search

clovaai / voxceleb_trainer

In defence of metric learning for speaker recognition
MIT License
1.02k stars 272 forks source link

get error when use --mixedprec get #116

Closed jingxuan9862 closed 2 years ago

jingxuan9862 commented 3 years ago

NotImplementedError: Could not run 'aten::_amp_foreach_non_finite_check_andunscale' with arguments from the 'CPU' backend. This could be because the operator doesn't exist for this backend, or w as omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutio ns. 'aten::_amp_foreach_non_finite_check_andunscale' is only available for these backends: [CUDA, BackendSelect, Named, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, UN KNOWN_TENSOR_TYPE_ID, AutogradMLC, AutogradHPU, AutogradNestedTensor, AutogradPrivateUse1, AutogradPrivateUse2, AutogradPrivateUse3, Tracer, Autocast, Batched, VmapMode].

KojimaIsBad commented 2 years ago

Maybe you can create a new env and install the requirements. When I use --mixedprec, it seems all right.

Jungjee commented 2 years ago

Thank you @KojimaIsBad. We checked again that this works fine.