Update app.py

app.py

@@ -8,16 +8,16 @@ from PIL import ImageDraw
 from torchvision.transforms.v2 import Resize
 
 import subprocess
-# 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_id = "vikhyatk/moondream2"
 revision = "2024-05-20"
 tokenizer = AutoTokenizer.from_pretrained(model_id, revision=revision)
 moondream = AutoModelForCausalLM.from_pretrained(
     model_id, trust_remote_code=True, revision=revision,
-    # torch_dtype=torch.bfloat16, device_map={"": "cuda"}
-    torch_dtype=torch.float32, device_map="cpu"
-    # attn_implementation="flash_attention_2"
+    torch_dtype=torch.bfloat16, device_map={"": "cuda"},
+    # torch_dtype=torch.float32, device_map="cpu"
+    attn_implementation="flash_attention_2"
 )
 moondream.eval()