Skip to content

Revert from native gelu. Breaks ONNX export. #447

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.

Sign up for GitHub

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

Merged
szmigacz merged 1 commit into from
Apr 7, 2020
Merged

Revert from native gelu. Breaks ONNX export. #447

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions PyTorch/LanguageModeling/BERT/modeling.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -117,11 +117,11 @@ def load_tf_weights_in_bert(model, tf_checkpoint_path):
return model

def gelu(x):
return x * 0.5 * (1.0 + torch.erf(x / 1.41421))
return x * 0.5 * (1.0 + torch.erf(x / 1.41421))

def bias_gelu(bias, y):
x = bias + y
return torch.nn.functional.gelu(x)# x * 0.5 * (1.0 + torch.erf(x / 1.41421))
return x * 0.5 * (1.0 + torch.erf(x / 1.41421))

def bias_tanh(bias, y):
x = bias + y
Expand All @@ -130,7 +130,7 @@ def bias_tanh(bias, y):
def swish(x):
return x * torch.sigmoid(x)

ACT2FN = {"gelu": torch.nn.functional.gelu, "bias_gelu": bias_gelu, "bias_tanh": bias_tanh, "relu": torch.nn.functional.relu, "swish": swish}
ACT2FN = {"gelu": gelu, "bias_gelu": bias_gelu, "bias_tanh": bias_tanh, "relu": torch.nn.functional.relu, "swish": swish}

class LinearActivation(Module):
r"""Fused Linear and activation Module.
Expand Down