Skip to content

Fix reshape in ViTAutoEnc #3841

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
wyli merged 4 commits into from
Feb 22, 2022
Merged

Fix reshape in ViTAutoEnc #3841

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5d79751
Add failing test
razorx89 Feb 22, 2022
4d15214
Fix reshape after ViT
razorx89 Feb 22, 2022
2f4cf1e
Remove unused import
razorx89 Feb 22, 2022
80c5005
Merge branch 'dev' into fix-vitautoenc-shape
wyli Feb 22, 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
6 changes: 4 additions & 2 deletions monai/networks/nets/vitautoenc.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,6 @@
# limitations under the License.


import math
from typing import Sequence, Union

import torch
Expand All @@ -19,6 +18,7 @@
from monai.networks.blocks.patchembedding import PatchEmbeddingBlock
from monai.networks.blocks.transformerblock import TransformerBlock
from monai.networks.layers import Conv
from monai.utils import ensure_tuple_rep

__all__ = ["ViTAutoEnc"]

Expand Down Expand Up @@ -74,6 +74,7 @@ def __init__(

super().__init__()

self.patch_size = ensure_tuple_rep(patch_size, spatial_dims)
self.spatial_dims = spatial_dims

self.patch_embedding = PatchEmbeddingBlock(
Expand Down Expand Up @@ -105,14 +106,15 @@ def forward(self, x):
x: input tensor must have isotropic spatial dimensions,
such as ``[batch_size, channels, sp_size, sp_size[, sp_size]]``.
"""
spatial_size = x.shape[2:]
x = self.patch_embedding(x)
hidden_states_out = []
for blk in self.blocks:
x = blk(x)
hidden_states_out.append(x)
x = self.norm(x)
x = x.transpose(1, 2)
d = [round(math.pow(x.shape[2], 1 / self.spatial_dims))] * self.spatial_dims
d = [s // p for s, p in zip(spatial_size, self.patch_size)]
x = torch.reshape(x, [x.shape[0], x.shape[1], *d])
x = self.conv3d_transpose(x)
x = self.conv3d_transpose_1(x)
Expand Down
19 changes: 19 additions & 0 deletions tests/test_vitautoenc.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,6 +41,25 @@

TEST_CASE_Vitautoenc.append(test_case)

TEST_CASE_Vitautoenc.append(
[
{
"in_channels": 1,
"img_size": (512, 512, 32),
"patch_size": (16, 16, 16),
"hidden_size": 768,
"mlp_dim": 3072,
"num_layers": 4,
"num_heads": 12,
"pos_embed": "conv",
"dropout_rate": 0.6,
"spatial_dims": 3,
},
(2, 1, 512, 512, 32),
(2, 1, 512, 512, 32),
]
)


class TestPatchEmbeddingBlock(unittest.TestCase):
@parameterized.expand(TEST_CASE_Vitautoenc)
Expand Down