Issue about the app.py

[tuning12]

Thank you for your great work.
when running the inpainting example in gradio, an error is occured:

CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

[tuning12]

Fail to install SAM2. Is there any solution?

[yxbian23]

Thank you for your great work. when running the inpainting example in gradio, an error is occured:

CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

This seems more likely a hardware problem. I haven't encountered such an issue before~ May be you can run the train/infer code in our repo to double-check your hardware works well.

[yxbian23]

Fail to install SAM2. Is there any solution?

Oh, I think this also may come from your hardware problem or a wrong CUDA installation. You can refer to SAM2 Official repo to debug.

[tuning12]

Actually I follow installation guidance and double-check the version of cuda. This problem still exists.
SAM is too difficult to install...

[amerk12]

For what it's worth also experiencing the same CUDA error in the gradio app. CUDA 12.4 here.

RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

[li-wenzhan]

I think the reason for the SAM-2 installation failure may be that the official setup.py file has been updated, refering to https://github.com/facebookresearch/sam2/blob/main/setup.py.

[beautifulchoi]

Hi, did you solve this issue?

[beautifulchoi]

I finally ran it after 1 hour...
what I did is,

1. go to sam2 official repo and, git clone + pip install -e .
2. delete _C file
3. run again