Update app.py

app.py

@@ -42,15 +42,6 @@ from transformers import (
 from PIL import ImageDraw
 from torchvision.transforms.v2 import Resize
 
-if IN_SPACES:
-    import subprocess
-
-    subprocess.run(
-        "pip install flash-attn --no-build-isolation",
-        env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
-        shell=True,
-    )
-
 os.environ["HF_TOKEN"] = os.environ.get("TOKEN_FROM_SECRET") or True
 tokenizer = AutoTokenizer.from_pretrained("vikhyatk/moondream-next", revision=REVISION)
 moondream = AutoModelForCausalLM.from_pretrained(
@@ -58,7 +49,6 @@ moondream = AutoModelForCausalLM.from_pretrained(
     trust_remote_code=True,
     torch_dtype=torch.float16,
     device_map={"": "cuda"},
-    attn_implementation="flash_attention_2",
     revision=REVISION
 )