Skip to content
New issue

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

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

Already on GitHub? # to your account

fix: add runtime_shape argument in batchnorm-related classes to match… #804

Open
wants to merge 2 commits into
base: dev
Choose a base branch
from

Conversation

euphoria0-0
Copy link

@euphoria0-0 euphoria0-0 commented Jan 23, 2024

This PR is related to issue #450 .
BatchNorm1dToQuantScaleBias would provide the same shape: (B, D) -> (B, D).

dim = 100
batch_norm = BatchNorm1dToQuantScaleBias(num_features=dim, runtime_shape=(1, -1))
inputs = torch.rand(size=(5, dim))
outputs = batch_norm(inputs)
print('Outputs:', outputs.shape)

Outputs: torch.Size([5, 100])

batch_norm = BatchNorm1dToQuantScaleBias(num_features=dim, runtime_shape=(1, -1, 1))
inputs = torch.rand(size=(5, dim, 32))
outputs = batch_norm(inputs)
print('Outputs:', outputs.shape)

Outputs: torch.Size([5, 100, 32])

@euphoria0-0 euphoria0-0 changed the base branch from master to dev January 24, 2024 01:05
@Giuseppe5
Copy link
Collaborator

Thanks, I will take a look at this!

# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants