Skip to content

[ViTHybrid] fix last accelerate slow test #20705

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
younesbelkada merged 3 commits into from
Dec 9, 2022
Merged

[ViTHybrid] fix last accelerate slow test #20705

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
cf2ae93
fix last slow test
younesbelkada Dec 9, 2022
96f2657
revert deletion
younesbelkada Dec 9, 2022
d1ac53a
Update src/transformers/models/vit_hybrid/modeling_vit_hybrid.py
younesbelkada Dec 9, 2022
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
4 changes: 2 additions & 2 deletions src/transformers/models/vit_hybrid/modeling_vit_hybrid.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -358,7 +358,6 @@ def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> to
return hidden_states


# Copied from transformers.models.vit.modeling_vit.ViTLayer with ViT->ViTHybrid
class ViTHybridLayer(nn.Module):
"""This corresponds to the Block class in the timm implementation."""

Expand Down Expand Up @@ -387,7 +386,8 @@ def forward(
outputs = self_attention_outputs[1:] # add self attentions if we output attention weights

# first residual connection
hidden_states = attention_output + hidden_states
# We assign to correct device for `accelerate`, check: https://github.com/huggingface/transformers/pull/20705/
hidden_states = attention_output + hidden_states.to(attention_output.device)

# in ViTHybrid, layernorm is also applied after self-attention
layer_output = self.layernorm_after(hidden_states)
Expand Down