Update app.py

app.py

@@ -7,7 +7,7 @@ import subprocess
 subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
 models = {
-    "microsoft/Phi-3.5-vision-instruct": AutoModelForCausalLM.from_pretrained("microsoft/Phi-3.5-vision-instruct", revision="main", trust_remote_code=True, torch_dtype="auto", _attn_implementation="default").cuda().eval()
+    "microsoft/Phi-3.5-vision-instruct": AutoModelForCausalLM.from_pretrained("microsoft/Phi-3.5-vision-instruct", trust_remote_code=True, torch_dtype="auto", _attn_implementation="flash_attention_2").cuda().eval()
 
 }