checkpoint update

config.json

@@ -1,13 +1,18 @@
 {
+  "architectures": [
+    "SpeechLLMModel"
+  ],
   "audio_enc_dim": 1024,
   "audio_encoder_name": "microsoft/wavlm-large",
   "audio_processor_name": "microsoft/wavlm-base",
   "auto_map": {
-    "AutoConfig": "config.SpeechLLMModelConfig"
+    "AutoConfig": "config.SpeechLLMModelConfig",
+    "AutoModel": "model.SpeechLLMModel"
   },
   "llm_dim": 2048,
   "llm_model_checkpoint": "hf_repo/llm_model_checkpoint",
   "llm_model_name": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
   "model_type": "custom_model",
+  "torch_dtype": "float32",
   "transformers_version": "4.41.2"
 }

model.py

@@ -114,6 +114,7 @@ Output:'''
             out = self.llm_model.generate(
                 inputs_embeds=combined_embeds,
                 max_new_tokens=max_new_tokens,
+                pad_token_id=self.llm_tokenizer.pad_token_id
             ).cpu().tolist()[0]
 
             output_text = self.llm_tokenizer.decode(out, skip_special_tokens=True)