Return with exit code when converting quantized-pytorch to onnx

[wcycqjy]

When I try to convert an INT8 quantized pytorch model to the onnx version, the error returns:
Process finished with exit code -1073741819 (0xC0000005).

No matter which export function I use, like the torch.onnx.export, and you provided export_to_onnx, the error always exists.

My os is Windows, the gpu is A4000.

[i-riyad]

Error code 0xC0000005 indicates a Windows access violation (segmentation fault), typically due to invalid memory access. In your case, PyTorch INT8-quantized models may include operators not fully supported by the ONNX exporter. To isolate the problem, try exporting a non-quantized model first, or export with different opset versions (e.g., 13, 14, 17). If the issue persists, please share a minimal quantization script and model snippet so we can reproduce it.