Updated app.py

app.py

@@ -6,13 +6,15 @@ from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 # Load the fine-tuned model and tokenizer
-model_name = "aarohanverma/text2sql-flan-t5-base-qlora-finetuned"  # Replace with your model repo name
+model_name = "aarohanverma/text2sql_flant5base_finetuned"  # Replace with your model repository name
 model = AutoModelForSeq2SeqLM.from_pretrained(model_name, torch_dtype=torch.bfloat16).to(device)
 tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-base")
 
 def generate_sql(context: str, query: str) -> str:
     """
-    Constructs a prompt using the user-provided context and query, then generates a SQL query.
+    Generates a SQL query given the provided context and natural language query.
+    Constructs a prompt from the inputs, then performs deterministic generation
+    with beam search.
     """
     prompt = f"""Context:
 {context}
@@ -22,17 +24,27 @@ Query:
 
 Response:
 """
+    # Tokenize the prompt and move to device
     inputs = tokenizer(prompt, return_tensors="pt").to(device)
+    
+    # Ensure decoder_start_token_id is set for encoder-decoder generation
+    if model.config.decoder_start_token_id is None:
+        model.config.decoder_start_token_id = tokenizer.pad_token_id
+
+    # Generate the SQL output
     generated_ids = model.generate(
         input_ids=inputs["input_ids"],
+        decoder_start_token_id=model.config.decoder_start_token_id,
         max_new_tokens=250,
         temperature=0.0,      # Deterministic output
-        num_beams=3,          # Beam search for quality output
-        early_stopping=True,
+        num_beams=3,          # Beam search for improved quality
+        early_stopping=True,  # Stop when output is complete
     )
+    
+    # Decode and return the generated SQL statement
     return tokenizer.decode(generated_ids[0], skip_special_tokens=True)
 
-# Create a Gradio interface with two input boxes: one for context, one for query.
+# Create Gradio interface with two input boxes: one for context and one for query
 iface = gr.Interface(
     fn=generate_sql,
     inputs=[