Update fine_tune_inference_test.py

fine_tune_inference_test.py

@@ -94,18 +94,23 @@ def chat(msg: Message):
         log(f"🔢 Tokenizer input_ids: {inputs['input_ids']}")
         log(f"📏 input shape: {inputs['input_ids'].shape}")
 
-        with torch.no_grad():
-            output = model.generate(
-                **inputs,
-                max_new_tokens=200,
-                do_sample=True,
-                temperature=0.7,
-                top_k=50,
-                top_p=0.95,
-                return_dict_in_generate=True,
-                output_scores=True,
-                suppress_tokens=[tokenizer.pad_token_id] if tokenizer.pad_token_id is not None else None
-            )
+        try:
+            with torch.no_grad():
+                output = model.generate(
+                    **inputs,
+                    max_new_tokens=200,
+                    do_sample=True,
+                    temperature=0.7,
+                    top_k=50,
+                    top_p=0.95,
+                    return_dict_in_generate=True,
+                    output_scores=True,
+                    suppress_tokens=[tokenizer.pad_token_id] if tokenizer.pad_token_id is not None else None
+                )
+        except Exception as e:
+            log(f"⚠️ generate() sırasında hata: {e}")
+            fallback = random.choice(FALLBACK_ANSWERS)
+            return {"answer": fallback, "chat_history": chat_history}
 
         generated_ids = output.sequences[0]
         generated_text = tokenizer.decode(generated_ids, skip_special_tokens=True)