Update app.py

app.py

@@ -41,27 +41,27 @@ try:
     rag = RAGWithCitations(model_path_or_name=MODEL_CACHE_DIR)
     
     # Fix the warnings by properly configuring generation parameters
-    if hasattr(rag, "model"):
+#    if hasattr(rag, "model"):
         # Configure tokenizer
-        if hasattr(rag, "tokenizer"):
-            if rag.tokenizer.pad_token is None:
-                rag.tokenizer.pad_token = rag.tokenizer.eos_token
-            rag.tokenizer.padding_side = "left"  # For batch generation
+#        if hasattr(rag, "tokenizer"):
+#            if rag.tokenizer.pad_token is None:
+#                rag.tokenizer.pad_token = rag.tokenizer.eos_token
+#            rag.tokenizer.padding_side = "left"  # For batch generation
             
         # Configure model generation settings
-        rag.model.config.pad_token_id = rag.tokenizer.pad_token_id
-        rag.model.generation_config.pad_token_id = rag.tokenizer.pad_token_id
+#        rag.model.config.pad_token_id = rag.tokenizer.pad_token_id
+#        rag.model.generation_config.pad_token_id = rag.tokenizer.pad_token_id
         
         # Fix the do_sample/top_p warning
-        rag.model.generation_config.do_sample = True
-        rag.model.generation_config.top_p = 0.95  # Explicitly set to match warning
+#        rag.model.generation_config.do_sample = True
+#        rag.model.generation_config.top_p = 0.95  # Explicitly set to match warning
         
         # Configure attention mask handling
-        rag.model.config.use_cache = True
+#        rag.model.config.use_cache = True
         
-        log_debug("✅ Model loaded successfully with configuration:")
-        log_debug(f" - Pad token: {rag.tokenizer.pad_token} (ID: {rag.tokenizer.pad_token_id})")
-        log_debug(f" - Generation config: {rag.model.generation_config}")
+#        log_debug("✅ Model loaded successfully with configuration:")
+#        log_debug(f" - Pad token: {rag.tokenizer.pad_token} (ID: {rag.tokenizer.pad_token_id})")
+#        log_debug(f" - Generation config: {rag.model.generation_config}")
         
 except Exception as e:
     log_debug(f"❌ Model initialization failed: {str(e)}")
@@ -89,7 +89,13 @@ log_debug("📄 Test Sources loaded successfully.")
 # Generate a response
 try:
     log_debug("🧠 Test rag model on simple example...")
-    response = rag.generate(query, sources)
+    response = rag.generate(query, 
+                            sources,
+                            do_sample=True,  # Enable sampling
+                            top_p=0.95,     # Set top_p for nucleus sampling
+                            pad_token_id=rag.tokenizer.eos_token_id,  # Set pad_token_id to eos_token_id
+                            attention_mask=None  # Ensure attention_mask is passed if needed
+                           )
     log_debug("✅ Test Answer generated successfully.")
     log_debug(response["processed"]["clean_answer"])
 except Exception as e: