Upload 10 files

Causal Language Classifier

all_thresholds.json

@@ -0,0 +1,20 @@
+{
+  "f1": {
+    "threshold": 0.3500000000000001,
+    "metrics": {
+      "accuracy": 0.8336666666666667,
+      "precision": 0.7284546805349182,
+      "recall": 0.88016157989228,
+      "f1": 0.7971544715447154
+    }
+  },
+  "balanced": {
+    "threshold": 0.5500000000000002,
+    "metrics": {
+      "accuracy": 0.847,
+      "precision": 0.7971869328493648,
+      "recall": 0.7885996409335727,
+      "f1": 0.7928700361010831
+    }
+  }
+}

config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "KM4STfulltext/SSCI-BERT-e4",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.48.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 28996
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d83ffc7765f5797aefd4327502bda64752563cd0ddd2c8ed025f2902e875548f
+size 433270768

predict_ft_model_causal.py

@@ -0,0 +1,90 @@
+def predict_with_threshold(text, model_dir="causal-classifier", threshold=None, threshold_type=None):
+    """
+    Make a prediction using the saved model and custom threshold.
+    
+    Args:
+        text: Input text to classify
+        model_dir: Directory where the model and threshold are saved
+        threshold: Custom threshold to use (if None, loads from saved config)
+        threshold_type: Type of threshold to use ('f1' or 'balanced') if loading from all_thresholds.json
+        
+    Returns:
+        Dictionary with prediction results
+    """
+    # Load the model and tokenizer
+    from transformers import AutoModelForSequenceClassification, AutoTokenizer
+    import torch
+    import json
+    
+    tokenizer = AutoTokenizer.from_pretrained(model_dir)
+    model = AutoModelForSequenceClassification.from_pretrained(model_dir)
+    
+    threshold_source = "custom"
+    
+    # Load the threshold if not provided
+    if threshold is None:
+        # Try multiple methods to get the threshold
+        if threshold_type is not None:
+            # First try all_thresholds.json if threshold_type is specified
+            try:
+                with open(f"{model_dir}/all_thresholds.json", "r") as f:
+                    all_thresholds = json.load(f)
+                    if threshold_type in all_thresholds:
+                        threshold = all_thresholds[threshold_type]["threshold"]
+                        threshold_source = f"all_thresholds.json ({threshold_type})"
+                    else:
+                        print(f"Threshold type '{threshold_type}' not found. Available types: {list(all_thresholds.keys())}")
+            except FileNotFoundError:
+                pass
+        
+        # If still no threshold, try threshold_config.json
+        if threshold is None:
+            try:
+                with open(f"{model_dir}/threshold_config.json", "r") as f:
+                    config = json.load(f)
+                    threshold = config["threshold"]
+                    threshold_source = "threshold_config.json"
+            except FileNotFoundError:
+                # Default to 0.5 if no threshold config is found
+                threshold = 0.5
+                threshold_source = "default"
+                print("No threshold configuration found. Using default threshold of 0.5.")
+    
+    # Tokenize the input text
+    inputs = tokenizer(text, padding="max_length", truncation=True, max_length=512, return_tensors="pt")
+    
+    # Get model prediction
+    model.eval()
+    with torch.no_grad():
+        outputs = model(**inputs)
+    
+    # Convert logits to probabilities
+    probs = torch.nn.functional.softmax(outputs.logits, dim=1).squeeze().tolist()
+    
+    # Apply threshold to get the final prediction
+    if isinstance(probs, list):  # Handle batch size of 1
+        prediction = 1 if probs[1] > threshold else 0
+        class_probs = {
+            "Causal": probs[1],
+            "Descriptive": probs[0]
+        }
+    else:  # Handle single prediction case
+        prediction = 1 if probs > threshold else 0
+        class_probs = {
+            "Causal": probs,
+            "Descriptive": 1 - probs
+        }
+    
+    # Map prediction back to original label
+    label_names = {1: "Causal", 0: "Descriptive"}
+    
+    return {
+        "prediction": label_names[prediction],
+        "probabilities": class_probs,
+        "threshold_used": threshold,
+        "threshold_source": threshold_source
+    }
+
+
+sample_text = 'This is a causal study that aims to investigate the relationship between smoking and lung cancer.'
+result = predict_with_threshold(sample_text)

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

threshold_config.json

@@ -0,0 +1,10 @@
+{
+  "threshold_type": "balanced",
+  "threshold": 0.5500000000000002,
+  "metrics": {
+    "accuracy": 0.847,
+    "precision": 0.7971869328493648,
+    "recall": 0.7885996409335727,
+    "f1": 0.7928700361010831
+  }
+}

tokenizer_config.json

@@ -0,0 +1,58 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4159057d9ab821b7ae87bf1469f0a43b53f84ca0c9b6a7a3b71f8376df8a4ea1
+size 5304