Spaces:

qubvel-hf
/

vjepa2-streaming-video-classification

Running on L4

qubvel-hf HF Staff commited on 2 days ago

Commit

5893d39

verified ·

1 Parent(s): 42718c3

Update app.py

Files changed (1) hide show

app.py CHANGED Viewed

@@ -30,7 +30,7 @@ from transformers import VJEPA2ForVideoClassification, AutoVideoProcessor
 # Model configuration
-CHECKPOINT = "qubvel-hf/vjepa2-vitl-fpc16-256-ssv2"  # Pre-trained VJEPA2 model checkpoint
 TORCH_DTYPE = torch.float16  # Use half precision for faster inference
 TORCH_DEVICE = "cuda"  # Use GPU for inference
 UPDATE_EVERY_N_FRAMES = 64  # How often to update predictions (in frames)

 # Model configuration
+CHECKPOINT = "facebook/vjepa2-vitl-fpc16-256-ssv2"  # Pre-trained VJEPA2 model checkpoint
 TORCH_DTYPE = torch.float16  # Use half precision for faster inference
 TORCH_DEVICE = "cuda"  # Use GPU for inference
 UPDATE_EVERY_N_FRAMES = 64  # How often to update predictions (in frames)