RuntimeError: CUDA error: CUDA-capable device(s) is/are busy or unavailable 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.

[NyanNat]

trafficstars

When I tried to load starcoder based on tutorial provided, RuntimeError emerges because of CUDA error, but torch.cuda.is_available() returns True. The gpu that I run this on is provided below, Can anyone help me with this ?