update

config.json

@@ -21,7 +21,7 @@
   "template_version": "base",
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.41.0",
+  "transformers_version": "4.43.1",
   "use_cache": true,
   "vision_config": {
     "dropout_prob": 0.0,

generation_config.json

@@ -7,5 +7,5 @@
   "max_length": 2048,
   "top_p": 0.1,
   "top_k": 1,
-  "transformers_version": "4.41.0"
+  "transformers_version": "4.43.1"
 }

modeling_cogvlm.py

@@ -780,9 +780,11 @@ class CogVLMVideoForCausalLM(CogVLMPreTrainedModel):
             standardize_cache_format: bool = False,
     ) -> Dict[str, Any]:
         # update past_key_values
-        model_kwargs["past_key_values"] = self._extract_past_from_model_output(
+        cache_name, cache = self._extract_past_from_model_output(
             outputs, standardize_cache_format=standardize_cache_format
         )
+        model_kwargs[cache_name] = cache
+
         if getattr(outputs, "state", None) is not None:
             model_kwargs["state"] = outputs.state