Initial upload of st0 RoBERTa-large binary causal classifier

classification_report.txt

@@ -0,0 +1,16 @@
+              precision    recall  f1-score   support
+
+           0       0.90      0.80      0.85       100
+           1       0.96      0.98      0.97       532
+
+    accuracy                           0.95       632
+   macro avg       0.93      0.89      0.91       632
+weighted avg       0.95      0.95      0.95       632
+
+# Notes
+# Best Model - Test Accuracy: 0.9541
+# Best epoch: 3 (val F1 0.9840)
+# Model: roberta-large, 10 epochs, binary single-label classification
+# Train/Dev/Test rows: 3396 / 627 / 632
+# Label semantics: 0 = no_relation, 1 = causal (positive)
+# Train label dist: 1=0.9167, 0=0.0833

config.json

@@ -0,0 +1,28 @@
+{
+  "_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",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 50265
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e4e46b1f3c294f970f7e7f0290588fb0b54d017e7f84209749c32eabaeb92e4
+size 1421495416

special_tokens_map.json

@@ -0,0 +1,15 @@
+{
+  "bos_token": "<s>",
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "unk_token": "<unk>"
+}

tokenizer_config.json

@@ -0,0 +1,57 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "50264": {
+      "content": "<mask>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "errors": "replace",
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "RobertaTokenizer",
+  "trim_offsets": true,
+  "unk_token": "<unk>"
+}

training_log.txt

@@ -0,0 +1,112 @@
+st0 — Relation_detection.py (binary causal-or-not)
+Model: roberta-large
+Data: Combined_dataset_CommonSense+News_Data
+Train/Dev/Test: 3396 / 627 / 632
+
+Label distribution
+  train: 1=0.9167, 0=0.0833
+  dev:   1=0.7879, 0=0.2121
+  test:  1=0.8418, 0=0.1582
+
+================================================================
+Epoch 1/10, Train Loss: 0.0438, Validation Loss: 0.4460, Validation Accuracy: 0.9282
+              precision    recall  f1-score   support
+           0       0.98      0.68      0.80       133
+           1       0.92      1.00      0.96       494
+    accuracy                           0.93       627
+   macro avg       0.95      0.84      0.88       627
+weighted avg       0.93      0.93      0.92       627
+Validation F1-score: 0.9563
+Saved best model to /content/drive/MyDrive/causalsense/checkpoints/st0_roberta_large (val F1=0.9563)
+
+Epoch 2/10, Train Loss: 0.0246, Validation Loss: 0.9863, Validation Accuracy: 0.8788
+              precision    recall  f1-score   support
+           0       1.00      0.43      0.60       133
+           1       0.87      1.00      0.93       494
+    accuracy                           0.88       627
+   macro avg       0.93      0.71      0.76       627
+weighted avg       0.89      0.88      0.86       627
+Validation F1-score: 0.9286
+
+Epoch 3/10, Train Loss: 0.0442, Validation Loss: 0.1793, Validation Accuracy: 0.9745
+              precision    recall  f1-score   support
+           0       0.98      0.89      0.94       133
+           1       0.97      1.00      0.98       494
+    accuracy                           0.97       627
+   macro avg       0.98      0.95      0.96       627
+weighted avg       0.97      0.97      0.97       627
+Validation F1-score: 0.9840
+Saved best model to /content/drive/MyDrive/causalsense/checkpoints/st0_roberta_large (val F1=0.9840)  *** BEST ***
+
+Epoch 4/10, Train Loss: 0.0134, Validation Loss: 0.6935, Validation Accuracy: 0.9187
+              precision    recall  f1-score   support
+           0       1.00      0.62      0.76       133
+           1       0.91      1.00      0.95       494
+    accuracy                           0.92       627
+   macro avg       0.95      0.81      0.86       627
+weighted avg       0.93      0.92      0.91       627
+Validation F1-score: 0.9509
+
+Epoch 5/10, Train Loss: 0.0058, Validation Loss: 0.7291, Validation Accuracy: 0.9123
+              precision    recall  f1-score   support
+           0       1.00      0.59      0.74       133
+           1       0.90      1.00      0.95       494
+    accuracy                           0.91       627
+   macro avg       0.95      0.79      0.84       627
+weighted avg       0.92      0.91      0.90       627
+Validation F1-score: 0.9473
+
+Epoch 6/10, Train Loss: 0.0004, Validation Loss: 0.8895, Validation Accuracy: 0.9075
+              precision    recall  f1-score   support
+           0       1.00      0.56      0.72       133
+           1       0.89      1.00      0.94       494
+    accuracy                           0.91       627
+   macro avg       0.95      0.78      0.83       627
+weighted avg       0.92      0.91      0.90       627
+Validation F1-score: 0.9446
+
+Epoch 7/10, Train Loss: 0.0031, Validation Loss: 0.6923, Validation Accuracy: 0.9155
+              precision    recall  f1-score   support
+           0       1.00      0.60      0.75       133
+           1       0.90      1.00      0.95       494
+    accuracy                           0.92       627
+   macro avg       0.95      0.80      0.85       627
+weighted avg       0.92      0.92      0.91       627
+Validation F1-score: 0.9491
+
+Epoch 8/10, Train Loss: 0.0003, Validation Loss: 0.6224, Validation Accuracy: 0.9362
+              precision    recall  f1-score   support
+           0       1.00      0.70      0.82       133
+           1       0.93      1.00      0.96       494
+    accuracy                           0.94       627
+   macro avg       0.96      0.85      0.89       627
+weighted avg       0.94      0.94      0.93       627
+Validation F1-score: 0.9611
+
+Epoch 9/10, Train Loss: 0.0001, Validation Loss: 0.6299, Validation Accuracy: 0.9362
+              precision    recall  f1-score   support
+           0       1.00      0.70      0.82       133
+           1       0.93      1.00      0.96       494
+    accuracy                           0.94       627
+   macro avg       0.96      0.85      0.89       627
+weighted avg       0.94      0.94      0.93       627
+Validation F1-score: 0.9611
+
+Epoch 10/10, Train Loss: 0.0009, Validation Loss: 0.8330, Validation Accuracy: 0.9091
+              precision    recall  f1-score   support
+           0       1.00      0.57      0.73       133
+           1       0.90      1.00      0.95       494
+    accuracy                           0.91       627
+   macro avg       0.95      0.79      0.84       627
+weighted avg       0.92      0.91      0.90       627
+Validation F1-score: 0.9455
+
+================================================================
+Best Model - Test Accuracy: 0.9541
+Best Model - Test Classification Report:
+              precision    recall  f1-score   support
+           0       0.90      0.80      0.85       100
+           1       0.96      0.98      0.97       532
+    accuracy                           0.95       632
+   macro avg       0.93      0.89      0.91       632
+weighted avg       0.95      0.95      0.95       632