First model version

README.md

@@ -1,3 +1,35 @@
 ---
 license: mit
 ---
+
+Base model: [roberta-large](https://huggingface.co/roberta-large)
+
+Fine tuned for persuadee donation detection on the [Persuasion For Good Dataset](https://gitlab.com/ucdavisnlp/persuasionforgood) (Wang et al., 2019):
+
+Given a complete dialogue from Persuasion For Good, the task is to predict the binary label:
+ - 0: the persuadee does not intend to donate
+ - 1: the persuadee intends to donate
+
+Only persuadee utterances are input to the model for this task - persuader utterances are discarded. Each training example is the concatenation of all persuadee utterances in a single dialogue, each separated by the `</s>` token.
+
+For example:
+
+**Input**: `<s>How are you?</s>Can you tell me more about the charity?</s>...</s>Sure, I'll donate a dollar.</s>...</s>`
+
+**Label**: 1
+
+**Input**: `<s>How are you?</s>Can you tell me more about the charity?</s>...</s>I am not interested.</s>...</s>`
+
+**Label**: 0
+
+The following Dialogues were excluded:
+ - 146 dialogues where a donation of 0 was made at the end of the task but a non-zero amount was pledged by the persuadee in the dialogue, per the following regular expression: `(?:\$(?:0\.)?[1-9]|[1-9][.0-9]*?(?: ?\$| dollars?| cents?))`
+ 
+Stats:
+ - **Training set**: 587 dialogues, using actual end-task donations as labels
+ - **Validation set**: 141 dialogues, using manual donation intention labels from Persuasion For Good 'AnnSet'
+ - **Test set**: 143 dialogues, using manual donation intention labels from Persuasion For Good 'AnnSet'
+
+**Test Macro F1**: 0.88
+
+**Test Accuracy**: 0.90

config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "roberta-large",
+  "architectures": [
+    "RobertaForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.11.3",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 50265
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a8a1a99878e330a2b226c429280e2c3d6800342792bf7ee41bae1c3a8cbbc90
+size 1421605805

runs/version_0/events.out.tfevents.1649251721.panacea.34025.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8f6d9f468458d8f0f946f1bb8048f2e0a406d005c6cc426a81fb0ebba953138
+size 21348

runs/version_0/events.out.tfevents.1649252717.panacea.34025.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebef0b5f87d987f5bf4a425df175d60b63f6ef6f623e17d13f94b29dc1d0b1ba
+size 4335

runs/version_0/hparams.yaml

@@ -0,0 +1,74 @@
+accelerator: null
+accumulate_grad_batches: 2
+adam_epsilon: 1.0e-08
+amp_backend: native
+amp_level: null
+auto_lr_find: false
+auto_scale_batch_size: false
+auto_select_gpus: false
+base_modelpath: roberta-large
+batch_size: 16
+benchmark: false
+check_val_every_n_epoch: 1
+checkpoint_callback: null
+checkpoint_save_top_k: 1
+checkpoint_save_weights_only: true
+data_random_state: 42
+datasetpath: ../data/persuasion_for_good/AnnotatedData
+default_root_dir: null
+detect_anomaly: false
+deterministic: false
+devices: null
+early_stopping_patience: 40
+enable_checkpointing: true
+enable_model_summary: true
+enable_progress_bar: true
+fast_dev_run: false
+flush_logs_every_n_steps: null
+gpus: 1
+gradient_clip_algorithm: null
+gradient_clip_val: 1.0
+ipus: null
+learning_rate: 3.0e-05
+limit_predict_batches: 1.0
+limit_test_batches: 1.0
+limit_train_batches: 1.0
+limit_val_batches: 1.0
+log_every_n_steps: 5
+log_gpu_memory: null
+logger: true
+max_epochs: 30
+max_steps: -1
+max_time: null
+min_epochs: null
+min_steps: null
+model_random_state: 42
+move_metrics_to_cpu: false
+multiple_trainloader_mode: max_size_cycle
+num_nodes: 1
+num_processes: 1
+num_sanity_val_steps: 2
+overfit_batches: 0.0
+plugins: null
+precision: 32
+prepare_data_per_node: null
+process_position: 0
+profiler: null
+progress_bar_refresh_rate: null
+reload_dataloaders_every_epoch: false
+reload_dataloaders_every_n_epochs: 0
+replace_sampler_ddp: true
+resume_from_checkpoint: null
+stochastic_weight_avg: false
+strategy: null
+sync_batchnorm: false
+terminate_on_nan: null
+test_split: 0.2
+tpu_cores: null
+track_grad_norm: -1
+val_check_interval: 0.25
+val_split: 0.1
+warmup_steps: 0
+weight_decay: 0.0
+weights_save_path: null
+weights_summary: top

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "<s>", "mask_token": {"content": "<mask>", "single_word": false, "lstrip": true, "rstrip": false, "normalized": false}}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "add_prefix_space": false, "errors": "replace", "sep_token": "</s>", "cls_token": "<s>", "pad_token": "<pad>", "mask_token": "<mask>", "model_max_length": 512, "special_tokens_map_file": null, "name_or_path": "roberta-large", "tokenizer_class": "RobertaTokenizer"}