mirror of
https://github.com/open-mmlab/mmsegmentation.git
synced 2025-06-03 22:03:48 +08:00
c11da07e18
* delete convert function and add instruction to README.md * unified model convert and README * remove url * fix import error * fix unittest * rename pretrain * rename vit and deit pretrain * Update upernet_deit-b16_512x512_160k_ade20k.py * Update upernet_deit-b16_512x512_80k_ade20k.py * Update upernet_deit-b16_ln_mln_512x512_160k_ade20k.py * Update upernet_deit-b16_mln_512x512_160k_ade20k.py * Update upernet_deit-s16_512x512_160k_ade20k.py * Update upernet_deit-s16_512x512_80k_ade20k.py * Update upernet_deit-s16_ln_mln_512x512_160k_ade20k.py * Update upernet_deit-s16_mln_512x512_160k_ade20k.py Co-authored-by: Jiarui XU <xvjiarui0826@gmail.com> Co-authored-by: Junjun2016 <hejunjun@sjtu.edu.cn>
62 lines
1.9 KiB
Python
62 lines
1.9 KiB
Python
# Copyright (c) OpenMMLab. All rights reserved.
|
|
import pytest
|
|
import torch
|
|
|
|
from mmseg.models.backbones import SwinTransformer
|
|
|
|
|
|
def test_swin_transformer():
|
|
"""Test Swin Transformer backbone."""
|
|
|
|
with pytest.raises(TypeError):
|
|
# Pretrained arg must be str or None.
|
|
model = SwinTransformer(pretrained=123)
|
|
|
|
with pytest.raises(AssertionError):
|
|
# Because swin use non-overlapping patch embed, so the stride of patch
|
|
# embed must be equal to patch size.
|
|
model = SwinTransformer(strides=(2, 2, 2, 2), patch_size=4)
|
|
|
|
# Test absolute position embedding
|
|
temp = torch.randn((1, 3, 224, 224))
|
|
model = SwinTransformer(pretrain_img_size=224, use_abs_pos_embed=True)
|
|
model.init_weights()
|
|
model(temp)
|
|
|
|
# Test patch norm
|
|
model = SwinTransformer(patch_norm=False)
|
|
model(temp)
|
|
|
|
# Test pretrain img size
|
|
model = SwinTransformer(pretrain_img_size=(224, ))
|
|
|
|
with pytest.raises(AssertionError):
|
|
model = SwinTransformer(pretrain_img_size=(224, 224, 224))
|
|
|
|
# Test normal inference
|
|
temp = torch.randn((1, 3, 512, 512))
|
|
model = SwinTransformer()
|
|
outs = model(temp)
|
|
assert outs[0].shape == (1, 96, 128, 128)
|
|
assert outs[1].shape == (1, 192, 64, 64)
|
|
assert outs[2].shape == (1, 384, 32, 32)
|
|
assert outs[3].shape == (1, 768, 16, 16)
|
|
|
|
# Test abnormal inference
|
|
temp = torch.randn((1, 3, 511, 511))
|
|
model = SwinTransformer()
|
|
outs = model(temp)
|
|
assert outs[0].shape == (1, 96, 128, 128)
|
|
assert outs[1].shape == (1, 192, 64, 64)
|
|
assert outs[2].shape == (1, 384, 32, 32)
|
|
assert outs[3].shape == (1, 768, 16, 16)
|
|
|
|
# Test abnormal inference
|
|
temp = torch.randn((1, 3, 112, 137))
|
|
model = SwinTransformer()
|
|
outs = model(temp)
|
|
assert outs[0].shape == (1, 96, 28, 35)
|
|
assert outs[1].shape == (1, 192, 14, 18)
|
|
assert outs[2].shape == (1, 384, 7, 9)
|
|
assert outs[3].shape == (1, 768, 4, 5)
|