Update custom model files, README, and requirements

asr_modeling.py

@@ -622,13 +622,18 @@ class ASRModel(PreTrainedModel, GenerationMixin):
                 user_content += " " + self.TRANSCRIBE_PROMPT
             messages.append({"role": "user", "content": user_content})
 
+            enable_thinking_val = getattr(self.config, "enable_thinking", False)
+            print(f"[DEBUG generate] enable_thinking={enable_thinking_val}, system_prompt={system_prompt[:100] if system_prompt else None}...")
             chat_result = self.tokenizer.apply_chat_template(
                 messages,
                 tokenize=True,
                 add_generation_prompt=True,
                 return_tensors="pt",
-                enable_thinking=getattr(self.config, "enable_thinking", False),
+                enable_thinking=enable_thinking_val,
             )
+            # Debug: show the formatted prompt
+            prompt_text = self.tokenizer.decode(chat_result.input_ids[0] if chat_result.input_ids.dim() > 1 else chat_result.input_ids)
+            print(f"[DEBUG generate] Formatted prompt: {prompt_text[:500]}...")
             input_ids = chat_result.input_ids.to(device)
 
             if input_ids.dim() == 1: