Training in progress, epoch 13, checkpoint

checkpoint-3250/adapter_config.json

@@ -1,6 +1,6 @@
 {
   "auto_mapping": null,
-  "base_model_name_or_path": "Qwen/Qwen1.5-4B",
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,

checkpoint-3250/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49da2a380542f82cef80568ad78ca503c9461ba4c299ace533b79f0fc8db0831
+size 143269386

checkpoint-3250/added_tokens.json

@@ -1,5 +1,5 @@
 {
-  "<|endoftext|>": 151643,
-  "<|im_end|>": 151645,
-  "<|im_start|>": 151644
+  "</s>": 2,
+  "<s>": 1,
+  "<unk>": 0
 }

checkpoint-3250/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81b2ee6ea7f07be6062352fcf5ead6c30bf6164c04bbe063b41a78f57cd901cb
-size 224537202
+oid sha256:45a4c8aac4bcf3a31ecc7c7eff4c062342a3bd1562a0cf35a9b598638d59dfa5
+size 286585234

checkpoint-3250/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cbe3eeb85ad6125e9c720931612782bfb33dc653e6508eafa576c018dbac7f6c
+oid sha256:935ebdc5646805a88df3e04672192b4f3ccea9db014368810615dd89e848e699
 size 15024

checkpoint-3250/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89cba79818c7c57715915330d6251dab794769fb455234c9dc6d7c2bd2e0d435
+oid sha256:e98a55aae31dcaa87e35253c8fa2ba86844f1a1d487c0b40f3eeb6626bf723e8
 size 15024

checkpoint-3250/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dbbd9086bcb785da51dcc2ef082d512acba7272152c32ad4fdc7691dfe2a8cfb
+oid sha256:2dd549b56dfe3d7f5f9611ea874d84e0534c3f1503b1e2ef8a07dc23a40cf751
 size 15024

checkpoint-3250/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce43dd6b881dd02a4e0a6eab49d4d8400e7e234b4f7472cc6ee52032dcd71428
+oid sha256:9ccdac159ba3210827c066a8e0292fc42f4fd3c83fd649f364187ed380a90e7a
 size 15024

checkpoint-3250/special_tokens_map.json

@@ -1,14 +1,6 @@
 {
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "eos_token": {
-    "content": "<|endoftext|>",
-    "lstrip": false,
-    "normalized": false,
-    "rstrip": false,
-    "single_word": false
-  },
-  "pad_token": "<|endoftext|>"
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "pad_token": "</s>",
+  "unk_token": "<unk>"
 }

checkpoint-3250/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

checkpoint-3250/tokenizer_config.json

@@ -1,24 +1,23 @@
 {
-  "add_prefix_space": false,
   "added_tokens_decoder": {
-    "151643": {
-      "content": "<|endoftext|>",
+    "0": {
+      "content": "<unk>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "151644": {
-      "content": "<|im_start|>",
+    "1": {
+      "content": "<s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "151645": {
-      "content": "<|im_end|>",
+    "2": {
+      "content": "</s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -26,18 +25,16 @@
       "special": true
     }
   },
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "bos_token": null,
-  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
   "clean_up_tokenization_spaces": false,
-  "eos_token": "<|endoftext|>",
-  "errors": "replace",
-  "model_max_length": 32768,
-  "pad_token": "<|endoftext|>",
-  "split_special_tokens": false,
-  "tokenizer_class": "Qwen2Tokenizer",
-  "unk_token": null
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": true
 }

checkpoint-3250/trainer_state.json

@@ -10,443 +10,397 @@
   "log_history": [
     {
       "epoch": 0.4,
-      "grad_norm": 0.5479037761688232,
       "learning_rate": 0.0005,
-      "loss": 2.3029,
+      "loss": 2.3712,
       "step": 100
     },
     {
       "epoch": 0.8,
-      "grad_norm": 0.5442906618118286,
       "learning_rate": 0.0005,
-      "loss": 2.2398,
+      "loss": 1.8732,
       "step": 200
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.5162539682539683,
-      "eval_loss": 2.3236427307128906,
-      "eval_runtime": 5.266,
-      "eval_samples_per_second": 94.949,
-      "eval_steps_per_second": 11.964,
+      "eval_accuracy": 0.593873417721519,
+      "eval_loss": 2.011091947555542,
+      "eval_runtime": 4.861,
+      "eval_samples_per_second": 102.859,
+      "eval_steps_per_second": 12.96,
       "step": 250
     },
     {
       "epoch": 1.0,
-      "eval_exact_match": 17.6,
-      "eval_f1": 26.59450937950939,
+      "eval_exact_match": 0.0,
+      "eval_f1": 0.32835692091264185,
       "step": 250
     },
     {
       "epoch": 1.2,
-      "grad_norm": 0.6918729543685913,
       "learning_rate": 0.0005,
-      "loss": 2.0245,
+      "loss": 1.7329,
       "step": 300
     },
     {
       "epoch": 1.6,
-      "grad_norm": 0.7950774431228638,
       "learning_rate": 0.0005,
-      "loss": 1.8286,
+      "loss": 1.6233,
       "step": 400
     },
     {
       "epoch": 2.0,
-      "grad_norm": 0.6913579106330872,
       "learning_rate": 0.0005,
-      "loss": 1.8301,
+      "loss": 1.6142,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.5124126984126984,
-      "eval_loss": 2.4219954013824463,
-      "eval_runtime": 5.2763,
-      "eval_samples_per_second": 94.763,
-      "eval_steps_per_second": 11.94,
+      "eval_accuracy": 0.6051139240506329,
+      "eval_loss": 1.844284176826477,
+      "eval_runtime": 4.8205,
+      "eval_samples_per_second": 103.724,
+      "eval_steps_per_second": 13.069,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_exact_match": 17.0,
-      "eval_f1": 24.808081363081367,
+      "eval_exact_match": 0.0,
+      "eval_f1": 0.7215873015873017,
       "step": 500
     },
     {
       "epoch": 2.4,
-      "grad_norm": 0.9923084378242493,
       "learning_rate": 0.0005,
-      "loss": 1.3144,
+      "loss": 1.1625,
       "step": 600
     },
     {
       "epoch": 2.8,
-      "grad_norm": 1.0025588274002075,
       "learning_rate": 0.0005,
-      "loss": 1.3626,
+      "loss": 1.206,
       "step": 700
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.5061904761904762,
-      "eval_loss": 2.6152632236480713,
-      "eval_runtime": 5.2745,
-      "eval_samples_per_second": 94.795,
-      "eval_steps_per_second": 11.944,
+      "eval_accuracy": 0.6007341772151898,
+      "eval_loss": 1.981817603111267,
+      "eval_runtime": 4.671,
+      "eval_samples_per_second": 107.044,
+      "eval_steps_per_second": 13.487,
       "step": 750
     },
     {
       "epoch": 3.0,
-      "eval_exact_match": 16.0,
-      "eval_f1": 24.234217449217454,
+      "eval_exact_match": 0.0,
+      "eval_f1": 0.4163636363636364,
       "step": 750
     },
     {
       "epoch": 3.2,
-      "grad_norm": 1.0352345705032349,
       "learning_rate": 0.0005,
-      "loss": 1.1681,
+      "loss": 1.0064,
       "step": 800
     },
     {
       "epoch": 3.6,
-      "grad_norm": 1.1078394651412964,
       "learning_rate": 0.0005,
-      "loss": 0.9584,
+      "loss": 0.8141,
       "step": 900
     },
     {
       "epoch": 4.0,
-      "grad_norm": 1.14862060546875,
       "learning_rate": 0.0005,
-      "loss": 1.0112,
+      "loss": 0.8693,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.49974603174603177,
-      "eval_loss": 2.834880828857422,
-      "eval_runtime": 5.1472,
-      "eval_samples_per_second": 97.141,
-      "eval_steps_per_second": 12.24,
+      "eval_accuracy": 0.594126582278481,
+      "eval_loss": 2.2100281715393066,
+      "eval_runtime": 5.2488,
+      "eval_samples_per_second": 95.26,
+      "eval_steps_per_second": 12.003,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_exact_match": 13.6,
-      "eval_f1": 20.6380957930958,
+      "eval_exact_match": 0.2,
+      "eval_f1": 1.559097642686839,
       "step": 1000
     },
     {
       "epoch": 4.4,
-      "grad_norm": 1.2231959104537964,
       "learning_rate": 0.0005,
-      "loss": 0.6614,
+      "loss": 0.5477,
       "step": 1100
     },
     {
       "epoch": 4.8,
-      "grad_norm": 1.1669212579727173,
       "learning_rate": 0.0005,
-      "loss": 0.7198,
+      "loss": 0.6023,
       "step": 1200
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.4963174603174603,
-      "eval_loss": 3.075563907623291,
-      "eval_runtime": 5.1584,
-      "eval_samples_per_second": 96.929,
-      "eval_steps_per_second": 12.213,
+      "eval_accuracy": 0.5910126582278481,
+      "eval_loss": 2.375615119934082,
+      "eval_runtime": 4.4156,
+      "eval_samples_per_second": 113.235,
+      "eval_steps_per_second": 14.268,
       "step": 1250
     },
     {
       "epoch": 5.0,
-      "eval_exact_match": 14.0,
-      "eval_f1": 20.88632090132091,
+      "eval_exact_match": 0.4,
+      "eval_f1": 2.41971862073413,
       "step": 1250
     },
     {
       "epoch": 5.2,
-      "grad_norm": 0.9650686979293823,
       "learning_rate": 0.0005,
-      "loss": 0.6299,
+      "loss": 0.514,
       "step": 1300
     },
     {
       "epoch": 5.6,
-      "grad_norm": 0.9634979963302612,
       "learning_rate": 0.0005,
-      "loss": 0.5491,
+      "loss": 0.4479,
       "step": 1400
     },
     {
       "epoch": 6.0,
-      "grad_norm": 1.0452007055282593,
       "learning_rate": 0.0005,
-      "loss": 0.589,
+      "loss": 0.4717,
       "step": 1500
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.49434920634920637,
-      "eval_loss": 3.2339136600494385,
-      "eval_runtime": 5.1512,
-      "eval_samples_per_second": 97.064,
-      "eval_steps_per_second": 12.23,
+      "eval_accuracy": 0.5895696202531645,
+      "eval_loss": 2.542104482650757,
+      "eval_runtime": 5.1447,
+      "eval_samples_per_second": 97.188,
+      "eval_steps_per_second": 12.246,
       "step": 1500
     },
     {
       "epoch": 6.0,
-      "eval_exact_match": 15.0,
-      "eval_f1": 21.66983199806731,
+      "eval_exact_match": 9.8,
+      "eval_f1": 17.19683736589267,
       "step": 1500
     },
     {
       "epoch": 6.4,
-      "grad_norm": 1.0834842920303345,
       "learning_rate": 0.0005,
-      "loss": 0.4679,
+      "loss": 0.3723,
       "step": 1600
     },
     {
       "epoch": 6.8,
-      "grad_norm": 0.9649925231933594,
       "learning_rate": 0.0005,
-      "loss": 0.4969,
+      "loss": 0.3938,
       "step": 1700
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.4934920634920635,
-      "eval_loss": 3.34251070022583,
-      "eval_runtime": 5.2951,
-      "eval_samples_per_second": 94.426,
-      "eval_steps_per_second": 11.898,
+      "eval_accuracy": 0.5891139240506329,
+      "eval_loss": 2.658656597137451,
+      "eval_runtime": 4.7175,
+      "eval_samples_per_second": 105.987,
+      "eval_steps_per_second": 13.354,
       "step": 1750
     },
     {
       "epoch": 7.0,
-      "eval_exact_match": 13.8,
-      "eval_f1": 20.9043167943168,
+      "eval_exact_match": 3.6,
+      "eval_f1": 8.743843476498272,
       "step": 1750
     },
     {
       "epoch": 7.2,
-      "grad_norm": 0.8583429455757141,
       "learning_rate": 0.0005,
-      "loss": 0.4737,
+      "loss": 0.3727,
       "step": 1800
     },
     {
       "epoch": 7.6,
-      "grad_norm": 0.9595094919204712,
       "learning_rate": 0.0005,
-      "loss": 0.4544,
+      "loss": 0.3538,
       "step": 1900
     },
     {
       "epoch": 8.0,
-      "grad_norm": 0.9260773062705994,
       "learning_rate": 0.0005,
-      "loss": 0.4786,
+      "loss": 0.3697,
       "step": 2000
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.49244444444444446,
-      "eval_loss": 3.4198057651519775,
-      "eval_runtime": 5.8813,
-      "eval_samples_per_second": 85.015,
-      "eval_steps_per_second": 10.712,
+      "eval_accuracy": 0.5872911392405064,
+      "eval_loss": 2.7531518936157227,
+      "eval_runtime": 4.7213,
+      "eval_samples_per_second": 105.902,
+      "eval_steps_per_second": 13.344,
       "step": 2000
     },
     {
       "epoch": 8.0,
-      "eval_exact_match": 13.4,
-      "eval_f1": 20.923260628260632,
+      "eval_exact_match": 0.0,
+      "eval_f1": 0.3733369257948794,
       "step": 2000
     },
     {
       "epoch": 8.4,
-      "grad_norm": 1.045862078666687,
       "learning_rate": 0.0005,
-      "loss": 0.4114,
+      "loss": 0.3431,
       "step": 2100
     },
     {
       "epoch": 8.8,
-      "grad_norm": 0.7850492000579834,
       "learning_rate": 0.0005,
-      "loss": 0.4399,
+      "loss": 0.3617,
       "step": 2200
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.49114285714285716,
-      "eval_loss": 3.4694714546203613,
-      "eval_runtime": 5.945,
-      "eval_samples_per_second": 84.104,
-      "eval_steps_per_second": 10.597,
+      "eval_accuracy": 0.5870126582278481,
+      "eval_loss": 2.7664403915405273,
+      "eval_runtime": 4.7203,
+      "eval_samples_per_second": 105.925,
+      "eval_steps_per_second": 13.347,
       "step": 2250
     },
     {
       "epoch": 9.0,
-      "eval_exact_match": 13.0,
-      "eval_f1": 20.096701018279973,
+      "eval_exact_match": 0.0,
+      "eval_f1": 0.18,
       "step": 2250
     },
     {
       "epoch": 9.2,
-      "grad_norm": 0.8059192895889282,
       "learning_rate": 0.0005,
-      "loss": 0.4299,
+      "loss": 0.3424,
       "step": 2300
     },
     {
       "epoch": 9.6,
-      "grad_norm": 0.9573262333869934,
       "learning_rate": 0.0005,
-      "loss": 0.4169,
+      "loss": 0.3418,
       "step": 2400
     },
     {
       "epoch": 10.0,
-      "grad_norm": 1.1837812662124634,
       "learning_rate": 0.0005,
-      "loss": 0.4481,
+      "loss": 0.3607,
       "step": 2500
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.49133333333333334,
-      "eval_loss": 3.5352694988250732,
-      "eval_runtime": 5.1504,
-      "eval_samples_per_second": 97.08,
-      "eval_steps_per_second": 12.232,
+      "eval_accuracy": 0.5866835443037974,
+      "eval_loss": 2.8513779640197754,
+      "eval_runtime": 4.5658,
+      "eval_samples_per_second": 109.511,
+      "eval_steps_per_second": 13.798,
       "step": 2500
     },
     {
       "epoch": 10.0,
-      "eval_exact_match": 14.0,
-      "eval_f1": 20.9187062937063,
+      "eval_exact_match": 15.8,
+      "eval_f1": 24.03880563880564,
       "step": 2500
     },
     {
       "epoch": 10.4,
-      "grad_norm": 1.1431084871292114,
       "learning_rate": 0.0005,
-      "loss": 0.3986,
+      "loss": 0.339,
       "step": 2600
     },
     {
       "epoch": 10.8,
-      "grad_norm": 1.0149098634719849,
       "learning_rate": 0.0005,
-      "loss": 0.4166,
+      "loss": 0.3414,
       "step": 2700
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.4893968253968254,
-      "eval_loss": 3.493849992752075,
-      "eval_runtime": 5.9644,
-      "eval_samples_per_second": 83.831,
-      "eval_steps_per_second": 10.563,
+      "eval_accuracy": 0.5860759493670886,
+      "eval_loss": 2.8931641578674316,
+      "eval_runtime": 5.1473,
+      "eval_samples_per_second": 97.139,
+      "eval_steps_per_second": 12.24,
       "step": 2750
     },
     {
       "epoch": 11.0,
-      "eval_exact_match": 14.0,
-      "eval_f1": 21.90111555111556,
+      "eval_exact_match": 0.0,
+      "eval_f1": 1.8004313814840132,
       "step": 2750
     },
     {
       "epoch": 11.2,
-      "grad_norm": 0.666008710861206,
       "learning_rate": 0.0005,
-      "loss": 0.4154,
+      "loss": 0.3302,
       "step": 2800
     },
     {
       "epoch": 11.6,
-      "grad_norm": 0.8676968812942505,
       "learning_rate": 0.0005,
-      "loss": 0.4081,
+      "loss": 0.329,
       "step": 2900
     },
     {
       "epoch": 12.0,
-      "grad_norm": 0.8371489644050598,
       "learning_rate": 0.0005,
-      "loss": 0.429,
+      "loss": 0.3439,
       "step": 3000
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.49057142857142855,
-      "eval_loss": 3.544997453689575,
-      "eval_runtime": 5.1712,
-      "eval_samples_per_second": 96.689,
-      "eval_steps_per_second": 12.183,
+      "eval_accuracy": 0.5854683544303797,
+      "eval_loss": 2.9544754028320312,
+      "eval_runtime": 4.5673,
+      "eval_samples_per_second": 109.473,
+      "eval_steps_per_second": 13.794,
       "step": 3000
     },
     {
       "epoch": 12.0,
-      "eval_exact_match": 13.0,
-      "eval_f1": 20.847710622710633,
+      "eval_exact_match": 0.8,
+      "eval_f1": 3.4374976657329603,
       "step": 3000
     },
     {
       "epoch": 12.4,
-      "grad_norm": 0.8826413750648499,
       "learning_rate": 0.0005,
-      "loss": 0.3877,
+      "loss": 0.3182,
       "step": 3100
     },
     {
       "epoch": 12.8,
-      "grad_norm": 0.9569880962371826,
       "learning_rate": 0.0005,
-      "loss": 0.4193,
+      "loss": 0.335,
       "step": 3200
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.4881587301587302,
-      "eval_loss": 3.5636065006256104,
-      "eval_runtime": 5.3238,
-      "eval_samples_per_second": 93.917,
-      "eval_steps_per_second": 11.834,
+      "eval_accuracy": 0.5842784810126582,
+      "eval_loss": 2.899094343185425,
+      "eval_runtime": 4.7188,
+      "eval_samples_per_second": 105.96,
+      "eval_steps_per_second": 13.351,
       "step": 3250
     },
     {
       "epoch": 13.0,
-      "eval_exact_match": 12.4,
-      "eval_f1": 19.792212787212797,
+      "eval_exact_match": 14.4,
+      "eval_f1": 23.30256743256743,
       "step": 3250
     }
   ],
   "logging_steps": 100,
   "max_steps": 12500,
-  "num_input_tokens_seen": 0,
   "num_train_epochs": 50,
   "save_steps": 500,
-  "stateful_callbacks": {
-    "TrainerControl": {
-      "args": {
-        "should_epoch_stop": false,
-        "should_evaluate": false,
-        "should_log": false,
-        "should_save": true,
-        "should_training_stop": false
-      },
-      "attributes": {}
-    }
-  },
-  "total_flos": 9.941801586655232e+16,
-  "train_batch_size": 2,
+  "total_flos": 2.0780698119543194e+17,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-3250/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b62f42960cf4fd93abfb3be317e2e8ef04964ff17a18b3862335bf3c3ff7041
-size 5304
+oid sha256:814f378bbfbf215f6c033d7b267886233699f8ce3ebf0a109279893fc2f542bf
+size 4728