38040d5e05
[Fix] Fix performance issue in convnext DDP train. ( #1098 )
...
to fix performance issue in convnext DDP train
2022-10-17 10:10:19 +08:00
91b85bb4a5
Merge remote-tracking branch 'origin/dev'
2022-09-30 18:06:17 +08:00
8c7b7b15a3
[Enhance] RepVGG for YOLOX-PAI. ( #1025 )
...
* repvgg add ppf for yoloxpai
* fix by review
* update stem_channels
* fix doc
Co-authored-by: Ezra-Yu <18586273+Ezra-Yu@users.noreply.github.com>
2022-09-30 14:20:53 +08:00
1047daa28e
[Feature] Support HorNet Backbone. ( #1013 )
...
* add hornet
* add hornet
* add hornet
* add hornet
* add hornet
* add hornet
* add hornet
* fix test for torch before 1.7.0
* del timm
* fix readme
* fix readme
* Update mmcls/models/backbones/hornet.py
Co-authored-by: Ezra-Yu <18586273+Ezra-Yu@users.noreply.github.com>
* fix docs
* fix docs
* s -> scale
* fix dims and dpr impl
* fix layer scale
* refactor gnconv
* add dw_cfg
* add convert tools
* update code
* update docs
* update readme
* update URLs
Co-authored-by: Ezra-Yu <18586273+Ezra-Yu@users.noreply.github.com>
2022-09-27 10:37:49 +08:00
56589ee280
[Enhancement] Update VAN. ( #1017 )
...
* update van
* fix init
* b4 result
* update van
* keep old config
* keep old config
* fix metafile
* update VAN configs
* update example
Co-authored-by: Ezra-Yu <18586273+Ezra-Yu@users.noreply.github.com>
2022-09-27 09:44:40 +08:00
6ebb3f77ad
[Fix] Fix attenstion clamp max params ( #1034 )
2022-09-26 14:12:51 +08:00
75ae8453ac
[Docs] Fix a typo in ImageClassifier ( #1050 )
2022-09-22 09:24:23 +08:00
0b4a67dd31
[Refactor] Re-write get_sinusoid_encoding from third-party implementation. ( #965 )
2022-09-13 15:24:29 +08:00
517bd3d34b
[Fix] Fix device mismatch in Swin-v2. ( #976 )
2022-09-01 18:03:49 +08:00
ec71d071d8
[Improve] Fixed typo in `RepVGG`. ( #985 )
...
* [Improve] Use `forward_dummy` to calculate FLOPS. (#953 )
* fixed
Co-authored-by: Ming-Hsuan-Tu <alec.tu@acer.com>
2022-08-22 10:28:33 +08:00
5ad3bed2cd
Merge remote-tracking branch 'origin/master' into dev
2022-08-22 10:12:24 +08:00
6474ea2fc0
[Feature] Support EfficientFormer. ( #954 )
...
* add efficient backbone
* Update Readme and metafile
* Add unit tests
* fix confict
* fix lint
* update efficientformer head unit tests
* update README
* fix unit test
* fix Readme
* fix example
* fix typo
* recover api modification
* Update EfficiemtFormer Backbone
* fix unit tests
* add efficientformer to readme and model zoo
2022-08-16 23:38:08 +08:00
b366897889
[Docs] Refine the docstring of RegNet ( #935 )
...
* Update regnet.py
In the example comment to print out the different layers of outputs, we need to indicate the `out_indices` to (0,1,2,3) to see all backbone layers output as the default argument is (3,)
* Update regnet.py
following changes proposal of maintainer
* fix linting
* fix blank space for docs
* fix blank space for docs
* fix blank space for docs
2022-08-10 18:17:36 +08:00
90254a8455
[Improve] Use `forward_dummy` to calculate FLOPS. ( #953 )
2022-08-08 18:34:09 +08:00
1a3d51acc2
[Feature] Support CSRA head. ( #881 )
...
* Support CSRA head.
* Add CSRA config.
* Improve training scheduler and Update cfg, ckpt, log
* Update metafile
* Rename config files and checkpoints
Co-authored-by: Ezra-Yu <1105212286@qq.com>
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-08-04 18:15:51 +08:00
b5bb86a357
[Fix] Fix the output position of Swin-Transformer. ( #947 )
...
* [Fix] Fix the output position of Swin-Transformer.
* Rename `downsample` argument to `do_downsample`.
2022-08-03 19:32:29 +08:00
6ec38fe742
[Feature] Support Swin Transform V2. ( #799 )
...
* init rough try for modify
* Init swin transform v2
* lint
* reformat
* init config
* refactor
* update config
* fix test
* add doc
* refact
* add model meta
* rename config
* add doc
* fix meta model name
* restruct
* rename embed_bims to out_channels
* fix ut and update model
2022-08-03 17:33:08 +08:00
556fa567a8
[Feature] Support MViT and add checkpoints. ( #924 )
...
* [Feature] Support MViT.
* Add MViT configs and docs
* Add unit test
* Fix unit tests.
2022-08-02 15:20:16 +08:00
3e198e63e8
[Docs] Update install tutorials. ( #854 )
...
* [Docs] Update install tutorials.
* [Docs] Improve dataset docs
* Add option to show the results in demo.
* fix typo
2022-06-01 18:31:57 +08:00
7c5ddb1e5b
Bump version to v0.23.0 ( #809 )
2022-05-01 21:58:33 +08:00
b4eefe4794
[Enhance] Support training on IPU and add fine-tuning configs of ViT. ( #723 )
...
* implement training and evaluation on IPU
* fp16 SOTA
* Tput reaches 5600
* 123
* add poptorch dataloder
* change ipu_replicas to ipu-replicas
* add noqa to config long line(website)
* remove ipu dataloder test code
* del one blank line in test_builder
* refine the dataloder initialization
* fix a typo
* refine args for dataloder
* remove an annoted line
* process one more conflict
* adjust code structure in mmcv.ipu
* adjust ipu code structure in mmcv
* IPUDataloader to IPUDataLoader
* align with mmcv
* adjust according to mmcv
* mmcv code structre fixed
Co-authored-by: hudi <dihu@graphcore.ai>
2022-04-29 22:22:19 +08:00
59292b315a
[Feature] Support DenseNet. ( #750 )
...
* init add densenet implementation
* Add config and converted models
* update meta
* add test for memory efficient
* Add docs
* add doc for jit
* Update checkpoint path
* Update readthedocs
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-04-28 13:06:03 +08:00
df6edd7f5a
[Feature] Support VAN. ( #739 )
...
* add van
* fix config
* add metafile
* add test
* model convert script
* fix review
* fix lint
* fix the configs and improve docs
* rm debug lines
* add VAN into api
Co-authored-by: Yu Zhaohui <1105212286@qq.com>
2022-04-28 09:35:17 +08:00
702c196514
[Docs] Refine API reference. ( #774 )
...
* [Docs] Refine API reference
* Add PoolFormer
2022-04-15 20:19:20 +08:00
29b882d4eb
Merge branch 'dev'
2022-04-15 20:10:45 +08:00
2eb27d0d4e
[Feature] Add PoolFormer backbone and checkpoints. ( #746 )
...
* add PoolFormer
* fix some typos in PoolFormer
* fix lint error
* modify out_indices and gap
* fix typo
* fix lint
* fix typo
* fix typo in poolforemr README
* fix lint
* Update some paths
* Refactor freeze_stages method
* Add unit tests
* Fix lint
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-04-15 11:35:15 +08:00
fbcf907dc6
[Feature] Support resize relative position embedding in `SwinTransformer`. ( #749 )
...
* [Feature]: Add resize rel pos embed
* [Refactor]: Create a separated resize_rel_pos_bias_table func
* [Refactor]: Refactor rel pos embed bias
* [Refactor]: Move interpolate into func
* Remove index buffer only when window_size changes
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-04-13 23:06:56 +08:00
875195efe9
[Enhance] Improve CPE performance by reduce memory copy. ( #762 )
2022-04-02 11:12:20 +08:00
922a31a68e
[Feature] ADD RepMLP Backbone and checkpoints. ( #709 )
...
* add repmlp
* refactor backbone code
* add checkpoint, readme, unit tests
* unchange demo
* improve docstring.
* update pic
* refactor patch embed
* refactor unit tests
* fix lint
* update tools
* update tools
* Update checkpoint path
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-03-30 20:40:25 +08:00
292127098f
[Feature] Add CSPNet and backbone and checkpoints ( #735 )
...
* add cspnet backbone
* try
* add backbones
* add cspnet
* add acc
* add unit tests
* chnage mode
* Update checkpoint path
* Fix typo
* Imporve CSPNet
* Update checkpoints
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-03-30 15:29:35 +08:00
1717efadbe
[Fix] Fix the discontiguous output feature map of ConvNeXt. ( #743 )
...
* [Fix] Fix the discontiguous output feature map of ConvNeXt.
* Add comment
2022-03-25 12:34:44 +08:00
04cb42a768
[Feature] Support ConvMixer. ( #716 )
...
* basic support for ConvMixer
* simplify
* add data pipeine config for timm
* Add model readme and metafile
* add unittest for convmixer
* add copyright
* modify
* add tests
* update model
* add conv2dAdaptivePadding replacement
* update model index
* fix comments
* Update checkpoint path
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-03-23 15:23:57 +08:00
c1534f9126
[Feature] Support resizemix. ( #676 )
...
* add resizemix
* skip torch.__version__ < 1.7.0
* Update mmcls/models/utils/augment/resizemix.py
Co-authored-by: Ma Zerun <mzr1996@163.com>
* Update mmcls/models/utils/augment/resizemix.py
Co-authored-by: Ma Zerun <mzr1996@163.com>
* resize -> F.interpolate
* fix docs
* fix test
* add Copyright
* add argument interpolation
Co-authored-by: Ma Zerun <mzr1996@163.com>
2022-03-07 12:11:20 +08:00
c708770b42
[Enhance] Support dynamic input shape for ViT-based algorithms. ( #706 )
...
* Move `resize_pos_embed` to `mmcls.models.utils`
* Refactor Vision Transformer
* Refactor DeiT
* Refactor MLP-Mixer
* Refactor Swin-Transformer
* Remove `indexing` arg
* Support dynamic inputs for t2t_vit
* Add copyright
* Fix bugs in swin transformer
* Add `pad_small_maps` option
* Update swin transformer
* Handle `attn_mask` in checkpoints of swin
* Imporve by comments
2022-03-03 13:10:12 +08:00
1214df083d
[Enhance] Use PyTorch official `one_hot` to implement `convert_to_one_hot`. ( #696 )
...
* some change to mmcls/models/losses/utils.py:convert_to_one_hot()
* fixed problem: line too long
* fixed wrong output shape
* fixed lint PEP8 E128
* fix lint
* fix lint
* add unit tests
Co-authored-by: Ezra-Yu <1105212286@qq.com>
2022-02-28 10:20:31 +08:00
5f7322c211
[Fix] Fix Conformer forward with irregular input size. ( #686 )
...
* Auto pad the feature map to be divisible by 4.
* Add unit tests
* Use `trans_down_stride` instead of 4.
2022-02-28 10:17:52 +08:00
7fcaedcbfb
[Feature] Add ResNetV1c. ( #692 )
...
* add ResNetV1c
* add unit tests
* fix lint
* update docstring
* fix lint
2022-02-23 11:36:33 +08:00
43024cda73
[Feature] Support gem pooling ( #677 )
...
* add gem pooling
* add example config
* fix params
* add assert
* add param clamp
* add test assert
* add clamp
* fix conflict
2022-02-17 02:17:36 +08:00
fcd57913ae
[Enhance] Upgrade isort pre-commit hooks. ( #687 )
...
* Use new version flake8 and isort hooks
* Fix missing copyright
2022-02-17 02:17:20 +08:00
bca695b684
[Refactor] Remove deprecation. ( #633 )
...
* Remove deprecated `--options` in some tools
* Remove deprecated eval hooks and fp16 hooks
* Remove deprecated mixup&cutmix, pretrained and return_tuple arguments.
* Remove deprecated `deprecated_options` in multi label dataset
* Remove deprecated `formating.py`
* Remove deprecated default mode of `LabelSmoothLoss`
* Revert pretrained argument, and remove it's deprecation sign.
* Update unit test
2022-02-10 09:30:18 +08:00
5de480ea9e
[Feature] Support HRNet and add pre-trained models ( #660 )
...
* Support HRNet
* Add HRNet configs
* Fix a bug in backward
* Add configs and update docs.
* Not use bias in conv before batch norm
* Defaults to use `norm_eval=False`
* Add unit tests and support out_channels in HRFuseScales
* Update checkpoint path
* Update docstring.
* Remove incorrect files
* Improve according to comments
2022-01-28 10:54:14 +08:00
dc456a0c2c
[Feature] Support ConvNeXt ( #670 )
...
* Support ConvNeXt
* Add configs of ConvNeXt
* Update dev scripts
* Update docs.
* Use new style README
* Add unit tests.
* Update README
* Imporve according to comments
* Modify refers to timm.
* Imporve according to comments
2022-01-28 10:36:45 +08:00
94eb3ff83e
[Feature] Add Twins backbone and convert checkpoints. ( #642 )
...
* add twins backbone
* add position_encoding
* refactor twins
* Supplemental unit tests
* update docstring and readme
* update docstring and readme
* update docstring and readme
* update docstring
* update docstring
* update docstring
* update docstring
* remove note
* update doc and docstring
* update docstring
* update docstring
* use abstract pdf link and rename yamlfile
* Update model link
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-01-27 10:25:05 +08:00
bd397f790f
[Enhance] Rewrite channel split operation in ShufflenetV2 ( #632 )
...
* replace chunk op
* shufflenetv2 config
2022-01-25 14:45:28 +08:00
d56170a734
[Feature] Support EfficientNet ( #649 )
...
* add config for resnest test
* fix config
* add label smoothing
* add memcached
* minor fix
* fix bug
* fix config
* add config
* minor fix
* fix configs
* use EResize
* change interpolation
* add more configs
* add docsting
* add unittest
* remove unnecessary changes
* minor fix
* add more docstring
* fix linting
* add efficient backbone
* add config
* add Edge Residual
* fix bug
* remove unnecessary files
* refactor
* add resize in crop to ensure crop size is output size
* fix bug and add comments
* test
* fix
* add more configs
* add more configs
* add more configs
* fix bug
* add model zoo
* fix
* reorganize code
* add edge tpu
* add edge tpu converter
* rename
* update readme
* reorganize code and config
* Rename configs of EfficientNet, and add metafile & model_zoo
* Remove `backend='pillow'`
* Add comments about EfficientNet-EdgeTPU
* Rename the convert tool of EfficientNet.
* Refactor EfficientNet and update docstring.
* Update EfficientNet-EdgeTPU config
* Fix unit tests
Co-authored-by: lixinran <lixr423@outlook.com>
Co-authored-by: lixinran <lixinran@sensetime.com>
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-01-25 12:14:17 +08:00
16864c7495
[Feature] Support features_only in TIMMBackbone ( #668 )
...
* Support features_only in TIMMBackbone
based on https://github.com/open-mmlab/mmsegmentation/pull/998
* update test for mmdet
* fix unit test for build_without_timm
* Update docstring
Co-authored-by: mzr1996 <mzr1996@163.com>
2022-01-25 11:22:23 +08:00
e694269c59
[Enhance] Support single-label, softmax, custom eps by asymmetric loss ( #609 )
...
Co-authored-by: Minyus <Minyus@users.noreply.github.com>
2022-01-21 11:35:14 +08:00
d29037e8d1
[Enhance] Suport Mixup&Cutmix for multi-label task.
2022-01-21 11:30:58 +08:00
7dfc9e4a85
Bump version to v0.19.0 ( #623 )
2021-12-31 12:55:47 +08:00
0bbbb04429
[Enhancement] Improve the accuracy of ResNet ( #572 )
...
* add itp timm
* minor update
* minor update
* minor update
* add rep aug, minor update on configs
* minor update
* add target threshold
* add decaymulti
* minor update
* minor update
* add lbl smooth
* update lr
* reorganize config files and code
* minor bugfixes
* remove unused parts and minor fixes on cfg
* critical bugfix, add test and cfg update
* refactor code
* update doc string
* remove duplicate code
* refactor drop path in resnet
* rename
* Modify configs and add README&metafile
* Update metafile
Co-authored-by: mzr1996 <mzr1996@163.com>
2021-12-29 10:58:42 +08:00
790475019
mzr1996
takuoko
Hubert
WRH
Kai Hu
Andrey Moskalenko
Jiahao Wang
Ezra-Yu
Timothy Lim
Ming-Hsuan-Tu
JiayuXu
Hu Di
Weihao Yu
Yuan Liu
XiaobingZhang
HumberMe
Zhicheng Chen
Yosuke Shinya
Yusuke Minami