use flashattn2

demo/infer.py

@@ -35,7 +35,7 @@ class LiveCCDemoInfer:
         self.model = Qwen2VLForConditionalGeneration.from_pretrained(
             model_path, torch_dtype="auto", 
             device_map=f'cuda:{device_id}', 
-            # attn_implementation='flash_attention_2'
+            attn_implementation='flash_attention_2'
         )
         self.processor = AutoProcessor.from_pretrained(model_path, use_fast=False)
         self.streaming_eos_token_id = self.processor.tokenizer(' ...').input_ids[-1]