Update app.py

app.py

@@ -11,11 +11,11 @@ import os
 from moviepy.editor import VideoFileClip 
 
 # Install flash-attn
-# subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
+subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
 # Model and Processor Loading (Done once at startup)
 MODEL_ID = "Qwen/Qwen2-VL-7B-Instruct"
-model = Qwen2VLForConditionalGeneration.from_pretrained(MODEL_ID, trust_remote_code=True, torch_dtype=torch.bfloat16, device_map="auto").cuda().eval()
+model = Qwen2VLForConditionalGeneration.from_pretrained(MODEL_ID, trust_remote_code=True, torch_dtype=torch.float16).to("cuda").eval()
 processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
 
 DESCRIPTION = "[Qwen2-VL-7B Demo](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct)"