Skip to content
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

[Fix] Remove fp16 folder in configs. #1031

Merged
merged 6 commits into from
Nov 15, 2021
Merged

[Fix] Remove fp16 folder in configs. #1031

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
0e603d3
remove fp16 folder
MengzhangLI Nov 11, 2021
025433c
remove fp16 in docs
MengzhangLI Nov 11, 2021
caa8df0
fix some typos
MengzhangLI Nov 12, 2021
0e4fc19
fix some typos
MengzhangLI Nov 12, 2021
ba1ee1c
Merge branch 'master' of https://github.com/open-mmlab/mmsegmentation…
MengzhangLI Nov 15, 2021
81c3ea4
fix fp16 in yml
MengzhangLI Nov 15, 2021
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
4 changes: 2 additions & 2 deletions .dev/batch_test_list.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,8 +116,8 @@
]
fp16 = [
dict(
config='configs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes.py', # noqa
checkpoint='deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes-cc58bc8d.pth', # noqa
config='configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py', # noqa
checkpoint='deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes_20200717_230920-f1104f4b.pth', # noqa
eval='mIoU',
metric=dict(mIoU=80.46),
)
Expand Down
2 changes: 1 addition & 1 deletion .dev/batch_train_list.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,5 +15,5 @@ configs/deeplabv3plus/deeplabv3plus_r50-d8_512x1024_80k_cityscapes.py
configs/deeplabv3plus/deeplabv3plus_r50-d8_769x769_80k_cityscapes.py
configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py
configs/vit/upernet_deit-s16_ln_mln_512x512_160k_ade20k.py
configs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes.py
configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py
configs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K.py
4 changes: 2 additions & 2 deletions .dev/benchmark_evaluation.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,7 @@ echo 'configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 tools/slurm_test.sh $PARTITION upernet_vit-b16_ln_mln_512x512_160k_ade20k configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py $CHECKPOINT_DIR/upernet_vit-b16_ln_mln_512x512_160k_ade20k-f444c077.pth --eval mIoU --work-dir work_dirs/benchmark_evaluation/upernet_vit-b16_ln_mln_512x512_160k_ade20k --options dist_params.port=28186 &
echo 'configs/vit/upernet_deit-s16_ln_mln_512x512_160k_ade20k.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 tools/slurm_test.sh $PARTITION upernet_deit-s16_ln_mln_512x512_160k_ade20k configs/vit/upernet_deit-s16_ln_mln_512x512_160k_ade20k.py $CHECKPOINT_DIR/upernet_deit-s16_ln_mln_512x512_160k_ade20k-c0cd652f.pth --eval mIoU --work-dir work_dirs/benchmark_evaluation/upernet_deit-s16_ln_mln_512x512_160k_ade20k --options dist_params.port=28187 &
echo 'configs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 tools/slurm_test.sh $PARTITION deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes configs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes.py $CHECKPOINT_DIR/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes-cc58bc8d.pth --eval mIoU --work-dir work_dirs/benchmark_evaluation/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes --options dist_params.port=28188 &
echo 'configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 tools/slurm_test.sh $PARTITION deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py $CHECKPOINT_DIR/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes-cc58bc8d.pth --eval mIoU --work-dir work_dirs/benchmark_evaluation/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes --options dist_params.port=28188 &
echo 'configs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 tools/slurm_test.sh $PARTITION upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K configs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K.py $CHECKPOINT_DIR/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K_20210531_112542-e380ad3e.pth --eval mIoU --work-dir work_dirs/benchmark_evaluation/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K --options dist_params.port=28189 &
4 changes: 2 additions & 2 deletions .dev/benchmark_train.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ echo 'configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py' &
GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=2 ./tools/slurm_train.sh $PARTITION upernet_vit-b16_ln_mln_512x512_160k_ade20k configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py --options checkpoint_config.max_keep_ckpts=1 dist_params.port=24742 --work-dir work_dirs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k >/dev/null &
echo 'configs/vit/upernet_deit-s16_ln_mln_512x512_160k_ade20k.py' &
GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=2 ./tools/slurm_train.sh $PARTITION upernet_deit-s16_ln_mln_512x512_160k_ade20k configs/vit/upernet_deit-s16_ln_mln_512x512_160k_ade20k.py --options checkpoint_config.max_keep_ckpts=1 dist_params.port=24743 --work-dir work_dirs/vit/upernet_deit-s16_ln_mln_512x512_160k_ade20k >/dev/null &
echo 'configs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 ./tools/slurm_train.sh $PARTITION deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes configs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes.py --options checkpoint_config.max_keep_ckpts=1 dist_params.port=24744 --work-dir work_dirs/fp16/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes >/dev/null &
echo 'configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py' &
GPUS=4 GPUS_PER_NODE=4 CPUS_PER_TASK=2 ./tools/slurm_train.sh $PARTITION deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py --options checkpoint_config.max_keep_ckpts=1 dist_params.port=24744 --work-dir work_dirs/deeplabv3plus/deeplabv3plus_r101-d8_512x1024_80k_fp16_cityscapes >/dev/null &
echo 'configs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K.py' &
GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=2 ./tools/slurm_train.sh $PARTITION upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K configs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K.py --options checkpoint_config.max_keep_ckpts=1 dist_params.port=24745 --work-dir work_dirs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K >/dev/null &
2 changes: 1 addition & 1 deletion .dev/md2yml.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -201,7 +201,7 @@ def parse_md(md_file):
'batch size':
1,
'mode':
'FP32',
'FP32' if 'fp16' not in config else 'FP16',
'resolution':
f'({crop_size[0]},{crop_size[1]})'
}]
Expand Down
1 change: 0 additions & 1 deletion README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,7 +73,6 @@ Supported methods:
- [x] [UNet (MICCAI'2016/Nat. Methods'2019)](configs/unet)
- [x] [PSPNet (CVPR'2017)](configs/pspnet)
- [x] [DeepLabV3 (ArXiv'2017)](configs/deeplabv3)
- [x] [Mixed Precision (FP16) Training (ArXiv'2017)](configs/fp16)
- [x] [BiSeNetV1 (ECCV'2018)](configs/bisenetv1)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we add one item named supported training tricks in the future?

- [x] [PSANet (ECCV'2018)](configs/psanet)
- [x] [DeepLabV3+ (CVPR'2018)](configs/deeplabv3plus)
Expand Down
1 change: 0 additions & 1 deletion README_zh-CN.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,7 +72,6 @@ MMSegmentation 是一个基于 PyTorch 的语义分割开源工具箱。它是 O
- [x] [UNet (MICCAI'2016/Nat. Methods'2019)](configs/unet)
- [x] [PSPNet (CVPR'2017)](configs/pspnet)
- [x] [DeepLabV3 (ArXiv'2017)](configs/deeplabv3)
- [x] [Mixed Precision (FP16) Training (ArXiv'2017)](configs/fp16)
- [x] [BiSeNetV1 (ECCV'2018)](configs/bisenetv1)
- [x] [PSANet (ECCV'2018)](configs/psanet)
- [x] [DeepLabV3+ (CVPR'2018)](configs/deeplabv3plus)
Expand Down
2 changes: 1 addition & 1 deletion configs/bisenetv2/bisenetv2.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,7 +75,7 @@ Models:
hardware: V100
backend: PyTorch
batch size: 1
mode: FP32
mode: FP16
resolution: (1024,1024)
memory (GB): 5.77
Results:
Expand Down
5 changes: 5 additions & 0 deletions configs/deeplabv3/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,7 @@
| DeepLabV3 | R-18-D8 | 512x1024 | 80000 | 1.7 | 13.78 | 76.70 | 78.27 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r18-d8_512x1024_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r18-d8_512x1024_80k_cityscapes/deeplabv3_r18-d8_512x1024_80k_cityscapes_20201225_021506-23dffbe2.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r18-d8_512x1024_80k_cityscapes/deeplabv3_r18-d8_512x1024_80k_cityscapes-20201225_021506.log.json) |
| DeepLabV3 | R-50-D8 | 512x1024 | 80000 | - | - | 79.32 | 80.57 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r50-d8_512x1024_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_512x1024_80k_cityscapes/deeplabv3_r50-d8_512x1024_80k_cityscapes_20200606_113404-b92cfdd4.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_512x1024_80k_cityscapes/deeplabv3_r50-d8_512x1024_80k_cityscapes_20200606_113404.log.json) |
| DeepLabV3 | R-101-D8 | 512x1024 | 80000 | - | - | 80.20 | 81.21 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes/deeplabv3_r101-d8_512x1024_80k_cityscapes_20200606_113503-9e428899.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes/deeplabv3_r101-d8_512x1024_80k_cityscapes_20200606_113503.log.json) |
| DeepLabV3 (FP16) | R-101-D8 | 512x1024 | 80000 | 5.75 | 3.86 | 80.48 | - | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes_20200717_230920-774d9cec.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes_20200717_230920.log.json) |
MengzhangLI marked this conversation as resolved.
Show resolved Hide resolved
| DeepLabV3 | R-18-D8 | 769x769 | 80000 | 1.9 | 5.55 | 76.60 | 78.26 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r18-d8_769x769_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r18-d8_769x769_80k_cityscapes/deeplabv3_r18-d8_769x769_80k_cityscapes_20201225_021506-6452126a.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r18-d8_769x769_80k_cityscapes/deeplabv3_r18-d8_769x769_80k_cityscapes-20201225_021506.log.json) |
| DeepLabV3 | R-50-D8 | 769x769 | 80000 | - | - | 79.89 | 81.06 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r50-d8_769x769_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_769x769_80k_cityscapes/deeplabv3_r50-d8_769x769_80k_cityscapes_20200606_221338-788d6228.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_769x769_80k_cityscapes/deeplabv3_r50-d8_769x769_80k_cityscapes_20200606_221338.log.json) |
| DeepLabV3 | R-101-D8 | 769x769 | 80000 | - | - | 79.67 | 80.81 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3/deeplabv3_r101-d8_769x769_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_769x769_80k_cityscapes/deeplabv3_r101-d8_769x769_80k_cityscapes_20200607_013353-60e95418.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_769x769_80k_cityscapes/deeplabv3_r101-d8_769x769_80k_cityscapes_20200607_013353.log.json) |
Expand Down Expand Up @@ -102,3 +103,7 @@
| DeepLabV3 | R-101-D8 | 512x512 | 160000 | - | - | 41.82 | 42.49 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3_r101-d8_512x512_4x4_160k_coco-stuff164k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x512_4x4_160k_coco-stuff164k/deeplabv3_r101-d8_512x512_4x4_160k_coco-stuff164k_20210709_155402-f035acfd.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x512_4x4_160k_coco-stuff164k/deeplabv3_r101-d8_512x512_4x4_160k_coco-stuff164k_20210709_155402.log.json) |
| DeepLabV3 | R-50-D8 | 512x512 | 320000 | - | - | 41.37 | 42.22 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3_r50-d8_512x512_4x4_320k_coco-stuff164k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_512x512_4x4_320k_coco-stuff164k/deeplabv3_r50-d8_512x512_4x4_320k_coco-stuff164k_20210709_155403-51b21115.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_512x512_4x4_320k_coco-stuff164k/deeplabv3_r50-d8_512x512_4x4_320k_coco-stuff164k_20210709_155403.log.json) |
| DeepLabV3 | R-101-D8 | 512x512 | 320000 | - | - | 42.61 | 43.42 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/deeplabv3_r101-d8_512x512_4x4_320k_coco-stuff164k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x512_4x4_320k_coco-stuff164k/deeplabv3_r101-d8_512x512_4x4_320k_coco-stuff164k_20210709_155402-3cbca14d.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x512_4x4_320k_coco-stuff164k/deeplabv3_r101-d8_512x512_4x4_320k_coco-stuff164k_20210709_155402.log.json) |

Note:

- `FP16` means Mixed Precision (FP16) is adopted in training.
21 changes: 21 additions & 0 deletions configs/deeplabv3/deeplabv3.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -157,6 +157,27 @@ Models:
mIoU(ms+flip): 81.21
Config: configs/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes.py
Weights: https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes/deeplabv3_r101-d8_512x1024_80k_cityscapes_20200606_113503-9e428899.pth
- Name: deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes
In Collection: deeplabv3
Metadata:
backbone: R-101-D8
crop size: (512,1024)
lr schd: 80000
inference time (ms/im):
- value: 259.07
hardware: V100
backend: PyTorch
batch size: 1
mode: FP16
resolution: (512,1024)
memory (GB): 5.75
Results:
- Task: Semantic Segmentation
Dataset: Cityscapes
Metrics:
mIoU: 80.48
Config: configs/deeplabv3/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes.py
Weights: https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes_20200717_230920-774d9cec.pth
- Name: deeplabv3_r18-d8_769x769_80k_cityscapes
In Collection: deeplabv3
Metadata:
Expand Down
2 changes: 1 addition & 1 deletion ...t_r101-d8_512x1024_80k_fp16_cityscapes.py → ...3_r101-d8_fp16_512x1024_80k_cityscapes.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
_base_ = '../pspnet/pspnet_r101-d8_512x1024_80k_cityscapes.py'
_base_ = './deeplabv3_r101-d8_512x1024_80k_cityscapes.py'
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale=512.)
# fp16 placeholder
Expand Down
Loading