Skip to content

input of batchnorm1d can also be (N,C) #934

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

input of batchnorm1d can also be (N,C) #934

Show file tree
Hide file tree
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
5 changes: 5 additions & 0 deletions tests/converter_tests/test_converters.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -203,6 +203,11 @@ def test_batch_norm_nd(nd, with_conv):
inputs = [torch.randn(*input_size).cuda()]
cross_validate(module, inputs, fp16_mode=False, tol=1e-1)

if nd == 1:
input_size = [2, 3]
inputs = [torch.randn(*input_size).cuda()]
cross_validate(module, inputs, fp16_mode=False, tol=1e-1)


@pytest.mark.parametrize("dim", [1, -1])
def test_cat(dim):
Expand Down
4 changes: 2 additions & 2 deletions torch2trt/converters/native_converters.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -156,7 +156,7 @@ def convert_batch_norm(ctx):
bias = bias.detach().cpu().numpy() - running_mean.detach().cpu().numpy() * scale
power = np.ones_like(scale)

if ndim == 1:
if ndim == 1 or ndim == 0:
# reshape to 2D
layer = ctx.network.add_shuffle(input_trt)

Expand All @@ -171,7 +171,7 @@ def convert_batch_norm(ctx):

layer = ctx.network.add_scale_nd(scale_input, trt.ScaleMode.CHANNEL, bias, scale, power, 1)

if ndim == 1:
if ndim == 1 or ndim == 0:
# reshape back to 1D
layer = ctx.network.add_shuffle(layer.get_output(0))
if len(input.shape) == 2:
Expand Down