Embedding max_norm wrong results on cuda when repeating indices are present #44792
Labels
high priority
module: cuda
Related to torch.cuda, and CUDA support in general
triaged
This issue has been looked at a team member, and triaged and prioritized into an appropriate module
馃悰 Bug
I need an reproducible output from model, but Embedding layer reproduces different results in some cases.
To Reproduce
Steps to reproduce the behavior:
cuda
deviceEmbedding
layer with largeembedding_dim
andmax_norm
enabledEmbeddings are different for different application runs.
Expected behavior
I expect the same output for different application runs.
Environment
Additional context
cc @ezyang @gchanan @zou3519 @ngimel
The text was updated successfully, but these errors were encountered: