[Reproduction] Reproduce RegNetX training accuracy. (#587)

* add preciseBN hook * add precise-bn config * add proority of preciseBN * Add precise BN hook * Update PreciseBN hook * add config example & pre-commit * Update PBN * Add unit test * Update unit tests * mv preciseBN to mmcls.runner * rm extra sleep() * merge latest * update code * fix lint * update unit tests * add regnetx configs * update prcise BN in configs * fix lint * add checkpoints and logs * fix conflict * fix lint * update model_zoo and model-index * remove some changes * update readme.md * remove precise_bn.py
2021-12-28 17:34:27 +08:00 · 2021-12-28 17:34:27 +08:00 · e7c06b8541
parent 159b38d276
commit e7c06b8541
28 changed files with 293 additions and 465 deletions
--- a/configs/regnet/README.md
+++ b/configs/regnet/README.md
@ -12,6 +12,7 @@ In this work, we present a new network design paradigm. Our goal is to help adva
 </div>

 ## Citation
+
 ```latex
@article{radosavovic2020designing,
    title={Designing Network Design Spaces},
@ -29,13 +30,21 @@ In this work, we present a new network design paradigm. Our goal is to help adva

 |         Model         | Params(M) | Flops(G) | Top-1 (%) | Top-5 (%) | Config | Download |
 |:---------------------:|:---------:|:--------:|:---------:|:---------:|:------:|:--------:|
-| RegNetX-400MF\*       |   5.16    |   0.41   |   72.55   |   90.91   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-400mf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-400MF-0db9f35c.pth) |
-| RegNetX-800MF\*       |   7.26    |   0.81   |   75.21   |   92.37   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-800mf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-800MF-4f9d1e8a.pth) |
-| RegNetX-1.6GF\*       |   9.19    |   1.63   |   77.04   |   93.51   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-1.6gf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-1.6GF-cfb32375.pth) |
-| RegNetX-3.2GF\*       |   15.3    |   3.21   |   78.26   |   94.20   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-3.2gf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-3.2GF-82c43fd5.pth) |
-| RegNetX-4.0GF\*       |   22.12   |   4.0    |   78.72   |   94.22   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-4.0gf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-4.0GF-ef8bb32c.pth) |
-| RegNetX-6.4GF\*       |   26.21   |   6.51   |   79.22   |   94.61   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-6.4gf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-6.4GF-6888c0ea.pth) |
-| RegNetX-8.0GF\*       |   39.57   |   8.03   |   79.31   |   94.57   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-8.0gf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-8.0GF-cb4c77ec.pth) |
-| RegNetX-12GF\*        |   46.11   |   12.15  |   79.91   |   94.78   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-12gf_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-12GF-0574538f.pth) |
+| RegNetX-400MF         |   5.16    |   0.41   |   72.56   |   90.78   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-400mf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-400mf_8xb128_in1k_20211213-89bfc226.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-400mf_8xb128_in1k_20211208_143316.log.json) |
+| RegNetX-800MF         |   7.26    |   0.81   |   74.76   |   92.32   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-800mf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-800mf_8xb128_in1k_20211213-222b0f11.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-800mf_8xb128_in1k_20211207_143037.log.json) |
+| RegNetX-1.6GF         |   9.19    |   1.63   |   76.84   |   93.31   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-1.6gf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-1.6gf_8xb128_in1k_20211213-d1b89758.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-1.6gf_8xb128_in1k_20211208_143018.log.json) |
+| RegNetX-3.2GF         |   15.3    |   3.21   |   78.09   |   94.08   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-3.2gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-3.2gf_8xb64_in1k_20211213-1fdd82ae.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-3.2gf_8xb64_in1k_20211208_142720.log.json) |
+| RegNetX-4.0GF         |   22.12   |   4.0    |   78.60   |   94.17   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-4.0gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-4.0gf_8xb64_in1k_20211213-efed675c.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-4.0gf_8xb64_in1k_20211207_150431.log.json) |
+| RegNetX-6.4GF         |   26.21   |   6.51   |   79.38   |   94.65   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-6.4gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-6.4gf_8xb64_in1k_20211215-5c6089da.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-6.4gf_8xb64_in1k_20211213_172748.log.json) |
+| RegNetX-8.0GF         |   39.57   |   8.03   |   79.12   |   94.51   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-8.0gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-8.0gf_8xb64_in1k_20211213-9a9fcc76.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-8.0gf_8xb64_in1k_20211208_103250.log.json) |
+| RegNetX-12GF          |   46.11   |   12.15  |   79.67   |   95.03   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-12gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-12gf_8xb64_in1k_20211213-5df8c2f8.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-12gf_8xb64_in1k_20211208_143713.log.json) |
+| RegNetX-400MF\*       |   5.16    |   0.41   |   72.55   |   90.91   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-400mf_8xb128_in1k) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-400MF-0db9f35c.pth) |
+| RegNetX-800MF\*       |   7.26    |   0.81   |   75.21   |   92.37   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-800mf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-800MF-4f9d1e8a.pth) |
+| RegNetX-1.6GF\*       |   9.19    |   1.63   |   77.04   |   93.51   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-1.6gf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-1.6GF-cfb32375.pth) |
+| RegNetX-3.2GF\*       |   15.3    |   3.21   |   78.26   |   94.20   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-3.2gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-3.2GF-82c43fd5.pth) |
+| RegNetX-4.0GF\*       |   22.12   |   4.0    |   78.72   |   94.22   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-4.0gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-4.0GF-ef8bb32c.pth) |
+| RegNetX-6.4GF\*       |   26.21   |   6.51   |   79.22   |   94.61   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-6.4gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-6.4GF-6888c0ea.pth) |
+| RegNetX-8.0GF\*       |   39.57   |   8.03   |   79.31   |   94.57   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-8.0gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-8.0GF-cb4c77ec.pth) |
+| RegNetX-12GF\*        |   46.11   |   12.15  |   79.91   |   94.78   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-12gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/convert/RegNetX-12GF-0574538f.pth) |

-*Models with \* are converted from [pycls](https://github.com/facebookresearch/pycls/blob/master/MODEL_ZOO.md). The config files of these models are only for validation. We don't ensure these config files' training accuracy and welcome you to contribute your reproduction results.*
+*Models with \* are converted from [pycls](https://github.com/facebookresearch/pycls/blob/master/MODEL_ZOO.md). The config files of these models are only for validation.*
--- a/configs/regnet/metafile.yml
+++ b/configs/regnet/metafile.yml
@ -0,0 +1,122 @@
+Collections:
+  - Name: RegNet
+    Metadata:
+      Training Data: ImageNet-1k
+      Architecture:
+        - Neural Architecture Search
+        - Design Space Design
+        - Precise BN
+        - SGD with nesterov
+    Paper:
+      URL: https://arxiv.org/abs/2003.13678
+      Title: Designing Network Design Spaces
+    README: configs/regnet/README.md
+    Code:
+      URL: https://github.com/open-mmlab/mmclassification/blob/v0.18.0/mmcls/models/backbones/regnet.py
+      Version: v0.18.0
+
+Models:
+  - Name: regnetx-400mf_8xb128_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-400mf_8xb128_in1k.py
+    Metadata:
+      FLOPs: 410000000     # 0.41G
+      Parameters: 5160000  # 5.16M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 72.56
+        Top 5 Accuracy: 90.78
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-400mf_8xb128_in1k_20211213-89bfc226.pth
+  - Name: regnetx-800mf_8xb128_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-800mf_8xb128_in1k.py
+    Metadata:
+      FLOPs: 810000000     # 0.81G
+      Parameters: 7260000  # 7.26M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 74.76
+        Top 5 Accuracy: 92.32
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-800mf_8xb128_in1k_20211213-222b0f11.pth
+  - Name: regnetx-1.6gf_8xb128_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-1.6gf_8xb128_in1k.py
+    Metadata:
+      FLOPs: 1630000000     # 1.63G
+      Parameters: 9190000   # 9.19M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 76.84
+        Top 5 Accuracy: 93.31
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-1.6gf_8xb128_in1k_20211213-d1b89758.pth
+  - Name: regnetx-3.2gf_8xb64_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-3.2gf_8xb64_in1k.py
+    Metadata:
+      FLOPs: 1530000000     # 1.53G
+      Parameters: 3210000   # 32.1M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 78.09
+        Top 5 Accuracy: 94.08
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-3.2gf_8xb64_in1k_20211213-1fdd82ae.pth
+  - Name: regnetx-4.0gf_8xb64_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-4.0gf_8xb64_in1k.py
+    Metadata:
+      FLOPs: 4000000000     # 4G
+      Parameters: 22120000  # 22.12M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 78.60
+        Top 5 Accuracy: 94.17
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-4.0gf_8xb64_in1k_20211213-efed675c.pth
+  - Name: regnetx-6.4gf_8xb64_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-6.4gf_8xb64_in1k.py
+    Metadata:
+      FLOPs: 6510000000      # 6.51G
+      Parameters: 26210000   # 26.21M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 79.38
+        Top 5 Accuracy: 94.65
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-6.4gf_8xb64_in1k_20211215-5c6089da.pth
+  - Name: regnetx-8.0gf_8xb64_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-8.0gf_8xb64_in1k.py
+    Metadata:
+      FLOPs: 8030000000     # 8.03G
+      Parameters: 39570000  # 39.57M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 79.12
+        Top 5 Accuracy: 94.51
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-8.0gf_8xb64_in1k_20211213-9a9fcc76.pth
+  - Name: regnetx-12gf_8xb64_in1k
+    In Collection: RegNet
+    Config: configs/regnet/regnetx-12gf_8xb64_in1k.py
+    Metadata:
+      FLOPs: 12150000000      # 12.15G
+      Parameters: 46110000    # 46.11M
+    Results:
+    - Dataset: ImageNet-1k
+      Task: Image Classification
+      Metrics:
+        Top 1 Accuracy: 79.67
+        Top 5 Accuracy: 95.03
+    Weights: https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-12gf_8xb64_in1k_20211213-5df8c2f8.pth
--- a/configs/regnet/regnetx-1.6gf_8xb128_in1k.py
+++ b/configs/regnet/regnetx-1.6gf_8xb128_in1k.py
@ -0,0 +1,6 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_1.6gf'),
+    head=dict(in_channels=912, ))
--- a/configs/regnet/regnetx-1.6gf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-1.6gf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_1.6gf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-12gf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-12gf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_12gf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-12gf_8xb64_in1k.py
+++ b/configs/regnet/regnetx-12gf_8xb64_in1k.py
@ -0,0 +1,11 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_12gf'),
+    head=dict(in_channels=2240, ))
+
+# for batch_size 512, use lr = 0.4
+optimizer = dict(lr=0.4)
+
+data = dict(samples_per_gpu=64, )
--- a/configs/regnet/regnetx-3.2gf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-3.2gf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_3.2gf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-3.2gf_8xb64_in1k.py
+++ b/configs/regnet/regnetx-3.2gf_8xb64_in1k.py
@ -0,0 +1,11 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_3.2gf'),
+    head=dict(in_channels=1008, ))
+
+# for batch_size 512, use lr = 0.4
+optimizer = dict(lr=0.4)
+
+data = dict(samples_per_gpu=64, )
--- a/configs/regnet/regnetx-4.0gf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-4.0gf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_4.0gf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-4.0gf_8xb64_in1k.py
+++ b/configs/regnet/regnetx-4.0gf_8xb64_in1k.py
@ -0,0 +1,11 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_4.0gf'),
+    head=dict(in_channels=1360, ))
+
+# for batch_size 512, use lr = 0.4
+optimizer = dict(lr=0.4)
+
+data = dict(samples_per_gpu=64, )
--- a/configs/regnet/regnetx-400mf_8xb128_in1k.py
+++ b/configs/regnet/regnetx-400mf_8xb128_in1k.py
@ -0,0 +1,77 @@
+_base_ = [
+    '../_base_/models/regnet/regnetx_400mf.py',
+    '../_base_/datasets/imagenet_bs32.py',
+    '../_base_/schedules/imagenet_bs1024_coslr.py',
+    '../_base_/default_runtime.py'
+]
+
+# Precise BN hook will update the bn stats, so this hook should be executed
+# before CheckpointHook, which has priority of 'NORMAL'. So set the
+# priority of PreciseBNHook to 'ABOVE_NORMAL' here.
+custom_hooks = [
+    dict(
+        type='PreciseBNHook',
+        num_samples=8192,
+        interval=1,
+        priority='ABOVE_NORMAL')
+]
+
+# sgd with nesterov, base ls is 0.8 for batch_size 1024,
+# 0.4 for batch_size 512 and 0.2 for batch_size 256 when training ImageNet1k
+optimizer = dict(lr=0.8, nesterov=True)
+
+# dataset settings
+dataset_type = 'ImageNet'
+
+# normalization params, in order of BGR
+NORM_MEAN = [103.53, 116.28, 123.675]
+NORM_STD = [57.375, 57.12, 58.395]
+
+# lighting params, in order of RGB, from repo. pycls
+EIGVAL = [0.2175, 0.0188, 0.0045]
+EIGVEC = [[-0.5675, 0.7192, 0.4009], [-0.5808, -0.0045, -0.814],
+          [-0.5836, -0.6948, 0.4203]]
+
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='RandomResizedCrop', size=224),
+    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
+    dict(
+        type='Lighting',
+        eigval=EIGVAL,
+        eigvec=EIGVEC,
+        alphastd=25.5,  # because the value range of images is [0,255]
+        to_rgb=True
+    ),  # BGR image from cv2 in LoadImageFromFile, convert to RGB here
+    dict(type='Normalize', mean=NORM_MEAN, std=NORM_STD,
+         to_rgb=True),  # RGB2BGR
+    dict(type='ImageToTensor', keys=['img']),
+    dict(type='ToTensor', keys=['gt_label']),
+    dict(type='Collect', keys=['img', 'gt_label'])
+]
+test_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='Resize', size=(256, -1)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Normalize', mean=NORM_MEAN, std=NORM_STD, to_rgb=False),
+    dict(type='ImageToTensor', keys=['img']),
+    dict(type='Collect', keys=['img'])
+]
+data = dict(
+    samples_per_gpu=128,
+    workers_per_gpu=8,
+    train=dict(
+        type=dataset_type,
+        data_prefix='data/imagenet/train',
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        data_prefix='data/imagenet/val',
+        ann_file='data/imagenet/meta/val.txt',
+        pipeline=test_pipeline),
+    test=dict(
+        # replace `data/val` with `data/test` for standard test
+        type=dataset_type,
+        data_prefix='data/imagenet/val',
+        ann_file='data/imagenet/meta/val.txt',
+        pipeline=test_pipeline))
--- a/configs/regnet/regnetx-400mf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-400mf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_400mf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-6.4gf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-6.4gf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_6.4gf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-6.4gf_8xb64_in1k.py
+++ b/configs/regnet/regnetx-6.4gf_8xb64_in1k.py
@ -0,0 +1,11 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_6.4gf'),
+    head=dict(in_channels=1624, ))
+
+# for batch_size 512, use lr = 0.4
+optimizer = dict(lr=0.4)
+
+data = dict(samples_per_gpu=64, )
--- a/configs/regnet/regnetx-8.0gf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-8.0gf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_8.0gf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx-8.0gf_8xb64_in1k.py
+++ b/configs/regnet/regnetx-8.0gf_8xb64_in1k.py
@ -0,0 +1,11 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_8.0gf'),
+    head=dict(in_channels=1920, ))
+
+# for batch_size 512, use lr = 0.4
+optimizer = dict(lr=0.4)
+
+data = dict(samples_per_gpu=64, )
--- a/configs/regnet/regnetx-800mf_8xb128_in1k.py
+++ b/configs/regnet/regnetx-800mf_8xb128_in1k.py
@ -0,0 +1,6 @@
+_base_ = ['./regnetx-400mf_8xb128_in1k.py']
+
+# model settings
+model = dict(
+    backbone=dict(type='RegNet', arch='regnetx_800mf'),
+    head=dict(in_channels=672, ))
--- a/configs/regnet/regnetx-800mf_8xb32_in1k.py
+++ b/configs/regnet/regnetx-800mf_8xb32_in1k.py
@ -1,51 +0,0 @@
-_base_ = [
-    '../_base_/models/regnet/regnetx_800mf.py',
-    '../_base_/datasets/imagenet_bs32.py',
-    '../_base_/schedules/imagenet_bs256.py', '../_base_/default_runtime.py'
-]
-
-# dataset settings
-dataset_type = 'ImageNet'
-
-img_norm_cfg = dict(
-    # The mean and std are used in PyCls when training RegNets
-    mean=[103.53, 116.28, 123.675],
-    std=[57.375, 57.12, 58.395],
-    to_rgb=False)
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', size=224),
-    dict(type='RandomFlip', flip_prob=0.5, direction='horizontal'),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='ToTensor', keys=['gt_label']),
-    dict(type='Collect', keys=['img', 'gt_label'])
-]
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='Resize', size=(256, -1)),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='Normalize', **img_norm_cfg),
-    dict(type='ImageToTensor', keys=['img']),
-    dict(type='Collect', keys=['img'])
-]
-data = dict(
-    samples_per_gpu=32,
-    workers_per_gpu=2,
-    train=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/train',
-        pipeline=train_pipeline),
-    val=dict(
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline),
-    test=dict(
-        # replace `data/val` with `data/test` for standard test
-        type=dataset_type,
-        data_prefix='data/imagenet/val',
-        ann_file='data/imagenet/meta/val.txt',
-        pipeline=test_pipeline))
-evaluation = dict(interval=1, metric='accuracy')
--- a/configs/regnet/regnetx_1.6gf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_1.6gf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-1.6gf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-1.6gf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_12gf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_12gf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-12gf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-12gf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_3.2gf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_3.2gf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-3.2gf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-3.2gf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_4.0gf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_4.0gf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-4.0gf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-4.0gf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_400mf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_400mf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-400mf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-400mf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_6.4gf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_6.4gf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-6.4gf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-6.4gf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_8.0gf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_8.0gf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-8.0gf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-8.0gf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/configs/regnet/regnetx_800mf_b32x8_imagenet.py
+++ b/configs/regnet/regnetx_800mf_b32x8_imagenet.py
@ -1,6 +0,0 @@
-_base_ = 'regnetx-800mf_8xb32_in1k.py'
-
-_deprecation_ = dict(
-    expected='regnetx-800mf_8xb32_in1k.py',
-    reference='https://github.com/open-mmlab/mmclassification/pull/508',
-)
--- a/docs/en/model_zoo.md
+++ b/docs/en/model_zoo.md
@ -48,6 +48,14 @@ The ResNet family models below are trained by standard data augmentations, i.e.,
 | ResNeXt-32x4d-152     | 59.95     | 11.8     | 78.93 | 94.41 | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnext/resnext152-32x4d_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/resnext/resnext152_32x4d_b32x8_imagenet_20210524-927787be.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/resnext/resnext152_32x4d_b32x8_imagenet_20210524-927787be.log.json) |
 | SE-ResNet-50          | 28.09     | 4.13     | 77.74 | 93.84 | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/seresnet/seresnet50_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/se-resnet/se-resnet50_batch256_imagenet_20200804-ae206104.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/se-resnet/se-resnet50_batch256_imagenet_20200708-657b3c36.log.json) |
 | SE-ResNet-101         | 49.33     | 7.86     | 78.26 | 94.07 | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/seresnet/seresnet101_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/se-resnet/se-resnet101_batch256_imagenet_20200804-ba5b51d4.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/se-resnet/se-resnet101_batch256_imagenet_20200708-038a4d04.log.json) |
+| RegNetX-400MF         |   5.16    |   0.41   |   72.56   |   90.78   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-400mf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-400mf_8xb128_in1k_20211213-89bfc226.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-400mf_8xb128_in1k_20211208_143316.log.json) |
+| RegNetX-800MF         |   7.26    |   0.81   |   74.76   |   92.32   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-800mf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-800mf_8xb128_in1k_20211213-222b0f11.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-800mf_8xb128_in1k_20211207_143037.log.json) |
+| RegNetX-1.6GF         |   9.19    |   1.63   |   76.84   |   93.31   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-1.6gf_8xb128_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-1.6gf_8xb128_in1k_20211213-d1b89758.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-1.6gf_8xb128_in1k_20211208_143018.log.json) |
+| RegNetX-3.2GF         |   15.3    |   3.21   |   78.09   |   94.08   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-3.2gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-3.2gf_8xb64_in1k_20211213-1fdd82ae.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-3.2gf_8xb64_in1k_20211208_142720.log.json) |
+| RegNetX-4.0GF         |   22.12   |   4.0    |   78.60   |   94.17   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-4.0gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-4.0gf_8xb64_in1k_20211213-efed675c.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-4.0gf_8xb64_in1k_20211207_150431.log.json) |
+| RegNetX-6.4GF         |   26.21   |   6.51   |   79.38   |   94.65   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-6.4gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-6.4gf_8xb64_in1k_20211215-5c6089da.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-6.4gf_8xb64_in1k_20211213_172748.log.json) |
+| RegNetX-8.0GF         |   39.57   |   8.03   |   79.12   |   94.51   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-8.0gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-8.0gf_8xb64_in1k_20211213-9a9fcc76.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-8.0gf_8xb64_in1k_20211208_103250.log.json) |
+| RegNetX-12GF          |   46.11   |   12.15  |   79.67   |   95.03   | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/regnet/regnetx-12gf_8xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-12gf_8xb64_in1k_20211213-5df8c2f8.pth)  &#124; [log](https://download.openmmlab.com/mmclassification/v0/regnet/regnetx-12gf_8xb64_in1k_20211208_143713.log.json) |
 | ShuffleNetV1 1.0x (group=3)   | 1.87      | 0.146    | 68.13 | 87.81 | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/shufflenet_v1/shufflenet-v1-1x_16xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/shufflenet_v1/shufflenet_v1_batch1024_imagenet_20200804-5d6cec73.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/shufflenet_v1/shufflenet_v1_batch1024_imagenet_20200804-5d6cec73.log.json) |
 | ShuffleNetV2 1.0x     | 2.28      | 0.149    | 69.55 | 88.92 | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/shufflenet_v2/shufflenet-v2-1x_16xb64_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/shufflenet_v2/shufflenet_v2_batch1024_imagenet_20200812-5bf4721e.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/shufflenet_v2/shufflenet_v2_batch1024_imagenet_20200804-8860eec9.log.json) |
 | MobileNet V2          | 3.5       | 0.319    | 71.86 | 90.42 | [config](https://github.com/open-mmlab/mmclassification/blob/master/configs/mobilenet_v2/mobilenet-v2_8xb32_in1k.py) | [model](https://download.openmmlab.com/mmclassification/v0/mobilenet_v2/mobilenet_v2_batch256_imagenet_20200708-3b2dc3af.pth) &#124; [log](https://download.openmmlab.com/mmclassification/v0/mobilenet_v2/mobilenet_v2_batch256_imagenet_20200708-3b2dc3af.log.json) |
--- a/model-index.yml
+++ b/model-index.yml
@ -14,4 +14,5 @@ Import:
  - configs/t2t_vit/metafile.yml
  - configs/mlp_mixer/metafile.yml
  - configs/conformer/metafile.yml
+  - configs/regnet/metafile.yml
  - configs/deit/metafile.yml