CUDA kernel error: GTX 1050 (sm_61) not supported by PyTorch 2.8 wheels

[nicolascordoba1]

Environment

• OS: Ubuntu 24.04.3 LTS
• GPU: NVIDIA GeForce GTX 1050 (compute capability 6.1)
• Driver: 550.163.01
• CUDA (via torch): 12.8
• PyTorch: 2.8.0+cu128
• App: digitalsreeni-image-annotator

Error

When running the annotator I get:

torch.AcceleratorError: 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.

Failed to generate prediction

Analysis

• The GTX 1050 is a Pascal GPU (sm_61).
• PyTorch 2.8 (CUDA 12.8 wheels) has dropped support for Pascal/Maxwell. It only supports sm_70 and above.
• This causes the app to fail when attempting to run inference on GPU.

Workaround

Downgrading to an earlier PyTorch release (e.g., 2.4.1 + cu121) that still includes sm_61 works:

pip install --index-url https://download.pytorch.org/whl/cu121 torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1