Error Message "max_pool2d_with_indices_out_cuda_frame failed with error code 0" Displayed in Logs

Symptom

"RuntimeError:max_pool2d_with_indices_out_cuda_frame failed with error code 0"

Possible Causes

The PyTorch 1.4 engine is incompatible with that of PyTorch 1.3.

Solution

1. Run the following commands to add contiguous data:

   images = images.cuda()  
   pred = model(images.permute(0, 3, 1, 2).contigous())

2. Roll back to PyTorch 1.3.
3. Use the local PyCharm to remotely access notebook for debugging.

Summary and Suggestions

Before creating a training job, use the ModelArts development environment to debug the training code to maximally eliminate errors in code migration.