.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1073.7601222397773,
+      "perplexity_stderr,none": 50.180031313469826,
+      "acc,none": 0.11274985445371628,
+      "acc_stderr,none": 0.00440649323200697,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step20000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step20000",
+    "model_sha": "cb17a08d30035e863620ed1bf43660ac61034a00",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723473335.1929982,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 593932.063681181,
+  "end_time": 594174.10247367,
+  "total_evaluation_time_seconds": "242.0387924889801"
+}

pythia-14m-seed1/step21000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-43-30.344693.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1315.570638563014,
+      "perplexity_stderr,none": 63.76335340304932,
+      "acc,none": 0.1121676693188434,
+      "acc_stderr,none": 0.004396543745902183,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.0002000252231615614,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.001148105625717566,
+      "exact_match_stderr,none": 0.0011481056257175658,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.0002000252231615614,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step21000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step21000",
+    "model_sha": "da14d73a2af1916540f2816a610007bce8ac08c3",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723473583.7465441,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 594181.919595051,
+  "end_time": 594422.551075061,
+  "total_evaluation_time_seconds": "240.63148000999354"
+}

pythia-14m-seed1/step22000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-47-38.819675.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1342.0689543431947,
+      "perplexity_stderr,none": 63.324437816541106,
+      "acc,none": 0.09819522608189404,
+      "acc_stderr,none": 0.004145849428034497,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step22000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step22000",
+    "model_sha": "299464f588430d6de40d815ce2455f356d27dd7b",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723473832.3860588,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 594430.683501378,
+  "end_time": 594671.025615628,
+  "total_evaluation_time_seconds": "240.34211424994282"
+}

pythia-14m-seed1/step23000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-51-49.064750.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1228.4784723168075,
+      "perplexity_stderr,none": 57.71083402349637,
+      "acc,none": 0.09974771977488842,
+      "acc_stderr,none": 0.004174896142165056,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step23000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step23000",
+    "model_sha": "2dc647a26d7124bca217b11f863a9c55244702ea",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723474080.5802898,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 594678.945846232,
+  "end_time": 594921.271703578,
+  "total_evaluation_time_seconds": "242.32585734594613"
+}

pythia-14m-seed1/step24000/EleutherAI__pythia-14m-seed1/results_2024-08-12T07-55-57.280297.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1248.5827910444752,
+      "perplexity_stderr,none": 58.28129910212816,
+      "acc,none": 0.09974771977488842,
+      "acc_stderr,none": 0.004174896142165056,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step24000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step24000",
+    "model_sha": "b4238dafabf80b0d2eec54cd35176af7d856a22b",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723474330.9848897,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 594929.263761255,
+  "end_time": 595169.48710706,
+  "total_evaluation_time_seconds": "240.22334580507595"
+}

pythia-14m-seed1/step25000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-00-23.273296.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1078.1037198020288,
+      "perplexity_stderr,none": 50.29663801073303,
+      "acc,none": 0.11080923733747332,
+      "acc_stderr,none": 0.004373181748568138,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step25000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step25000",
+    "model_sha": "",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723474586.227733,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 595177.544644202,
+  "end_time": 595435.480625855,
+  "total_evaluation_time_seconds": "257.93598165304866"
+}

pythia-14m-seed1/step256/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-12-30.518046.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1557200.1507235158,
+      "perplexity_stderr,none": 120989.57362868116,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step256,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step256",
+    "model_sha": "cd1a420123c542aa6480a57f5fef34e1c2503505",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723468127.1369855,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 588724.930293508,
+  "end_time": 588962.725103075,
+  "total_evaluation_time_seconds": "237.79480956704356"
+}

pythia-14m-seed1/step26000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-04-43.508436.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1143.0351018704325,
+      "perplexity_stderr,none": 51.58348486426997,
+      "acc,none": 0.09703085581214826,
+      "acc_stderr,none": 0.004123855686199016,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step26000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step26000",
+    "model_sha": "4b9cba39f4a8a56c34ecd6ca25e24b3648b821fb",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723474855.5270512,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 595453.977653388,
+  "end_time": 595695.715247744,
+  "total_evaluation_time_seconds": "241.73759435594548"
+}

pythia-14m-seed1/step27000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-08-54.479591.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1390.8543046931316,
+      "perplexity_stderr,none": 64.72963625566457,
+      "acc,none": 0.09004463419367359,
+      "acc_stderr,none": 0.003987962235095557,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step27000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step27000",
+    "model_sha": "ced4ebf986f79f1cc6e2e70f4307cfd7a2bc3b21",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723475105.3772542,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 595703.674144688,
+  "end_time": 595946.686523621,
+  "total_evaluation_time_seconds": "243.0123789330246"
+}

pythia-14m-seed1/step28000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-13-03.437554.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1183.8921931457348,
+      "perplexity_stderr,none": 54.52426093268814,
+      "acc,none": 0.10382301571899864,
+      "acc_stderr,none": 0.0042496756618570835,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step28000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step28000",
+    "model_sha": "76defabe487a6e50961940526ec36a31f4d4b12b",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723475356.888963,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 595954.866322523,
+  "end_time": 596195.644005189,
+  "total_evaluation_time_seconds": "240.7776826659683"
+}

pythia-14m-seed1/step29000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-17-10.235976.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1156.7714106994533,
+      "perplexity_stderr,none": 53.46807045970772,
+      "acc,none": 0.10886862022123035,
+      "acc_stderr,none": 0.004339446122928107,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.000200055824375816,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0008424599831508003,
+      "exact_match_stderr,none": 0.0008424599831508003,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.000200055824375816,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step29000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step29000",
+    "model_sha": "66c5b2e40d1100ce83f810ef45daabcd39a2eeba",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723475605.492433,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 596203.750955566,
+  "end_time": 596442.443272417,
+  "total_evaluation_time_seconds": "238.69231685099658"
+}

pythia-14m-seed1/step3000/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-29-07.176670.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 4557.244384016776,
+      "perplexity_stderr,none": 195.26897772477193,
+      "acc,none": 0.044440131961963905,
+      "acc_stderr,none": 0.002870970259614857,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step3000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step3000",
+    "model_sha": "9f5304a0d268fe9d3734d81200b6f63880a1ec31",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723469122.1623023,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 589720.380247604,
+  "end_time": 589959.382942748,
+  "total_evaluation_time_seconds": "239.00269514403772"
+}

pythia-14m-seed1/step30000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-21-21.161950.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1036.9991390279656,
+      "perplexity_stderr,none": 47.555329652720005,
+      "acc,none": 0.11022705220260043,
+      "acc_stderr,none": 0.00436310604106385,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step30000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step30000",
+    "model_sha": "06aebfad24380c55705b157ab403b7ef58838edf",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723475852.0692358,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 596450.437902112,
+  "end_time": 596693.369136271,
+  "total_evaluation_time_seconds": "242.93123415892478"
+}

pythia-14m-seed1/step31000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-25-31.758593.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 942.471093909997,
+      "perplexity_stderr,none": 44.99705295268038,
+      "acc,none": 0.1280807296720357,
+      "acc_stderr,none": 0.004655776323580886,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step31000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step31000",
+    "model_sha": "bcede1861a7058b4f117dc76a645556cbe0a4f67",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723476105.5894463,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 596701.277227848,
+  "end_time": 596943.965335081,
+  "total_evaluation_time_seconds": "242.68810723302886"
+}

pythia-14m-seed1/step32/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-00-10.088125.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 3445742.0970824086,
+      "perplexity_stderr,none": 330851.13687632605,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step32,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step32",
+    "model_sha": "3dc82a1d607c52fe5bcc1abf2c4e73760a4830bf",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723467381.9868443,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 587980.078017756,
+  "end_time": 588222.294495094,
+  "total_evaluation_time_seconds": "242.21647733799182"
+}

pythia-14m-seed1/step32000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-29-41.037818.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1160.8794826047688,
+      "perplexity_stderr,none": 53.486068122081036,
+      "acc,none": 0.10343489229575005,
+      "acc_stderr,none": 0.004242643325764933,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.00020002929690168824,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0011074197120708748,
+      "exact_match_stderr,none": 0.0011074197120708748,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.00020002929690168824,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step32000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step32000",
+    "model_sha": "987bb293e032cbd453e11b08655ee1e4ece27250",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723476355.902467,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 596951.993184742,
+  "end_time": 597193.243859167,
+  "total_evaluation_time_seconds": "241.25067442492582"
+}

pythia-14m-seed1/step33000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-33-49.879234.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1087.5960183425093,
+      "perplexity_stderr,none": 49.77281778231847,
+      "acc,none": 0.10130021346788279,
+      "acc_stderr,none": 0.0042036308668715675,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step33000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step33000",
+    "model_sha": "8e748cbf70be374b3f021a5fe9f8e732cc234c93",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723476603.089929,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 597201.285561259,
+  "end_time": 597442.086597274,
+  "total_evaluation_time_seconds": "240.80103601503652"
+}

pythia-14m-seed1/step34000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-38-10.348821.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1215.6802065018167,
+      "perplexity_stderr,none": 53.86288705444756,
+      "acc,none": 0.0894624490588007,
+      "acc_stderr,none": 0.003976320615111653,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.000200055824375816,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0008424599831508003,
+      "exact_match_stderr,none": 0.0008424599831508003,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.000200055824375816,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step34000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step34000",
+    "model_sha": "cfa1333d98ac8757f8c48150d0a2cf854f104b11",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723476865.3848665,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 597449.948562086,
+  "end_time": 597702.555639231,
+  "total_evaluation_time_seconds": "252.60707714501768"
+}

pythia-14m-seed1/step35000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-42-22.390553.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1165.123090953555,
+      "perplexity_stderr,none": 52.749636148758206,
+      "acc,none": 0.10168833689113138,
+      "acc_stderr,none": 0.004210766549718483,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step35000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step35000",
+    "model_sha": "4d7b052635e59a22f5b68148ef89a6759006aedd",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723477114.2603142,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 597710.476242525,
+  "end_time": 597954.596861367,
+  "total_evaluation_time_seconds": "244.1206188419601"
+}

pythia-14m-seed1/step36000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-46-35.988527.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1421.4078171093663,
+      "perplexity_stderr,none": 62.81381033079262,
+      "acc,none": 0.08732777023093344,
+      "acc_stderr,none": 0.0039331967952010785,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step36000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step36000",
+    "model_sha": "cb765c5781bf9b9eb1f345976a95644a6b691a67",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723477370.389327,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 597962.724644432,
+  "end_time": 598208.195816519,
+  "total_evaluation_time_seconds": "245.47117208701093"
+}

pythia-14m-seed1/step37000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-50-44.879723.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1217.9659264756383,
+      "perplexity_stderr,none": 55.009084692314545,
+      "acc,none": 0.09994178148651271,
+      "acc_stderr,none": 0.004178504912005431,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step37000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step37000",
+    "model_sha": "2a2d8d1029b462415958aff9167bb1d8a152ab5e",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723477617.9696627,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 598216.258751823,
+  "end_time": 598457.087299166,
+  "total_evaluation_time_seconds": "240.82854734302964"
+}

pythia-14m-seed1/step38000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-54-53.146841.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1111.7325934316825,
+      "perplexity_stderr,none": 52.59070829519442,
+      "acc,none": 0.10983892877935184,
+      "acc_stderr,none": 0.004356367566472683,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step38000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step38000",
+    "model_sha": "b6a18fa195cfef08d8f5ec22a754d86db34f62e1",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723477866.4832387,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 598465.033826554,
+  "end_time": 598705.353701372,
+  "total_evaluation_time_seconds": "240.31987481808756"
+}

pythia-14m-seed1/step39000/EleutherAI__pythia-14m-seed1/results_2024-08-12T08-59-06.449694.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1182.0839232543688,
+      "perplexity_stderr,none": 55.96832739868955,
+      "acc,none": 0.10459926256549583,
+      "acc_stderr,none": 0.004263684964802846,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.0001999547463471357,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.001851851851851852,
+      "exact_match_stderr,none": 0.001851851851851852,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.0001999547463471357,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step39000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step39000",
+    "model_sha": "5a4c489df54902660037150c9efc7fe288ba5445",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723478118.404584,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 598713.237571929,
+  "end_time": 598958.656903778,
+  "total_evaluation_time_seconds": "245.41933184897061"
+}

pythia-14m-seed1/step4/EleutherAI__pythia-14m-seed1/results_2024-08-12T05-47-23.908503.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 3577077.782029242,
+      "perplexity_stderr,none": 346451.99565970706,
+      "acc,none": 0.0,
+      "acc_stderr,none": 0.0,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step4,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step4",
+    "model_sha": "72f661aa74ff6f4226f9b6830e8888c1312664e2",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723466616.3978312,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 587213.770161715,
+  "end_time": 587456.114898593,
+  "total_evaluation_time_seconds": "242.34473687806167"
+}

pythia-14m-seed1/step4000/EleutherAI__pythia-14m-seed1/results_2024-08-12T06-33-15.998610.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 2808.885365372563,
+      "perplexity_stderr,none": 124.03134252171543,
+      "acc,none": 0.06384630312439356,
+      "acc_stderr,none": 0.0034060696844330452,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step4000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step4000",
+    "model_sha": "fada4a82218d9b0254c96b1c369a423e827a409f",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723469369.1975448,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 589967.38387155,
+  "end_time": 590208.205939946,
+  "total_evaluation_time_seconds": "240.8220683960244"
+}

pythia-14m-seed1/step40000/EleutherAI__pythia-14m-seed1/results_2024-08-12T09-03-22.077168.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1204.9911371450303,
+      "perplexity_stderr,none": 54.28190867159313,
+      "acc,none": 0.09392586842615952,
+      "acc_stderr,none": 0.004064307401732146,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step40000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step40000",
+    "model_sha": "8ff1656834308baaecb093c2f3172c9198144382",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723478377.1336043,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 598965.37537771,
+  "end_time": 599214.284706319,
+  "total_evaluation_time_seconds": "248.90932860900648"
+}

pythia-14m-seed1/step41000/EleutherAI__pythia-14m-seed1/results_2024-08-12T09-07-29.805237.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1066.3853643932846,
+      "perplexity_stderr,none": 48.51192891663625,
+      "acc,none": 0.10421113914224724,
+      "acc_stderr,none": 0.004256689511125182,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.00019999999999999998,
+      "exact_match_stderr,none": 0.00019993112346228413,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0020876826722338203,
+      "exact_match_stderr,none": 0.0020876826722338207,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.00019999999999999998,
+      "exact_match_stderr,none": 0.00019993112346228413,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step41000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step41000",
+    "model_sha": "f99142bc12d9ade53a2f94c4c036aac75c7519ec",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723478622.908552,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               2200.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 599220.025245513,
+  "end_time": 599462.007075518,
+  "total_evaluation_time_seconds": "241.98183000506833"
+}

pythia-14m-seed1/step42000/EleutherAI__pythia-14m-seed1/results_2024-08-12T09-11-40.178060.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1120.2234090661466,
+      "perplexity_stderr,none": 50.333654631556755,
+      "acc,none": 0.09664273238889967,
+      "acc_stderr,none": 0.004116484120110667,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step42000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step42000",
+    "model_sha": "3ab79635fe11167d8cb1170a167c937f0c49838b",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723478870.766831,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 599467.765480389,
+  "end_time": 599712.385375487,
+  "total_evaluation_time_seconds": "244.61989509803243"
+}

pythia-14m-seed1/step43000/EleutherAI__pythia-14m-seed1/results_2024-08-12T09-15-48.237659.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1222.7750559057145,
+      "perplexity_stderr,none": 55.19618389771133,
+      "acc,none": 0.09761304094702115,
+      "acc_stderr,none": 0.0041348751357468405,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.00019995678465298491,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0018315018315018315,
+      "exact_match_stderr,none": 0.0018315018315018313,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0002,
+      "exact_match_stderr,none": 0.00019995678465298491,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step43000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step43000",
+    "model_sha": "a5868a08e71e59908f5b17551b31db6eca0fd211",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723479121.3723686,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 599718.335975932,
+  "end_time": 599960.440780523,
+  "total_evaluation_time_seconds": "242.10480459104292"
+}

pythia-14m-seed1/step44000/EleutherAI__pythia-14m-seed1/results_2024-08-12T09-20-11.460936.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1054.0902326469609,
+      "perplexity_stderr,none": 47.94516895248167,
+      "acc,none": 0.10440520085387153,
+      "acc_stderr,none": 0.004260189531743511,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step44000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step44000",
+    "model_sha": "84f55295500644f84cf599c6c755a88927fe6371",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723479380.6053634,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 599967.536589297,
+  "end_time": 600223.662270896,
+  "total_evaluation_time_seconds": "256.12568159902"
+}

pythia-14m-seed1/step45000/EleutherAI__pythia-14m-seed1/results_2024-08-12T09-24-41.702741.json

@@ -0,0 +1,482 @@
+{
+  "results": {
+    "lambada_openai": {
+      "perplexity,none": 1151.4880964856604,
+      "perplexity_stderr,none": 51.8272862471497,
+      "acc,none": 0.09838928779351834,
+      "acc_stderr,none": 0.004149497552485457,
+      "alias": "lambada_openai"
+    },
+    "hendrycks_math": {
+      "exact_match,none": 0.00019999999999999998,
+      "exact_match_stderr,none": 0.00019993112346228413,
+      "alias": "hendrycks_math"
+    },
+    "hendrycks_math_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_algebra"
+    },
+    "hendrycks_math_counting_and_prob": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_counting_and_prob"
+    },
+    "hendrycks_math_geometry": {
+      "exact_match,none": 0.0020876826722338203,
+      "exact_match_stderr,none": 0.0020876826722338207,
+      "alias": " - hendrycks_math_geometry"
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_intermediate_algebra"
+    },
+    "hendrycks_math_num_theory": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_num_theory"
+    },
+    "hendrycks_math_prealgebra": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_prealgebra"
+    },
+    "hendrycks_math_precalc": {
+      "exact_match,none": 0.0,
+      "exact_match_stderr,none": 0.0,
+      "alias": " - hendrycks_math_precalc"
+    }
+  },
+  "groups": {
+    "hendrycks_math": {
+      "exact_match,none": 0.00019999999999999998,
+      "exact_match_stderr,none": 0.00019993112346228413,
+      "alias": "hendrycks_math"
+    }
+  },
+  "group_subtasks": {
+    "hendrycks_math": [
+      "hendrycks_math_precalc",
+      "hendrycks_math_prealgebra",
+      "hendrycks_math_num_theory",
+      "hendrycks_math_intermediate_algebra",
+      "hendrycks_math_geometry",
+      "hendrycks_math_counting_and_prob",
+      "hendrycks_math_algebra"
+    ],
+    "lambada_openai": []
+  },
+  "configs": {
+    "hendrycks_math_algebra": {
+      "task": "hendrycks_math_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_counting_and_prob": {
+      "task": "hendrycks_math_counting_and_prob",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "counting_and_probability",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_geometry": {
+      "task": "hendrycks_math_geometry",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "geometry",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "task": "hendrycks_math_intermediate_algebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "intermediate_algebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_num_theory": {
+      "task": "hendrycks_math_num_theory",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "number_theory",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_prealgebra": {
+      "task": "hendrycks_math_prealgebra",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "prealgebra",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "hendrycks_math_precalc": {
+      "task": "hendrycks_math_precalc",
+      "group": [
+        "math_word_problems"
+      ],
+      "dataset_path": "EleutherAI/hendrycks_math",
+      "dataset_name": "precalculus",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "training_split": "train",
+      "test_split": "test",
+      "process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n    def _process_doc(doc: dict) -> dict:\n        out_doc = {\n            \"problem\": doc[\"problem\"],\n            \"solution\": doc[\"solution\"],\n            \"answer\": remove_boxed(last_boxed_only_string(doc[\"solution\"])),\n        }\n        return out_doc\n\n    return dataset.map(_process_doc)\n",
+      "doc_to_text": "Problem: {{problem}}\nAnswer:",
+      "doc_to_target": "{{answer}}",
+      "process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n    retval = 0\n    indices = [pos for pos, char in enumerate(results[0]) if char == \"$\"]\n    if len(indices) <= 1:\n        answer = results[0]\n    else:\n        answer = results[0][indices[0] + 1 : indices[-1]]\n\n    if is_equiv(answer, remove_boxed(last_boxed_only_string(doc[\"solution\"]))):\n        retval = 1\n\n    results = {\n        \"exact_match\": retval,\n    }\n    return results\n",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "exact_match",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "generate_until",
+      "generation_kwargs": {
+        "until": [
+          "Problem:"
+        ],
+        "do_sample": false,
+        "temperature": 0.0
+      },
+      "repeats": 1,
+      "should_decontaminate": false,
+      "metadata": {
+        "version": 1.0
+      }
+    },
+    "lambada_openai": {
+      "task": "lambada_openai",
+      "group": [
+        "lambada"
+      ],
+      "dataset_path": "EleutherAI/lambada_openai",
+      "dataset_name": "default",
+      "dataset_kwargs": {
+        "trust_remote_code": true
+      },
+      "test_split": "test",
+      "doc_to_text": "{{text.split(' ')[:-1]|join(' ')}}",
+      "doc_to_target": "{{' '+text.split(' ')[-1]}}",
+      "description": "",
+      "target_delimiter": " ",
+      "fewshot_delimiter": "\n\n",
+      "num_fewshot": 0,
+      "metric_list": [
+        {
+          "metric": "perplexity",
+          "aggregation": "perplexity",
+          "higher_is_better": false
+        },
+        {
+          "metric": "acc",
+          "aggregation": "mean",
+          "higher_is_better": true
+        }
+      ],
+      "output_type": "loglikelihood",
+      "repeats": 1,
+      "should_decontaminate": true,
+      "doc_to_decontamination_query": "{{text}}",
+      "metadata": {
+        "version": 1.0
+      }
+    }
+  },
+  "versions": {
+    "hendrycks_math_algebra": 1.0,
+    "hendrycks_math_counting_and_prob": 1.0,
+    "hendrycks_math_geometry": 1.0,
+    "hendrycks_math_intermediate_algebra": 1.0,
+    "hendrycks_math_num_theory": 1.0,
+    "hendrycks_math_prealgebra": 1.0,
+    "hendrycks_math_precalc": 1.0,
+    "lambada_openai": 1.0
+  },
+  "n-shot": {
+    "hendrycks_math": 0,
+    "hendrycks_math_algebra": 0,
+    "hendrycks_math_counting_and_prob": 0,
+    "hendrycks_math_geometry": 0,
+    "hendrycks_math_intermediate_algebra": 0,
+    "hendrycks_math_num_theory": 0,
+    "hendrycks_math_prealgebra": 0,
+    "hendrycks_math_precalc": 0,
+    "lambada_openai": 0
+  },
+  "n-samples": {
+    "lambada_openai": {
+      "original": 5153,
+      "effective": 5153
+    },
+    "hendrycks_math_precalc": {
+      "original": 546,
+      "effective": 546
+    },
+    "hendrycks_math_prealgebra": {
+      "original": 871,
+      "effective": 871
+    },
+    "hendrycks_math_num_theory": {
+      "original": 540,
+      "effective": 540
+    },
+    "hendrycks_math_intermediate_algebra": {
+      "original": 903,
+      "effective": 903
+    },
+    "hendrycks_math_geometry": {
+      "original": 479,
+      "effective": 479
+    },
+    "hendrycks_math_counting_and_prob": {
+      "original": 474,
+      "effective": 474
+    },
+    "hendrycks_math_algebra": {
+      "original": 1187,
+      "effective": 1187
+    }
+  },
+  "config": {
+    "model": "hf",
+    "model_args": "pretrained=EleutherAI/pythia-14m-seed1,revision=step45000,",
+    "model_num_parameters": 14067712,
+    "model_dtype": "torch.float16",
+    "model_revision": "step45000",
+    "model_sha": "6a56a88dbb56f315216c31814f5c54c2ebbd757f",
+    "batch_size": "128",
+    "batch_sizes": [],
+    "device": "cuda",
+    "use_cache": null,
+    "limit": null,
+    "bootstrap_iters": 100000,
+    "gen_kwargs": null,
+    "random_seed": 0,
+    "numpy_seed": 1234,
+    "torch_seed": 1234,
+    "fewshot_seed": 1234
+  },
+  "git_hash": "51a7ca9",
+  "date": 1723479654.2901926,
+  "pretty_env_info": "PyTorch version: 2.3.0+cu121\nIs debug build: False\nCUDA used to build PyTorch: 12.1\nROCM used to build PyTorch: N/A\n\nOS: CentOS Linux release 7.9.2009 (Core) (x86_64)\nGCC version: (GCC) 12.1.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.17\n\nPython version: 3.12.1 (main, Jan 12 2024, 16:49:08) [GCC 12.1.0] (64-bit runtime)\nPython platform: Linux-3.10.0-1160.119.1.el7.tuxcare.els2.x86_64-x86_64-with-glibc2.17\nIs CUDA available: True\nCUDA runtime version: 12.4.99\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: GPU 0: Tesla V100S-PCIE-32GB\nNvidia driver version: 550.90.07\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture:          x86_64\nCPU op-mode(s):        32-bit, 64-bit\nByte Order:            Little Endian\nCPU(s):                32\nOn-line CPU(s) list:   0-31\nThread(s) per core:    1\nCore(s) per socket:    32\nSocket(s):             1\nNUMA node(s):          2\nVendor ID:             AuthenticAMD\nCPU family:            23\nModel:                 49\nModel name:            AMD EPYC 7502P 32-Core Processor\nStepping:              0\nCPU MHz:               1500.000\nCPU max MHz:           2500.0000\nCPU min MHz:           1500.0000\nBogoMIPS:              4999.78\nVirtualization:        AMD-V\nL1d cache:             32K\nL1i cache:             32K\nL2 cache:              512K\nL3 cache:              16384K\nNUMA node0 CPU(s):     0-15\nNUMA node1 CPU(s):     16-31\nFlags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc art rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_l2 cpb cat_l3 cdp_l3 hw_pstate sme ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip overflow_recov succor smca\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.3.0\n[conda] Could not collect",
+  "transformers_version": "4.40.2",
+  "upper_git_hash": null,
+  "task_hashes": {},
+  "model_source": "hf",
+  "model_name": "EleutherAI/pythia-14m-seed1",
+  "model_name_sanitized": "EleutherAI__pythia-14m-seed1",
+  "start_time": 600230.934252729,
+  "end_time": 600493.901946463,
+  "total_evaluation_time_seconds": "262.96769373398274"
+}