add part mmgeneration code

configs/sagan/sagan_128_woReLUinplace_Glr-1e-4_Dlr-4e-4_ndisc1_imagenet1k_b64x4.py

+_base_ = [
+    '../_base_/models/sagan/sagan_128x128.py',
+    '../_base_/datasets/imagenet_128.py', '../_base_/default_runtime.py'
+]
+
+init_cfg = dict(type='studio')
+model = dict(
+    generator=dict(num_classes=1000, init_cfg=init_cfg),
+    discriminator=dict(num_classes=1000, init_cfg=init_cfg),
+)
+
+lr_config = None
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=1000)
+]
+
+inception_pkl = './work_dirs/inception_pkl/imagenet.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='ema'))
+
+n_disc = 1
+total_iters = 1000000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0001, betas=(0.0, 0.999)),
+    discriminator=dict(type='Adam', lr=0.0004, betas=(0.0, 0.999)))
+
+# train on 4 gpus
+data = dict(samples_per_gpu=64)

configs/sagan/sagan_128_woReLUinplace_noaug_bigGAN_Glr-1e-4_Dlr-4e-4_ndisc1_imagenet1k_b32x8.py

+# In this config, we follow the setting `launch_SAGAN_bz128x2_ema.sh` from
+# BigGAN's repo. Please refer to https://github.com/ajbrock/BigGAN-PyTorch/blob/master/scripts/launch_SAGAN_bs128x2_ema.sh  # noqa
+# In summary, in this config:
+# 1. use eps=1e-8 for Spectral Norm
+# 2. not use syncBN
+# 3. not use  Spectral Norm for embedding layers in cBN
+# 4. start EMA at iterations
+# 5. use xavier_uniform for weight initialization
+# 6. no data augmentation
+
+_base_ = [
+    '../_base_/models/sagan/sagan_128x128.py',
+    '../_base_/datasets/imagenet_noaug_128.py', '../_base_/default_runtime.py'
+]
+
+init_cfg = dict(type='BigGAN')
+model = dict(
+    num_classes=1000,
+    generator=dict(
+        num_classes=1000,
+        init_cfg=init_cfg,
+        norm_eps=1e-5,
+        sn_eps=1e-8,
+        auto_sync_bn=False,
+        with_embedding_spectral_norm=False),
+    discriminator=dict(num_classes=1000, init_cfg=init_cfg, sn_eps=1e-8),
+)
+
+n_disc = 1
+train_cfg = dict(disc_step=n_disc, use_ema=True)
+
+lr_config = None
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=1000),
+    dict(
+        type='ExponentialMovingAverageHook',
+        module_keys=('generator_ema'),
+        interval=n_disc,
+        start_iter=2000 * n_disc,
+        interp_cfg=dict(momentum=0.999),
+        priority='VERY_HIGH')
+]
+
+inception_pkl = './work_dirs/inception_pkl/imagenet.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='ema'))
+
+total_iters = 1000000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0001, betas=(0.0, 0.999)),
+    discriminator=dict(type='Adam', lr=0.0004, betas=(0.0, 0.999)))
+
+# train on 8 gpus
+data = dict(samples_per_gpu=32)

configs/sagan/sagan_32_cvt_studioGAN.py

+_base_ = ['../_base_/models/sagan/sagan_32x32.py']

configs/sagan/sagan_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py

+_base_ = [
+    '../_base_/models/sagan/sagan_32x32.py',
+    '../_base_/datasets/cifar10_nopad.py', '../_base_/default_runtime.py'
+]
+
+init_cfg = dict(type='studio')
+model = dict(
+    num_classes=10,
+    generator=dict(
+        num_classes=10,
+        act_cfg=dict(type='ReLU', inplace=True),
+        init_cfg=init_cfg),
+    discriminator=dict(
+        num_classes=10,
+        act_cfg=dict(type='ReLU', inplace=True),
+        init_cfg=init_cfg),
+)
+
+lr_config = None
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=1000)
+]
+
+inception_pkl = './work_dirs/inception_pkl/cifar10.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='orig'))
+
+n_disc = 5
+total_iters = 100000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)),
+    discriminator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)))
+
+data = dict(samples_per_gpu=64)

configs/sagan/sagan_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py

+_base_ = [
+    '../_base_/models/sagan/sagan_32x32.py',
+    '../_base_/datasets/cifar10_nopad.py', '../_base_/default_runtime.py'
+]
+
+init_cfg = dict(type='studio')
+model = dict(
+    num_classes=10,
+    generator=dict(num_classes=10, init_cfg=init_cfg),
+    discriminator=dict(num_classes=10, init_cfg=init_cfg))
+
+lr_config = None
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=1000)
+]
+
+inception_pkl = './work_dirs/inception_pkl/cifar10.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='orig'))
+
+n_disc = 5
+total_iters = 100000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)),
+    discriminator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)))
+
+data = dict(samples_per_gpu=64)

configs/singan/README.md

+# SinGAN
+
+> [Singan: Learning a Generative Model from a Single Natural Image](https://openaccess.thecvf.com/content_ICCV_2019/html/Shaham_SinGAN_Learning_a_Generative_Model_From_a_Single_Natural_Image_ICCV_2019_paper.html)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+<!-- [ABSTRACT] -->
+
+We introduce SinGAN, an unconditional generative model that can be learned from a single natural image. Our model is trained to capture the internal distribution of patches within the image, and is then able to generate high quality, diverse samples that carry the same visual content as the image. SinGAN contains a pyramid of fully convolutional GANs, each responsible for learning the patch distribution at a different scale of the image. This allows generating new samples of arbitrary size and aspect ratio, that have significant variability, yet maintain both the global structure and the fine textures of the training image. In contrast to previous single image GAN schemes, our approach is not limited to texture images, and is not conditional (i.e. it generates samples from noise). User studies confirm that the generated samples are commonly confused to be real images. We illustrate the utility of SinGAN in a wide range of image manipulation tasks.
+
+<!-- [IMAGE] -->
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/28132635/143054395-000ceec1-3be9-4447-b4b9-effc9de94c62.JPG"/>
+</div>
+
+## Results and Models
+
+<div align="center">
+  <b> SinGAN balloons</b>
+  <br/>
+  <img src="https://user-images.githubusercontent.com/12726765/113702715-7861a900-970c-11eb-9dd8-0743cc30701f.png" width="800"/>
+</div>
+
+| Model  |                            Data                             | Num Scales |                            Config                             |                             Download                             |
+| :----: | :---------------------------------------------------------: | :--------: | :-----------------------------------------------------------: | :--------------------------------------------------------------: |
+| SinGAN | [balloons.png](https://download.openmmlab.com/mmgen/dataset/singan/balloons.png) |     8      | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/singan/singan_balloons.py) | [ckpt](https://download.openmmlab.com/mmgen/singan/singan_balloons_20210406_191047-8fcd94cf.pth) \| [pkl](https://download.openmmlab.com/mmgen/singan/singan_balloons_20210406_191047-8fcd94cf.pkl) |
+| SinGAN | [fish.jpg](https://download.openmmlab.com/mmgen/dataset/singan/fish-crop.jpg) |     10     | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/singan/singan_fish.py) | [ckpt](https://download.openmmlab.com/mmgen/singan/singan_fis_20210406_201006-860d91b6.pth) \| [pkl](https://download.openmmlab.com/mmgen/singan/singan_fis_20210406_201006-860d91b6.pkl) |
+| SinGAN | [bohemian.png](https://download.openmmlab.com/mmgen/dataset/singan/bohemian.png) |     10     | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/singan/singan_bohemian.py) | [ckpt](https://download.openmmlab.com/mmgen/singan/singan_bohemian_20210406_175439-f964ee38.pth) \| [pkl](https://download.openmmlab.com/mmgen/singan/singan_bohemian_20210406_175439-f964ee38.pkl) |
+
+## Notes for using SinGAN
+
+When training SinGAN models, users may obtain the number of scales (stages) in advance via the following commands. This number is important for constructing config file, which is related to the generator, discriminator, the training iterations and so on.
+
+```shell
+>>> from mmgen.datasets.singan_dataset import create_real_pyramid
+>>> import mmcv
+>>> real = mmcv.imread('real_img_path')
+>>> _, _, num_scales = create_real_pyramid(real, min_size=25, max_size=300, scale_factor_init=0.75)
+```
+
+When testing SinGAN models, users have to modify the config file to add the `test_cfg`. As shown in `configs/singan/singan_balloons.py`, the only thing you need to do is add the path for `pkl` data. There are some important data containing in the pickle files which you can download from our website.
+
+```python
+test_cfg = dict(
+    _delete_ = True
+    pkl_data = 'path to pkl data'
+)
+```
+
+## Citation
+
+```latex
+@inproceedings{shaham2019singan,
+  title={Singan: Learning a generative model from a single natural image},
+  author={Shaham, Tamar Rott and Dekel, Tali and Michaeli, Tomer},
+  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
+  pages={4570--4580},
+  year={2019},
+  url={https://openaccess.thecvf.com/content_ICCV_2019/html/Shaham_SinGAN_Learning_a_Generative_Model_From_a_Single_Natural_Image_ICCV_2019_paper.html},
+}
+```

configs/singan/metafile.yml

+Collections:
+- Metadata:
+    Architecture:
+    - SinGAN
+  Name: SinGAN
+  Paper:
+  - https://openaccess.thecvf.com/content_ICCV_2019/html/Shaham_SinGAN_Learning_a_Generative_Model_From_a_Single_Natural_Image_ICCV_2019_paper.html
+  README: configs/singan/README.md
+Models:
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/singan/singan_balloons.py
+  In Collection: SinGAN
+  Metadata:
+    Training Data: Others
+  Name: singan_balloons
+  Results:
+  - Dataset: Others
+    Metrics:
+      Num Scales: 8.0
+    Task: Internal Learning
+  Weights: https://download.openmmlab.com/mmgen/singan/singan_balloons_20210406_191047-8fcd94cf.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/singan/singan_fish.py
+  In Collection: SinGAN
+  Metadata:
+    Training Data: Others
+  Name: singan_fish
+  Results:
+  - Dataset: Others
+    Metrics:
+      Num Scales: 10.0
+    Task: Internal Learning
+  Weights: https://download.openmmlab.com/mmgen/singan/singan_fis_20210406_201006-860d91b6.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/singan/singan_bohemian.py
+  In Collection: SinGAN
+  Metadata:
+    Training Data: Others
+  Name: singan_bohemian
+  Results:
+  - Dataset: Others
+    Metrics:
+      Num Scales: 10.0
+    Task: Internal Learning
+  Weights: https://download.openmmlab.com/mmgen/singan/singan_bohemian_20210406_175439-f964ee38.pth

configs/singan/singan_balloons.py

+_base_ = [
+    '../_base_/models/singan/singan.py', '../_base_/datasets/singan.py',
+    '../_base_/default_runtime.py'
+]
+
+num_scales = 8  # start from zero
+model = dict(
+    generator=dict(num_scales=num_scales),
+    discriminator=dict(num_scales=num_scales))
+
+train_cfg = dict(
+    noise_weight_init=0.1,
+    iters_per_scale=2000,
+)
+
+# test_cfg = dict(
+#     _delete_ = True
+#     pkl_data = 'path to pkl data'
+# )
+
+data = dict(train=dict(img_path='./data/singan/balloons.png'))
+
+optimizer = None
+lr_config = None
+checkpoint_config = dict(by_epoch=False, interval=2000, max_keep_ckpts=3)
+
+custom_hooks = [
+    dict(
+        type='MMGenVisualizationHook',
+        output_dir='visual',
+        interval=500,
+        bgr2rgb=True,
+        res_name_list=['fake_imgs', 'recon_imgs', 'real_imgs']),
+    dict(
+        type='PickleDataHook',
+        output_dir='pickle',
+        interval=-1,
+        after_run=True,
+        data_name_list=['noise_weights', 'fixed_noises', 'curr_stage'])
+]
+
+total_iters = 18000

configs/singan/singan_bohemian.py

+_base_ = [
+    '../_base_/models/singan/singan.py', '../_base_/datasets/singan.py',
+    '../_base_/default_runtime.py'
+]
+
+num_scales = 10  # start from zero
+model = dict(
+    generator=dict(num_scales=num_scales),
+    discriminator=dict(num_scales=num_scales))
+
+train_cfg = dict(
+    noise_weight_init=0.1,
+    iters_per_scale=2000,
+)
+
+# test_cfg = dict(
+#     _delete_ = True
+#     pkl_data = 'path to pkl data'
+# )
+
+data = dict(
+    train=dict(
+        img_path='./data/singan/bohemian.png', min_size=25, max_size=500))
+
+optimizer = None
+lr_config = None
+checkpoint_config = dict(by_epoch=False, interval=2000, max_keep_ckpts=3)
+
+custom_hooks = [
+    dict(
+        type='MMGenVisualizationHook',
+        output_dir='visual',
+        interval=500,
+        bgr2rgb=True,
+        res_name_list=['fake_imgs', 'recon_imgs', 'real_imgs']),
+    dict(
+        type='PickleDataHook',
+        output_dir='pickle',
+        interval=-1,
+        after_run=True,
+        data_name_list=['noise_weights', 'fixed_noises', 'curr_stage'])
+]
+
+total_iters = 22000

configs/singan/singan_fish.py

+_base_ = [
+    '../_base_/models/singan/singan.py', '../_base_/datasets/singan.py',
+    '../_base_/default_runtime.py'
+]
+
+num_scales = 10  # start from zero
+model = dict(
+    generator=dict(num_scales=num_scales),
+    discriminator=dict(num_scales=num_scales))
+
+train_cfg = dict(
+    noise_weight_init=0.1,
+    iters_per_scale=2000,
+)
+
+# test_cfg = dict(
+#     _delete_ = True
+#     pkl_data = 'path to pkl data'
+# )
+
+data = dict(
+    train=dict(
+        img_path='./data/singan/fish-crop.jpg', min_size=25, max_size=300))
+
+optimizer = None
+lr_config = None
+checkpoint_config = dict(by_epoch=False, interval=2000, max_keep_ckpts=3)
+
+custom_hooks = [
+    dict(
+        type='MMGenVisualizationHook',
+        output_dir='visual',
+        interval=500,
+        bgr2rgb=True,
+        res_name_list=['fake_imgs', 'recon_imgs', 'real_imgs']),
+    dict(
+        type='PickleDataHook',
+        output_dir='pickle',
+        interval=-1,
+        after_run=True,
+        data_name_list=['noise_weights', 'fixed_noises', 'curr_stage'])
+]
+
+total_iters = 22000

configs/sngan_proj/README.md

+# SNGAN
+
+> [Spectral Normalization for Generative Adversarial Networks](https://openreview.net/forum?id=B1QRgziT-)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+<!-- [ABSTRACT] -->
+
+One of the challenges in the study of generative adversarial networks is the instability of its training. In this paper, we propose a novel weight normalization technique called spectral normalization to stabilize the training of the discriminator. Our new normalization technique is computationally light and easy to incorporate into existing implementations. We tested the efficacy of spectral normalization on CIFAR10, STL-10, and ILSVRC2012 dataset, and we experimentally confirmed that spectrally normalized GANs (SN-GANs) is capable of generating images of better or equal quality relative to the previous training stabilization techniques.
+
+<!-- [IMAGE] -->
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/28132635/143154496-6a03def4-4507-4d80-a948-89a5b747d916.png"/>
+</div>
+
+## Results and models
+
+<div align="center">
+  <b> Results from our SNGAN-PROJ trained in CIFAR10 and ImageNet</b>
+  <br/>
+  <img src="https://user-images.githubusercontent.com/28132635/125151484-14220b80-e179-11eb-81f7-9391ccaeb841.png" width="400"/> &nbsp;&nbsp;
+  <img src="https://user-images.githubusercontent.com/28132635/127621152-7b7a0f2c-c743-485a-bf2e-2beca849a6e6.png" width="400"/>
+</div>
+
+|               Models               | Dataset  | Inplace ReLU | disc_step | Total Iters\* |  Iter  |   IS    |   FID   |               Config                |               Download                |               Log                |
+| :--------------------------------: | :------: | :----------: | :-------: | :-----------: | :----: | :-----: | :-----: | :---------------------------------: | :-----------------------------------: | :------------------------------: |
+| SNGAN_Proj-32x32-woInplaceReLU Best IS | CIFAR10  |     w/o      |     5     |    500000     | 400000 | 9.6919  | 9.8203  | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_woReLUinplace_is-iter400000_20210709_163823-902ce1ae.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_woReLUinplace_20210624_065306_fid-ba0862a0_is-902ce1ae.json) |
+| SNGAN_Proj-32x32-woInplaceReLU Best FID | CIFAR10  |     w/o      |     5     |    500000     | 490000 | 9.5659  | 8.1158  | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_woReLUinplace_fid-iter490000_20210709_163329-ba0862a0.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_woReLUinplace_20210624_065306_fid-ba0862a0_is-902ce1ae.json) |
+| SNGAN_Proj-32x32-wInplaceReLU Best IS | CIFAR10  |      w       |     5     |    500000     | 490000 | 9.5564  | 8.3462  | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_wReLUinplace_is-iter490000_20210709_202230-cd863c74.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_wReLUinplace_20210624_065306_fid-ba0862a0_is-902ce1ae.json) |
+| SNGAN_Proj-32x32-wInplaceReLU Best FID | CIFAR10  |      w       |     5     |    500000     | 490000 | 9.5564  | 8.3462  | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4-b64x1_wReLUinplace_fid-iter490000_20210709_203038-191b2648.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_wReLUinplace_20210624_065306_fid-ba0862a0_is-902ce1ae.json) |
+| SNGAN_Proj-128x128-woInplaceReLU Best IS | ImageNet |     w/o      |     5     |    1000000    | 952000 | 30.0651 | 33.4682 | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_woReLUinplace_is-iter952000_20210730_132027-9c884a21.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_woReLUinplace_20210730_131424_fid-061bf803_is-9c884a21.json) |
+| SNGAN_Proj-128x128-woInplaceReLU Best FID | ImageNet |     w/o      |     5     |    1000000    | 989000 | 29.5779 | 32.6193 | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_woReLUinplace_fid-iter988000_20210730_131424-061bf803.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_woReLUinplace_20210730_131424_fid-061bf803_is-9c884a21.json) |
+| SNGAN_Proj-128x128-wInplaceReLU Best IS | ImageNet |      w       |     5     |    1000000    | 944000 | 28.1799 | 34.3383 | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_wReLUinplace_is-iter944000_20210730_132714-ca0ccd07.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_wReLUinplace_20210730_132401_fid-9a682411_is-ca0ccd07.json) |
+| SNGAN_Proj-128x128-wInplaceReLU Best FID | ImageNet |      w       |     5     |    1000000    | 988000 | 27.7948 | 33.4821 | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py) | [ckpt](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_wReLUinplace_fid-iter988000_20210730_132401-9a682411.pth) | [Log](https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_wReLUinplace_20210730_132401_fid-9a682411_is-ca0ccd07.json) |
+
+'\*' Iteration counting rule in our implementation is different from others. If you want to align with other codebases, you can use the following conversion formula:
+
+```
+total_iters (biggan/pytorch studio gan) = our_total_iters / disc_step
+```
+
+We also provide converted pre-train models from [Pytorch-StudioGAN](https://github.com/POSTECH-CVLab/PyTorch-StudioGAN).
+To be noted that, in Pytorch Studio GAN, **inplace ReLU** is used in generator and discriminator.
+
+|       Models        | Dataset  | Inplace ReLU | disc_step | Total Iters | IS (Our Pipeline) | FID (Our Pipeline) | IS (StudioGAN) | FID (StudioGAN) |       Config        |        Download        |        Original Download link        |
+| :-----------------: | :------: | :----------: | :-------: | :---------: | :---------------: | :----------------: | :------------: | :-------------: | :-----------------: | :--------------------: | :----------------------------------: |
+| SAGAN_Proj-32x32 StudioGAN | CIFAR10  |      w       |     5     |   100000    |       9.372       |      10.2011       |     8.677      |     13.248      | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_cvt_studioGAN.py) | [model](https://download.openmmlab.com/mmgen/sngan_proj/sngan_cifar10_convert-studio-rgb_20210709_111346-2979202d.pth) | [model](https://drive.google.com/drive/folders/16s5Cr-V-NlfLyy_uyXEkoNxLBt-8wYSM) |
+| SAGAN_Proj-128x128 StudioGAN | ImageNet |      w       |     2     |   1000000   |      30.218       |      29.8199       |     32.247     |     26.792      | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_cvt_studioGAN.py) | [model](https://download.openmmlab.com/mmgen/sngan_proj/sngan_imagenet1k_convert-studio-rgb_20210709_111406-877b1130.pth) | [model](https://drive.google.com/drive/folders/1Ek2wAMlxpajL_M8aub4DKQ9B313K8XhS) |
+
+- `Our Pipeline` denote results evaluated with our pipeline.
+- `StudioGAN` denote results released by Pytorch-StudioGAN.
+
+For IS metric, our implementation is different from PyTorch-Studio GAN in the following aspects:
+
+1. We use [Tero's Inception](https://nvlabs-fi-cdn.nvidia.com/stylegan2-ada-pytorch/pretrained/metrics/inception-2015-12-05.pt) for feature extraction.
+2. We use bicubic interpolation with PIL backend to resize image before feed them to Inception.
+
+For FID evaluation, we follow the pipeline of [BigGAN](https://github.com/ajbrock/BigGAN-PyTorch/blob/98459431a5d618d644d54cd1e9fceb1e5045648d/calculate_inception_moments.py#L52), where the whole training set is adopted to extract inception statistics, and Pytorch Studio GAN uses 50000 randomly selected samples. Besides, we also use [Tero's Inception](https://nvlabs-fi-cdn.nvidia.com/stylegan2-ada-pytorch/pretrained/metrics/inception-2015-12-05.pt) for feature extraction.
+
+You can download the preprocessed inception state by the following url: [CIFAR10](https://download.openmmlab.com/mmgen/evaluation/fid_inception_pkl/cifar10.pkl) and [ImageNet1k](https://download.openmmlab.com/mmgen/evaluation/fid_inception_pkl/imagenet.pkl).
+
+You can use following commands to extract those inception states by yourself.
+
+```
+# For CIFAR10
+python tools/utils/inception_stat.py --data-cfg configs/_base_/datasets/cifar10_inception_stat.py --pklname cifar10.pkl --no-shuffle --inception-style stylegan --num-samples -1 --subset train
+
+# For ImageNet1k
+python tools/utils/inception_stat.py --data-cfg configs/_base_/datasets/imagenet_128x128_inception_stat.py --pklname imagenet.pkl --no-shuffle --inception-style stylegan --num-samples -1 --subset train
+```
+
+## Citation
+
+```latex
+@inproceedings{miyato2018spectral,
+  title={Spectral Normalization for Generative Adversarial Networks},
+  author={Miyato, Takeru and Kataoka, Toshiki and Koyama, Masanori and Yoshida, Yuichi},
+  booktitle={International Conference on Learning Representations},
+  year={2018},
+  url={https://openreview.net/forum?id=B1QRgziT-},
+}
+```

configs/sngan_proj/metafile.yml

+Collections:
+- Metadata:
+    Architecture:
+    - SNGAN
+  Name: SNGAN
+  Paper:
+  - https://openreview.net/forum?id=B1QRgziT-
+  README: configs/sngan_proj/README.md
+Models:
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: CIFAR
+  Name: sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1
+  Results:
+  - Dataset: CIFAR
+    Metrics:
+      FID: 9.8203
+      IS: 9.6919
+      Inplace ReLU: w/o
+      Iter: 400000.0
+      Log: '[Log]'
+      Total Iters\*: 500000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_woReLUinplace_is-iter400000_20210709_163823-902ce1ae.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: CIFAR
+  Name: sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1
+  Results:
+  - Dataset: CIFAR
+    Metrics:
+      FID: 8.1158
+      IS: 9.5659
+      Inplace ReLU: w/o
+      Iter: 490000.0
+      Log: '[Log]'
+      Total Iters\*: 500000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_woReLUinplace_fid-iter490000_20210709_163329-ba0862a0.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: CIFAR
+  Name: sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1
+  Results:
+  - Dataset: CIFAR
+    Metrics:
+      FID: 8.3462
+      IS: 9.5564
+      Inplace ReLU: w
+      Iter: 490000.0
+      Log: '[Log]'
+      Total Iters\*: 500000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4_b64x1_wReLUinplace_is-iter490000_20210709_202230-cd863c74.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: CIFAR
+  Name: sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1
+  Results:
+  - Dataset: CIFAR
+    Metrics:
+      FID: 8.3462
+      IS: 9.5564
+      Inplace ReLU: w
+      Iter: 490000.0
+      Log: '[Log]'
+      Total Iters\*: 500000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_cifar10_32_lr-2e-4-b64x1_wReLUinplace_fid-iter490000_20210709_203038-191b2648.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: IMAGENET
+  Name: sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2
+  Results:
+  - Dataset: IMAGENET
+    Metrics:
+      FID: 33.4682
+      IS: 30.0651
+      Inplace ReLU: w/o
+      Iter: 952000.0
+      Log: '[Log]'
+      Total Iters\*: 1000000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_woReLUinplace_is-iter952000_20210730_132027-9c884a21.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: IMAGENET
+  Name: sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2
+  Results:
+  - Dataset: IMAGENET
+    Metrics:
+      FID: 32.6193
+      IS: 29.5779
+      Inplace ReLU: w/o
+      Iter: 989000.0
+      Log: '[Log]'
+      Total Iters\*: 1000000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_woReLUinplace_fid-iter988000_20210730_131424-061bf803.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: IMAGENET
+  Name: sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2
+  Results:
+  - Dataset: IMAGENET
+    Metrics:
+      FID: 34.3383
+      IS: 28.1799
+      Inplace ReLU: w
+      Iter: 944000.0
+      Log: '[Log]'
+      Total Iters\*: 1000000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_wReLUinplace_is-iter944000_20210730_132714-ca0ccd07.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: IMAGENET
+  Name: sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2
+  Results:
+  - Dataset: IMAGENET
+    Metrics:
+      FID: 33.4821
+      IS: 27.7948
+      Inplace ReLU: w
+      Iter: 988000.0
+      Log: '[Log]'
+      Total Iters\*: 1000000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_proj_imagenet1k_128_Glr2e-4_Dlr5e-5_ndisc5_b128x2_wReLUinplace_fid-iter988000_20210730_132401-9a682411.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_cvt_studioGAN.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: Others
+  Name: sngan_proj_32_cvt_studioGAN
+  Results:
+  - Dataset: Others
+    Metrics:
+      FID (Our Pipeline): 10.2011
+      FID (StudioGAN): 13.248
+      IS (Our Pipeline): 9.372
+      IS (StudioGAN): 8.677
+      Inplace ReLU: w
+      Total Iters: 100000.0
+      disc_step: 5.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_cifar10_convert-studio-rgb_20210709_111346-2979202d.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/sngan_proj/sngan_proj_32_cvt_studioGAN.py
+  In Collection: SNGAN
+  Metadata:
+    Training Data: Others
+  Name: sngan_proj_32_cvt_studioGAN
+  Results:
+  - Dataset: Others
+    Metrics:
+      FID (Our Pipeline): 29.8199
+      FID (StudioGAN): 26.792
+      IS (Our Pipeline): 30.218
+      IS (StudioGAN): 32.247
+      Inplace ReLU: w
+      Total Iters: 1000000.0
+      disc_step: 2.0
+    Task: Conditional GANs
+  Weights: https://download.openmmlab.com/mmgen/sngan_proj/sngan_imagenet1k_convert-studio-rgb_20210709_111406-877b1130.pth

configs/sngan_proj/sngan_proj_128_cvt_studioGAN.py

+_base_ = ['../_base_/models/sngan_proj/sngan_proj_128x128.py']

configs/sngan_proj/sngan_proj_128_wReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py

+_base_ = [
+    '../_base_/models/sngan_proj/sngan_proj_128x128.py',
+    '../_base_/datasets/imagenet_128.py', '../_base_/default_runtime.py'
+]
+
+num_classes = 1000
+init_cfg = dict(type='studio')
+model = dict(
+    num_classes=num_classes,
+    generator=dict(
+        num_classes=num_classes,
+        act_cfg=dict(type='ReLU', inplace=True),
+        init_cfg=init_cfg),
+    discriminator=dict(
+        num_classes=num_classes,
+        act_cfg=dict(type='ReLU', inplace=True),
+        init_cfg=init_cfg))
+
+n_disc = 5
+train_cfg = dict(disc_steps=n_disc)
+lr_config = None
+
+checkpoint_config = dict(interval=50000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000)
+]
+
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+
+inception_pkl = './work_dirs/inception_pkl/imagenet.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=dict(milestones=[800000], interval=[10000, 4000]),
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='orig'))
+
+total_iters = 500000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0002, betas=(0.0, 0.999)),
+    discriminator=dict(type='Adam', lr=0.00005, betas=(0.0, 0.999)))
+
+# train on 2 gpus
+data = dict(samples_per_gpu=128)

configs/sngan_proj/sngan_proj_128_woReLUinplace_Glr-2e-4_Dlr-5e-5_ndisc5_imagenet1k_b128x2.py

+_base_ = [
+    '../_base_/models/sngan_proj/sngan_proj_128x128.py',
+    '../_base_/datasets/imagenet_128.py', '../_base_/default_runtime.py'
+]
+
+num_classes = 1000
+init_cfg = dict(type='studio')
+model = dict(
+    num_classes=num_classes,
+    generator=dict(num_classes=num_classes, init_cfg=init_cfg),
+    discriminator=dict(num_classes=num_classes, init_cfg=init_cfg))
+
+n_disc = 5
+train_cfg = dict(disc_steps=n_disc)
+lr_config = None
+
+checkpoint_config = dict(interval=50000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000)
+]
+
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+
+inception_pkl = './work_dirs/inception_pkl/imagenet.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=dict(milestones=[800000], interval=[10000, 4000]),
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='orig'))
+
+total_iters = 500000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0002, betas=(0.0, 0.999)),
+    discriminator=dict(type='Adam', lr=0.00005, betas=(0.0, 0.999)))
+
+# train on 2 gpus
+data = dict(samples_per_gpu=128)

configs/sngan_proj/sngan_proj_32_cvt_studioGAN.py

+_base_ = ['../_base_/models/sngan_proj/sngan_proj_32x32.py']

configs/sngan_proj/sngan_proj_32_wReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py

+# follow pytorch GAN-Studio, random flip is used in the dataset
+
+_base_ = [
+    '../_base_/models/sngan_proj/sngan_proj_32x32.py',
+    '../_base_/datasets/cifar10_nopad.py', '../_base_/default_runtime.py'
+]
+
+num_classes = 10
+init_cfg = dict(type='studio')
+model = dict(
+    num_classes=num_classes,
+    generator=dict(
+        act_cfg=dict(type='ReLU', inplace=True),
+        num_classes=num_classes,
+        init_cfg=init_cfg),
+    discriminator=dict(
+        act_cfg=dict(type='ReLU', inplace=True),
+        num_classes=num_classes,
+        init_cfg=init_cfg))
+
+n_disc = 5
+lr_config = None
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000)
+]
+
+inception_pkl = './work_dirs/inception_pkl/cifar10.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='orig'))
+
+total_iters = 100000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)),
+    discriminator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)))
+
+data = dict(samples_per_gpu=64)

configs/sngan_proj/sngan_proj_32_woReLUinplace_lr-2e-4_ndisc5_cifar10_b64x1.py

+# follow pytorch GAN-Studio, random flip is used in the dataset
+
+_base_ = [
+    '../_base_/models/sngan_proj/sngan_proj_32x32.py',
+    '../_base_/datasets/cifar10_nopad.py', '../_base_/default_runtime.py'
+]
+
+num_classes = 10
+init_cfg = dict(type='studio')
+model = dict(
+    num_classes=num_classes,
+    generator=dict(num_classes=num_classes, init_cfg=init_cfg),
+    discriminator=dict(num_classes=num_classes, init_cfg=init_cfg))
+
+n_disc = 5
+lr_config = None
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=20)
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000)
+]
+
+inception_pkl = './work_dirs/inception_pkl/cifar10.pkl'
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=[
+        dict(
+            type='FID',
+            num_images=50000,
+            inception_pkl=inception_pkl,
+            bgr2rgb=True,
+            inception_args=dict(type='StyleGAN')),
+        dict(type='IS', num_images=50000)
+    ],
+    best_metric=['fid', 'is'],
+    sample_kwargs=dict(sample_model='orig'))
+
+total_iters = 100000 * n_disc
+# use ddp wrapper for faster training
+use_ddp_wrapper = True
+find_unused_parameters = False
+
+runner = dict(
+    type='DynamicIterBasedRunner',
+    is_dynamic_ddp=False,  # Note that this flag should be False.
+    pass_training_status=True)
+
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN')),
+    IS50k=dict(type='IS', num_images=50000))
+
+optimizer = dict(
+    generator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)),
+    discriminator=dict(type='Adam', lr=0.0002, betas=(0.5, 0.999)))
+
+data = dict(samples_per_gpu=64)

configs/styleganv1/README.md

+# StyleGANv1
+
+> [A Style-Based Generator Architecture for Generative Adversarial Networks](https://openaccess.thecvf.com/content_CVPR_2019/html/Karras_A_Style-Based_Generator_Architecture_for_Generative_Adversarial_Networks_CVPR_2019_paper.html)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+<!-- [ABSTRACT] -->
+
+We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces.
+
+<!-- [IMAGE] -->
+
+<div align=center>
+<img src="https://user-images.githubusercontent.com/28132635/143055313-f4988870-2963-4a2f-916e-0de0e04eb474.JPG"/>
+</div>
+
+## Results and Models
+
+<div align="center">
+  <b> Results (compressed) from StyleGANv1 trained by MMGeneration</b>
+  <br/>
+  <img src="https://user-images.githubusercontent.com/12726765/113845642-4f9ee980-97c8-11eb-85c7-49d6d21dd46b.png" width="800"/>
+</div>
+
+|        Model         | FID50k |  P&R50k_full  |                                     Config                                      |                                      Download                                      |
+| :------------------: | :----: | :-----------: | :-----------------------------------------------------------------------------: | :--------------------------------------------------------------------------------: |
+| styleganv1_ffhq_256  | 6.090  | 70.228/27.050 | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv1/styleganv1_ffhq_256_g8_25Mimg.py) | [model](https://download.openmmlab.com/mmgen/styleganv1/styleganv1_ffhq_256_g8_25Mimg_20210407_161748-0094da86.pth) |
+| styleganv1_ffhq_1024 | 4.056  | 70.302/36.869 | [config](https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv1/styleganv1_ffhq_1024_g8_25Mimg.py) | [model](https://download.openmmlab.com/mmgen/styleganv1/styleganv1_ffhq_1024_g8_25Mimg_20210407_161627-850a7234.pth) |
+
+## Citation
+
+```latex
+@inproceedings{karras2019style,
+  title={A style-based generator architecture for generative adversarial networks},
+  author={Karras, Tero and Laine, Samuli and Aila, Timo},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={4401--4410},
+  year={2019},
+  url={https://openaccess.thecvf.com/content_CVPR_2019/html/Karras_A_Style-Based_Generator_Architecture_for_Generative_Adversarial_Networks_CVPR_2019_paper.html},
+}
+```

configs/styleganv1/metafile.yml

+Collections:
+- Metadata:
+    Architecture:
+    - StyleGANv1
+  Name: StyleGANv1
+  Paper:
+  - https://openaccess.thecvf.com/content_CVPR_2019/html/Karras_A_Style-Based_Generator_Architecture_for_Generative_Adversarial_Networks_CVPR_2019_paper.html
+  README: configs/styleganv1/README.md
+Models:
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv1/styleganv1_ffhq_256_g8_25Mimg.py
+  In Collection: StyleGANv1
+  Metadata:
+    Training Data: FFHQ
+  Name: styleganv1_ffhq_256_g8_25Mimg
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      FID50k: 6.09
+      P&R50k_full: 70.228/27.050
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/styleganv1/styleganv1_ffhq_256_g8_25Mimg_20210407_161748-0094da86.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv1/styleganv1_ffhq_1024_g8_25Mimg.py
+  In Collection: StyleGANv1
+  Metadata:
+    Training Data: FFHQ
+  Name: styleganv1_ffhq_1024_g8_25Mimg
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      FID50k: 4.056
+      P&R50k_full: 70.302/36.869
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/styleganv1/styleganv1_ffhq_1024_g8_25Mimg_20210407_161627-850a7234.pth