Skip to content

[Frontend][Pytorch] Add axis N when maxpool3d layout is (C,D,H,W) #12467

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
masahi merged 2 commits into from
Aug 18, 2022
Merged

[Frontend][Pytorch] Add axis N when maxpool3d layout is (C,D,H,W) #12467

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9266ba4
Add axis N if input is (C,D,H,W) layout.
blackkker Aug 17, 2022
63e0520
Add (C,D,H,W) test case.
blackkker Aug 17, 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
7 changes: 6 additions & 1 deletion python/tvm/relay/frontend/pytorch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1071,20 +1071,25 @@ def maxpool_1d(self, inputs, input_types):
def maxpool_3d(self, inputs, input_types):
data = inputs[0]

need_squeeze = False
if len(self.get_dims(data)) == 4:
need_squeeze = True
data = _op.expand_dims(data, 0)
pool_size = inputs[1]
strides = inputs[2] if inputs[2] else pool_size
padding = inputs[3]
dilation = inputs[4]
ceil_mode = int(inputs[5])

return _op.nn.max_pool3d(
res = _op.nn.max_pool3d(
data,
pool_size=pool_size,
strides=strides,
dilation=dilation,
padding=padding,
ceil_mode=ceil_mode,
)
return res if not need_squeeze else _op.squeeze(res, [0])

def hardtanh(self, inputs, input_types):
a = inputs[0]
Expand Down
16 changes: 10 additions & 6 deletions tests/python/frontend/pytorch/test_forward.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -906,13 +906,17 @@ def forward(self, *args):
def test_forward_maxpool3d():
"""test_forward_maxpool3d"""
torch.set_grad_enabled(False)
input_shape = [1, 3, 10, 10, 10]
input_data = torch.rand(input_shape).float()
for input_shape in [(1, 3, 10, 10, 10), (3, 10, 10, 10)]:
input_data = torch.rand(input_shape).float()

verify_model(torch.nn.MaxPool3d(kernel_size=[1, 1, 1]).eval(), input_data)
verify_model(torch.nn.MaxPool3d(kernel_size=[2, 2, 2], dilation=[1, 2, 3]).eval(), input_data)
verify_model(torch.nn.MaxPool3d(kernel_size=[10, 10, 10]).eval(), input_data)
verify_model(torch.nn.MaxPool3d(kernel_size=[4, 4, 4], padding=2, stride=2).eval(), input_data)
verify_model(torch.nn.MaxPool3d(kernel_size=[1, 1, 1]).eval(), input_data)
verify_model(
torch.nn.MaxPool3d(kernel_size=[2, 2, 2], dilation=[1, 2, 3]).eval(), input_data
)
verify_model(torch.nn.MaxPool3d(kernel_size=[10, 10, 10]).eval(), input_data)
verify_model(
torch.nn.MaxPool3d(kernel_size=[4, 4, 4], padding=2, stride=2).eval(), input_data
)

# A functional variant (default strides = None case)
class MaxPool3D(Module):
Expand Down