Update models/videochat.py

models/videochat.py

@@ -109,7 +109,7 @@ class VideoChat(Blip2Base):
         print('Loading Q-Former Done')
 
         print('Loading LLAMA')
-        self.llama_tokenizer = LlamaTokenizer.from_pretrained(llama_model_path, use_fast=False)
+        self.llama_tokenizer = LlamaTokenizer.from_pretrained(llama_model_path, use_fast=False, use_auth_token=os.environ["HF_TOKEN"])
         self.llama_tokenizer.pad_token = self.llama_tokenizer.eos_token
 
         if self.low_resource:
@@ -117,12 +117,14 @@ class VideoChat(Blip2Base):
                     llama_model_path,
                     torch_dtype=torch.float16,
                     load_in_8bit=True,
-                    device_map="auto"
+                    device_map="auto",
+                    use_auth_token=os.environ["HF_TOKEN"],
                 )
         else:
                 self.llama_model = LlamaForCausalLM.from_pretrained(
                     llama_model_path,
                     torch_dtype=torch.float16,
+                    use_auth_token=os.environ["HF_TOKEN"],
                 )
 
         print("freeze LLAMA")