Aifred-Health/VulcanAI

[CUDA/GPU] tensor to numpy handling when CUDA present

Closed this issue · 1 comments

priyatharsan commented

  • in metrics: apparently the incoming targets to the function get_score is torch.Tensor sometimes which needs to be brought to cpu and then converted to numpy when calculating metric_score
    1

  • Receptive fields: when CUDA present, the torch.Tensor must be moved to cpu before converting to numpy
    2

priyatharsan commented

this issue is resolved in #113