search

RetroCirce / HTS-Audio-Transformer

The official code repo of "HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound Classification and Detection"
https://arxiv.org/abs/2202.00874
MIT License
341 stars 62 forks source link

upsample_bicubic2d_backward_out_cuda #33

Closed DuQingChen closed 1 year ago

DuQingChen commented 1 year ago

When model trained on esc-50, run into this error: RuntimeError: upsample_bicubic2d_backward_out_cuda does not have a deterministic implementation, but you set 'torch.use_deterministic_algorithms(True)'.

If set torch.use_deterministic_algorithms =False, acc can only achieve 0.65.

env : torch 1.12.1 pytorch-lightning 1.5.9

DuQingChen commented 1 year ago

https://github.com/RetroCirce/HTS-Audio-Transformer/issues/13