Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -29,6 +29,7 @@ model = AutoModelForVision2Seq.from_pretrained(
|
|
29 |
MODEL_ID,
|
30 |
token=hf_token,
|
31 |
torch_dtype=torch.bfloat16,
|
|
|
32 |
).to("cuda") # Ensure model loads on CUDA for Spaces
|
33 |
|
34 |
@spaces.GPU # Add spaces.GPU decorator for GPU acceleration
|
|
|
29 |
MODEL_ID,
|
30 |
token=hf_token,
|
31 |
torch_dtype=torch.bfloat16,
|
32 |
+
_attn_implementation="flash_attention_2" if DEVICE == "cuda" else "eager",
|
33 |
).to("cuda") # Ensure model loads on CUDA for Spaces
|
34 |
|
35 |
@spaces.GPU # Add spaces.GPU decorator for GPU acceleration
|