Delete main.py

main.py

@@ -1,55 +0,0 @@
-from model import load_model, load_tokenizer
-from utils import clean_output
-import torch
-import shap
-from huggingface_hub import login
-
-def generate_questions(context, num_questions=3, max_length=64):
-    tokenizer = load_tokenizer()
-    model = load_model()
-
-    input_prompt = f"generate question: {context.strip()}"
-    inputs = tokenizer(input_prompt, return_tensors="pt", truncation=True, padding="longest").to(model.device)
-
-    outputs = model.generate(
-        input_ids=inputs["input_ids"],
-        attention_mask=inputs["attention_mask"],
-        max_length=max_length,
-        num_return_sequences=num_questions,
-        do_sample=True,
-        top_p=0.95,
-        temperature=1.0
-    )
-
-    decoded = tokenizer.batch_decode(outputs, skip_special_tokens=True)
-    return clean_output(decoded)
-
-def get_shap_values(tokenizer, model, prompt):
-    """
-    Compute SHAP token attributions for a given prompt.
-    """
-    # Define wrapper prediction function
-    def f(texts):
-        # Tokenize the list of texts
-        inputs = tokenizer(list(texts), return_tensors="pt", truncation=True, padding=True).to(model.device)
-        with torch.no_grad():
-            out = model.generate(
-                input_ids=inputs["input_ids"],
-                attention_mask=inputs["attention_mask"],
-                max_length=64,
-                do_sample=False,
-                num_beams=2
-            )
-        # Return something SHAP can use (e.g., output logits or decoded text)
-        # Here, we return the first token's id for each output as a simple example
-        return out[:, 0].detach().cpu().numpy()
-
-    explainer = shap.Explainer(f, tokenizer)
-    shap_values = explainer([prompt])
-
-    # Get tokens for visualization
-    tokens = tokenizer.convert_ids_to_tokens(tokenizer(prompt, return_tensors="pt")["input_ids"][0])
-    return shap_values.values[0], tokens
-
-
-