Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Bias Gelu Fusion + GEGLU errors #237

Closed
sdtblck opened this issue Apr 16, 2021 · 0 comments · Fixed by #304
Closed

Bias Gelu Fusion + GEGLU errors #237

sdtblck opened this issue Apr 16, 2021 · 0 comments · Fixed by #304
Labels
bug Something isn't working

Comments

@sdtblck
Copy link
Contributor

sdtblck commented Apr 16, 2021

Describe the bug
the bias gelu fusion is expecting a tensor, but we pass it an int if bias is None. Should just be able to pass it a zeros tensor to fix.

 File "/nobackup/users/andonian/gpt-neox/megatron/model/transformer.py", line 117, in forward
    bias_gelu_impl(gate, bias_2)
  File "/nobackup/users/andonian/gpt-neox/megatron/model/fused_bias_gelu.py", line 52, in forward
    return bias_gelu(bias, input)
RuntimeError: bias_gelu() Expected a value of type 'Tensor (inferred)' for argument 'bias' but instead found type 'int'.
Inferred 'bias' to be of type 'Tensor' because it was not annotated with an explicit type.
Position: 0
Value: 0
Declaration: bias_gelu(Tensor bias, Tensor y) -> (Tensor)
Cast error details: Unable to cast Python instance to C++ type (compile in debug mode for details)
@sdtblck sdtblck added the bug Something isn't working label Apr 16, 2021
@sdtblck sdtblck linked a pull request May 7, 2021 that will close this issue
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

Successfully merging a pull request may close this issue.

1 participant