Update app.py

app.py

@@ -74,15 +74,15 @@ OBJ_ID = 0
 sam2_checkpoint = "checkpoints/edgetam.pt"
 model_cfg = "edgetam.yaml"
 predictor = build_sam2_video_predictor(model_cfg, sam2_checkpoint, device="cpu")
-predictor.to("cuda")
+predictor.to("cpu")
 print("predictor loaded")
 
 # use bfloat16 for the entire demo
-torch.autocast(device_type="cuda", dtype=torch.bfloat16).__enter__()
-if torch.cuda.get_device_properties(0).major >= 8:
-    # turn on tfloat32 for Ampere GPUs (https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices)
-    torch.backends.cuda.matmul.allow_tf32 = True
-    torch.backends.cudnn.allow_tf32 = True
+torch.autocast(device_type="cpu", dtype=torch.bfloat16).__enter__()
+# if torch.cuda.get_device_properties(0).major >= 8:
+#     # turn on tfloat32 for Ampere GPUs (https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices)
+#     torch.backends.cuda.matmul.allow_tf32 = True
+#     torch.backends.cudnn.allow_tf32 = True
 
 
 def get_video_fps(video_path):
@@ -189,7 +189,7 @@ def preprocess_video_in(video_path, session_state):
     ]
 
 
-@spaces.GPU
+
 def segment_with_points(
     point_type,
     session_state,
@@ -244,7 +244,7 @@ def segment_with_points(
     mask_image = show_mask((out_mask_logits[0] > 0.0).cpu().numpy())
     first_frame_output = Image.alpha_composite(transparent_background, mask_image)
 
-    torch.cuda.empty_cache()
+    # torch.cuda.empty_cache()
     return selected_point_map, first_frame_output, session_state
 
 
@@ -303,7 +303,7 @@ def propagate_to_all(
         output_frame = np.array(output_frame)
         output_frames.append(output_frame)
 
-    torch.cuda.empty_cache()
+    # torch.cuda.empty_cache()
 
     # Create a video clip from the image sequence
     original_fps = get_video_fps(video_in)