stylegan2_mmcv

build/lib/mmgen/.mim/configs/styleganv2/stylegan2_c2_lsun-horse_256_b4x8_800k.py

+"""Note that this config is just for testing."""
+
+_base_ = [
+    '../_base_/datasets/lsun_stylegan.py',
+    '../_base_/models/stylegan/stylegan2_base.py',
+    '../_base_/default_runtime.py'
+]
+
+model = dict(generator=dict(out_size=256), discriminator=dict(in_size=256))
+
+data = dict(
+    samples_per_gpu=4, train=dict(dataset=dict(imgs_root='./data/lsun-horse')))
+
+ema_half_life = 10.  # G_smoothing_kimg
+
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000),
+    dict(
+        type='ExponentialMovingAverageHook',
+        module_keys=('generator_ema', ),
+        interval=1,
+        interp_cfg=dict(momentum=0.5**(32. / (ema_half_life * 1000.))),
+        priority='VERY_HIGH')
+]
+
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=30)
+lr_config = None
+
+log_config = dict(
+    interval=100,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook'),
+    ])
+
+total_iters = 800002  # need to modify
+
+metrics = dict(
+    fid50k=dict(
+        type='FID', num_images=50000, inception_pkl=None, bgr2rgb=True),
+    pr50k3=dict(type='PR', num_images=50000, k=3),
+    ppl_wend=dict(type='PPL', space='W', sampling='end', num_images=50000))

build/lib/mmgen/.mim/configs/styleganv3/metafile.yml

+Collections:
+- Metadata:
+    Architecture:
+    - StyleGANv3
+  Name: StyleGANv3
+  Paper:
+  - https://nvlabs-fi-cdn.nvidia.com/stylegan3/stylegan3-paper.pdf
+  README: configs/styleganv3/README.md
+Models:
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv3/stylegan3_t_noaug_fp16_gamma32.8_ffhq_1024_b4x8.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: FFHQ
+  Name: stylegan3_t_noaug_fp16_gamma32.8_ffhq_1024_b4x8
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      FID50k: 3.37<sup>\</sup>
+      Iter: 490000.0
+      Log: '[log]'
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_noaug_fp16_gamma32.8_ffhq_1024_b4x8_best_fid_iter_490000_20220401_120733-4ff83434.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv3/stylegan3_t_ada_fp16_gamma6.6_metfaces_1024_b4x8.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: Others
+  Name: stylegan3_t_ada_fp16_gamma6.6_metfaces_1024_b4x8
+  Results:
+  - Dataset: Others
+    Metrics:
+      FID50k: 15.09
+      Iter: 130000.0
+      Log: '[log]'
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_ada_fp16_gamma6.6_metfaces_1024_b4x8_best_fid_iter_130000_20220401_115101-f2ef498e.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/styleganv3/stylegan3_t_noaug_fp16_gamma2.0_ffhq_256_b4x8.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: FFHQ
+  Name: stylegan3_t_noaug_fp16_gamma2.0_ffhq_256_b4x8
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      FID50k: 7.65
+      Iter: 740000.0
+      Log: '[log]'
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_noaug_fp16_gamma2.0_ffhq_256_b4x8_best_fid_iter_740000_20220401_122456-730e1fba.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/_base_/models/stylegan/stylegan3_t_ffhqu_256_b4x8_cvt_official_rgb.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: FFHQ
+  Name: stylegan3_t_ffhqu_256_b4x8_cvt_official_rgb
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      Comment: official weight
+      EQ-R: 13.12
+      EQ-T: 63.01
+      FID50k: 4.62
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_ffhqu_256_b4x8_cvt_official_rgb_20220329_235046-153df4c8.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/_base_/models/stylegan/stylegan3_t_afhqv2_512_b4x8_cvt_official_rgb.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: Others
+  Name: stylegan3_t_afhqv2_512_b4x8_cvt_official_rgb
+  Results:
+  - Dataset: Others
+    Metrics:
+      Comment: official weight
+      EQ-R: 13.51
+      EQ-T: 60.15
+      FID50k: 4.04
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_afhqv2_512_b4x8_cvt_official_rgb_20220329_235017-ee6b037a.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/_base_/models/stylegan/stylegan3_t_ffhq_1024_b4x8_cvt_official_rgb.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: FFHQ
+  Name: stylegan3_t_ffhq_1024_b4x8_cvt_official_rgb
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      Comment: official weight
+      EQ-R: 13.82
+      EQ-T: 61.21
+      FID50k: 2.79
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_ffhq_1024_b4x8_cvt_official_rgb_20220329_235113-db6c6580.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/_base_/models/stylegan/stylegan3_r_ffhqu_256_b4x8_cvt_official_rgb.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: FFHQ
+  Name: stylegan3_r_ffhqu_256_b4x8_cvt_official_rgb
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      Comment: official weight
+      EQ-R: 40.48
+      EQ-T: 66.65
+      FID50k: 4.5
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_r_ffhqu_256_b4x8_cvt_official_rgb_20220329_234909-4521d963.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/_base_/models/stylegan/stylegan3_r_afhqv2_512_b4x8_cvt_official_rgb.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: Others
+  Name: stylegan3_r_afhqv2_512_b4x8_cvt_official_rgb
+  Results:
+  - Dataset: Others
+    Metrics:
+      Comment: official weight
+      EQ-R: 40.34
+      EQ-T: 64.89
+      FID50k: 4.4
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_r_afhqv2_512_b4x8_cvt_official_rgb_20220329_234829-f2eaca72.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/_base_/models/stylegan/stylegan3_r_ffhq_1024_b4x8_cvt_official_rgb.py
+  In Collection: StyleGANv3
+  Metadata:
+    Training Data: FFHQ
+  Name: stylegan3_r_ffhq_1024_b4x8_cvt_official_rgb
+  Results:
+  - Dataset: FFHQ
+    Metrics:
+      Comment: official weight
+      EQ-R: 46.62
+      EQ-T: 64.76
+      FID50k: 3.07
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/stylegan3/stylegan3_r_ffhq_1024_b4x8_cvt_official_rgb_20220329_234933-ac0500a1.pth

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_r_ada_fp16_gamma3.3_metfaces_1024_b4x8.py

+_base_ = [
+    '../_base_/models/stylegan/stylegan3_base.py',
+    '../_base_/datasets/ffhq_flip.py', '../_base_/default_runtime.py'
+]
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 65536,
+    'channel_max': 1024,
+    'magnitude_ema_beta': 0.999,
+    'conv_kernel': 1,
+    'use_radial_filters': True
+}
+r1_gamma = 3.3  # set by user
+d_reg_interval = 16
+
+load_from = 'https://download.openmmlab.com/mmgen/stylegan3/stylegan3_r_ffhq_1024_b4x8_cvt_official_rgb_20220329_234933-ac0500a1.pth'  # noqa
+
+# ada settings
+aug_kwargs = {
+    'xflip': 1,
+    'rotate90': 1,
+    'xint': 1,
+    'scale': 1,
+    'rotate': 1,
+    'aniso': 1,
+    'xfrac': 1,
+    'brightness': 1,
+    'contrast': 1,
+    'lumaflip': 1,
+    'hue': 1,
+    'saturation': 1
+}
+
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=1024,
+        img_channels=3,
+        rgb2bgr=True,
+        synthesis_cfg=synthesis_cfg),
+    discriminator=dict(
+        type='ADAStyleGAN2Discriminator',
+        in_size=1024,
+        input_bgr2rgb=True,
+        data_aug=dict(type='ADAAug', aug_pipeline=aug_kwargs, ada_kimg=100)),
+    gan_loss=dict(type='GANLoss', gan_type='wgan-logistic-ns'),
+    disc_auxiliary_loss=dict(loss_weight=r1_gamma / 2.0 * d_reg_interval))
+
+imgs_root = 'data/metfaces/images/'
+data = dict(
+    samples_per_gpu=4,
+    train=dict(dataset=dict(imgs_root=imgs_root)),
+    val=dict(imgs_root=imgs_root))
+
+ema_half_life = 10.  # G_smoothing_kimg
+
+ema_kimg = 10
+ema_nimg = ema_kimg * 1000
+ema_beta = 0.5**(32 / max(ema_nimg, 1e-8))
+
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000),
+    dict(
+        type='ExponentialMovingAverageHook',
+        module_keys=('generator_ema', ),
+        interp_mode='lerp',
+        interp_cfg=dict(momentum=ema_beta),
+        interval=1,
+        start_iter=0,
+        priority='VERY_HIGH')
+]
+
+inception_pkl = 'work_dirs/inception_pkl/metface_1024x1024_noflip.pkl'
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True))
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=dict(milestones=[100000], interval=[10000, 5000]),
+    metrics=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True),
+    sample_kwargs=dict(sample_model='ema'))
+
+lr_config = None
+
+total_iters = 160000

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_r_afhqv2_512_b4x8_cvt_official_rgb.py

+_base_ = ['./stylegan3_base.py']
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 65536,
+    'channel_max': 1024,
+    'magnitude_ema_beta': 0.999,
+    'conv_kernel': 1,
+    'use_radial_filters': True
+}
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        type='StyleGANv3Generator',
+        noise_size=512,
+        style_channels=512,
+        out_size=512,
+        img_channels=3,
+        rgb2bgr=True,
+        synthesis_cfg=synthesis_cfg),
+    discriminator=dict(type='StyleGAN2Discriminator', in_size=512))

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_r_ffhq_1024_b4x8_cvt_official_rgb.py

+_base_ = ['./stylegan3_base.py']
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 65536,
+    'channel_max': 1024,
+    'magnitude_ema_beta': 0.999,
+    'conv_kernel': 1,
+    'use_radial_filters': True
+}
+
+r1_gamma = 32.8
+d_reg_interval = 16
+
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=1024,
+        img_channels=3,
+        synthesis_cfg=synthesis_cfg,
+        rgb2bgr=True),
+    discriminator=dict(type='StyleGAN2Discriminator', in_size=1024))

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_r_ffhqu_256_b4x8_cvt_official_rgb.py

+_base_ = ['./stylegan3_base.py']
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 32768,
+    'channel_max': 1024,
+    'magnitude_ema_beta': 0.999,
+    'conv_kernel': 1,
+    'use_radial_filters': True
+}
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=256,
+        img_channels=3,
+        rgb2bgr=True,
+        synthesis_cfg=synthesis_cfg),
+    discriminator=dict(in_size=256, channel_multiplier=1))

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_t_ada_fp16_gamma6.6_metfaces_1024_b4x8.py

+_base_ = [
+    '../_base_/models/stylegan/stylegan3_base.py',
+    '../_base_/datasets/ffhq_flip.py', '../_base_/default_runtime.py'
+]
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 32768,
+    'channel_max': 512,
+    'magnitude_ema_beta': 0.999
+}
+r1_gamma = 6.6  # set by user
+d_reg_interval = 16
+
+load_from = 'https://download.openmmlab.com/mmgen/stylegan3/stylegan3_t_ffhq_1024_b4x8_cvt_official_rgb_20220329_235113-db6c6580.pth'  # noqa
+# ada settings
+aug_kwargs = {
+    'xflip': 1,
+    'rotate90': 1,
+    'xint': 1,
+    'scale': 1,
+    'rotate': 1,
+    'aniso': 1,
+    'xfrac': 1,
+    'brightness': 1,
+    'contrast': 1,
+    'lumaflip': 1,
+    'hue': 1,
+    'saturation': 1
+}
+
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=1024,
+        img_channels=3,
+        rgb2bgr=True,
+        synthesis_cfg=synthesis_cfg),
+    discriminator=dict(
+        type='ADAStyleGAN2Discriminator',
+        in_size=1024,
+        input_bgr2rgb=True,
+        data_aug=dict(type='ADAAug', aug_pipeline=aug_kwargs, ada_kimg=100)),
+    gan_loss=dict(type='GANLoss', gan_type='wgan-logistic-ns'),
+    disc_auxiliary_loss=dict(loss_weight=r1_gamma / 2.0 * d_reg_interval))
+
+imgs_root = 'data/metfaces/images/'
+data = dict(
+    samples_per_gpu=4,
+    train=dict(dataset=dict(imgs_root=imgs_root)),
+    val=dict(imgs_root=imgs_root))
+
+ema_half_life = 10.  # G_smoothing_kimg
+
+ema_kimg = 10
+ema_nimg = ema_kimg * 1000
+ema_beta = 0.5**(32 / max(ema_nimg, 1e-8))
+
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000),
+    dict(
+        type='ExponentialMovingAverageHook',
+        module_keys=('generator_ema', ),
+        interp_mode='lerp',
+        interp_cfg=dict(momentum=ema_beta),
+        interval=1,
+        start_iter=0,
+        priority='VERY_HIGH')
+]
+
+inception_pkl = 'work_dirs/inception_pkl/metface_1024x1024_noflip.pkl'
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True))
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=dict(milestones=[80000], interval=[10000, 5000]),
+    metrics=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True),
+    sample_kwargs=dict(sample_model='ema'))
+
+lr_config = None
+
+total_iters = 160000

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_t_afhqv2_512_b4x8_cvt_official_rgb.py

+_base_ = ['./stylegan3_base.py']
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 32768,
+    'channel_max': 512,
+    'magnitude_ema_beta': 0.999
+}
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=512,
+        img_channels=3,
+        rgb2bgr=True,
+        synthesis_cfg=synthesis_cfg),
+    discriminator=dict(in_size=512))

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_t_ffhq_1024_b4x8_cvt_official_rgb.py

+_base_ = ['./stylegan3_base.py']
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 32768,
+    'channel_max': 512,
+    'magnitude_ema_beta': 0.999
+}
+
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=1024,
+        img_channels=3,
+        synthesis_cfg=synthesis_cfg,
+        rgb2bgr=True),
+    discriminator=dict(in_size=1024))

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_t_ffhqu_256_b4x8_cvt_official_rgb.py

+_base_ = ['./stylegan3_base.py']
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 16384,
+    'channel_max': 512,
+    'magnitude_ema_beta': 0.999
+}
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(
+        out_size=256,
+        img_channels=3,
+        rgb2bgr=True,
+        synthesis_cfg=synthesis_cfg),
+    discriminator=dict(in_size=256, channel_multiplier=1))

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_t_noaug_fp16_gamma2.0_ffhq_256_b4x8.py

+_base_ = [
+    '../_base_/models/stylegan/stylegan3_base.py',
+    '../_base_/datasets/unconditional_imgs_flip_lanczos_resize_256x256.py',
+    '../_base_/default_runtime.py'
+]
+
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 16384,
+    'channel_max': 512,
+    'magnitude_ema_beta': 0.999
+}
+r1_gamma = 2.  # set by user
+d_reg_interval = 16
+
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(out_size=256, img_channels=3, synthesis_cfg=synthesis_cfg),
+    discriminator=dict(in_size=256, channel_multiplier=1),
+    gan_loss=dict(type='GANLoss', gan_type='wgan-logistic-ns'),
+    disc_auxiliary_loss=dict(loss_weight=r1_gamma / 2.0 * d_reg_interval))
+
+imgs_root = 'data/ffhq/images'
+data = dict(
+    samples_per_gpu=4,
+    train=dict(dataset=dict(imgs_root=imgs_root)),
+    val=dict(imgs_root=imgs_root))
+
+ema_half_life = 10.  # G_smoothing_kimg
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000),
+    dict(
+        type='ExponentialMovingAverageHook',
+        module_keys=('generator_ema', ),
+        interp_mode='lerp',
+        interval=1,
+        start_iter=0,
+        momentum_policy='rampup',
+        momentum_cfg=dict(
+            ema_kimg=10, ema_rampup=0.05, batch_size=32, eps=1e-8),
+        priority='VERY_HIGH')
+]
+
+inception_pkl = 'work_dirs/inception_pkl/ffhq-lanczos-256x256.pkl'
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True))
+
+inception_path = None
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN', inception_path=inception_path),
+        bgr2rgb=True),
+    sample_kwargs=dict(sample_model='ema'))
+
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=30)
+
+lr_config = None
+
+total_iters = 800002

build/lib/mmgen/.mim/configs/styleganv3/stylegan3_t_noaug_fp16_gamma32.8_ffhq_1024_b4x8.py

+_base_ = [
+    '../_base_/models/stylegan/stylegan3_base.py',
+    '../_base_/datasets/ffhq_flip.py', '../_base_/default_runtime.py'
+]
+
+batch_size = 32
+magnitude_ema_beta = 0.5**(batch_size / (20 * 1e3))
+synthesis_cfg = {
+    'type': 'SynthesisNetwork',
+    'channel_base': 32768,
+    'channel_max': 512,
+    'magnitude_ema_beta': 0.999
+}
+r1_gamma = 32.8
+d_reg_interval = 16
+
+model = dict(
+    type='StaticUnconditionalGAN',
+    generator=dict(out_size=1024, img_channels=3, synthesis_cfg=synthesis_cfg),
+    discriminator=dict(in_size=1024),
+    gan_loss=dict(type='GANLoss', gan_type='wgan-logistic-ns'),
+    disc_auxiliary_loss=dict(loss_weight=r1_gamma / 2.0 * d_reg_interval))
+
+imgs_root = None  # set by user
+
+data = dict(
+    samples_per_gpu=4,
+    train=dict(dataset=dict(imgs_root=imgs_root)),
+    val=dict(imgs_root=imgs_root))
+
+ema_half_life = 10.  # G_smoothing_kimg
+
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=5000),
+    dict(
+        type='ExponentialMovingAverageHook',
+        module_keys=('generator_ema', ),
+        interp_mode='lerp',
+        interval=1,
+        start_iter=0,
+        momentum_policy='rampup',
+        momentum_cfg=dict(
+            ema_kimg=10, ema_rampup=0.05, batch_size=batch_size, eps=1e-8),
+        priority='VERY_HIGH')
+]
+
+inception_pkl = 'work_dirs/inception_pkl/ffhq_noflip_1024x1024.pkl'
+metrics = dict(
+    fid50k=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True))
+
+evaluation = dict(
+    type='GenerativeEvalHook',
+    interval=10000,
+    metrics=dict(
+        type='FID',
+        num_images=50000,
+        inception_pkl=inception_pkl,
+        inception_args=dict(type='StyleGAN'),
+        bgr2rgb=True),
+    sample_kwargs=dict(sample_model='ema'))
+
+checkpoint_config = dict(interval=10000, by_epoch=False, max_keep_ckpts=30)
+lr_config = None
+
+total_iters = 800002

build/lib/mmgen/.mim/configs/wgan-gp/metafile.yml

+Collections:
+- Metadata:
+    Architecture:
+    - WGAN-GP
+  Name: WGAN-GP
+  Paper:
+  - https://arxiv.org/abs/1704.00028
+  README: configs/wgan-gp/README.md
+Models:
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/wgan-gp/wgangp_GN_celeba-cropped_128_b64x1_160kiter.py
+  In Collection: WGAN-GP
+  Metadata:
+    Training Data: CELEBA
+  Name: wgangp_GN_celeba-cropped_128_b64x1_160kiter
+  Results:
+  - Dataset: CELEBA
+    Metrics:
+      Details: GN
+      MS-SSIM: 0.2601
+      SWD: 5.87, 9.76, 9.43, 18.84/10.97
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/wgangp/wgangp_GN_celeba-cropped_128_b64x1_160k_20210408_170611-f8a99336.pth
+- Config: https://github.com/open-mmlab/mmgeneration/tree/master/configs/wgan-gp/wgangp_GN_GP-50_lsun-bedroom_128_b64x1_160kiter.py
+  In Collection: WGAN-GP
+  Metadata:
+    Training Data: LSUN
+  Name: wgangp_GN_GP-50_lsun-bedroom_128_b64x1_160kiter
+  Results:
+  - Dataset: LSUN
+    Metrics:
+      Details: GN, GP-lambda = 50
+      MS-SSIM: 0.059
+      SWD: 11.7, 7.87, 9.82, 25.36/13.69
+    Task: Unconditional GANs
+  Weights: https://download.openmmlab.com/mmgen/wgangp/wgangp_GN_GP-50_lsun-bedroom_128_b64x1_130k_20210408_170509-56f2a37c.pth

build/lib/mmgen/.mim/configs/wgan-gp/wgangp_GN_GP-50_lsun-bedroom_128_b64x1_160kiter.py

+_base_ = ['./wgangp_GN_celeba-cropped_128_b64x1_160kiter.py']
+
+model = dict(disc_auxiliary_loss=[
+    dict(
+        type='GradientPenaltyLoss',
+        loss_weight=50,
+        norm_mode='HWC',
+        data_info=dict(
+            discriminator='disc', real_data='real_imgs',
+            fake_data='fake_imgs'))
+])
+
+data = dict(
+    samples_per_gpu=64, train=dict(imgs_root='./data/lsun/bedroom_train'))

build/lib/mmgen/.mim/configs/wgan-gp/wgangp_GN_celeba-cropped_128_b64x1_160kiter.py

+_base_ = [
+    '../_base_/datasets/unconditional_imgs_128x128.py',
+    '../_base_/models/wgangp/wgangp_base.py'
+]
+
+data = dict(
+    samples_per_gpu=64,
+    train=dict(imgs_root='./data/celeba-cropped/cropped_images_aligned_png/'))
+
+checkpoint_config = dict(interval=10000, by_epoch=False)
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+
+custom_hooks = [
+    dict(
+        type='VisualizeUnconditionalSamples',
+        output_dir='training_samples',
+        interval=1000)
+]
+
+lr_config = None
+total_iters = 160000
+
+metrics = dict(
+    ms_ssim10k=dict(type='MS_SSIM', num_images=10000),
+    swd16k=dict(type='SWD', num_images=16384, image_shape=(3, 128, 128)))

build/lib/mmgen/.mim/model-index.yml

+Import:
+- configs/ada/metafile.yml
+- configs/biggan/metafile.yml
+- configs/cyclegan/metafile.yml
+- configs/dcgan/metafile.yml
+- configs/ggan/metafile.yml
+- configs/improved_ddpm/metafile.yml
+- configs/lsgan/metafile.yml
+- configs/pggan/metafile.yml
+- configs/pix2pix/metafile.yml
+- configs/positional_encoding_in_gans/metafile.yml
+- configs/sagan/metafile.yml
+- configs/singan/metafile.yml
+- configs/sngan_proj/metafile.yml
+- configs/styleganv1/metafile.yml
+- configs/styleganv2/metafile.yml
+- configs/styleganv3/metafile.yml
+- configs/wgan-gp/metafile.yml

build/lib/mmgen/.mim/tools/deployment/mmgen2torchserver.py

+# Copyright (c) OpenMMLab. All rights reserved.
+from argparse import ArgumentParser, Namespace
+from pathlib import Path
+from tempfile import TemporaryDirectory
+
+import mmcv
+
+try:
+    from model_archiver.model_packaging import package_model
+    from model_archiver.model_packaging_utils import ModelExportUtils
+except ImportError:
+    package_model = None
+
+
+def mmgen2torchserver(config_file: str,
+                      checkpoint_file: str,
+                      output_folder: str,
+                      model_name: str,
+                      model_version: str = '1.0',
+                      model_type: str = 'unconditional',
+                      force: bool = False):
+    """Converts MMGeneration model (config + checkpoint) to TorchServe `.mar`.
+
+    Args:
+        config_file (str): Path of config file. The config should in
+            MMGeneration format.
+        checkpoint_file (str): Path of checkpoint. The checkpoint should in
+            MMGeneration checkpoint format.
+        output_folder (str): Folder where `{model_name}.mar` will be created.
+            The file created will be in TorchServe archive format.
+        model_name (str): Name of the generated ``'mar'`` file. If not None,
+            used for naming the `{model_name}.mar` file that will be created
+            under `output_folder`. If None, `{Path(checkpoint_file).stem}`
+            will be used.
+        model_version (str, optional): Model's version. Defaults to '1.0'.
+        model_type (str, optional): Type of the model to be convert. Handler
+            named ``{model_type}_handler`` would be used to generate ``mar``
+            file. Defaults to 'unconditional'.
+        force (bool, optional): If True, existing `{model_name}.mar` will be
+            overwritten. Default to False.
+    """
+    mmcv.mkdir_or_exist(output_folder)
+
+    config = mmcv.Config.fromfile(config_file)
+
+    with TemporaryDirectory() as tmpdir:
+        config.dump(f'{tmpdir}/config.py')
+
+        args = Namespace(
+            **{
+                'model_file': f'{tmpdir}/config.py',
+                'serialized_file': checkpoint_file,
+                'handler':
+                f'{Path(__file__).parent}/mmgen_{model_type}_handler.py',
+                'model_name': model_name or Path(checkpoint_file).stem,
+                'version': model_version,
+                'export_path': output_folder,
+                'force': force,
+                'requirements_file': None,
+                'extra_files': None,
+                'runtime': 'python',
+                'archive_format': 'default'
+            })
+        manifest = ModelExportUtils.generate_manifest_json(args)
+        package_model(args, manifest)
+
+
+def parse_args():
+    parser = ArgumentParser(
+        description='Convert MMGeneration models to TorchServe `.mar` format.')
+    parser.add_argument('config', type=str, help='config file path')
+    parser.add_argument('checkpoint', type=str, help='checkpoint file path')
+    parser.add_argument(
+        '--output-folder',
+        type=str,
+        required=True,
+        help='Folder where `{model_name}.mar` will be created.')
+    parser.add_argument(
+        '--model-name',
+        type=str,
+        default=None,
+        help='If not None, used for naming the `{model_name}.mar`'
+        'file that will be created under `output_folder`.'
+        'If None, `{Path(checkpoint_file).stem}` will be used.')
+    parser.add_argument(
+        '--model-type',
+        type=str,
+        default='unconditional',
+        help='Which model type and handler to be used.')
+    parser.add_argument(
+        '--model-version',
+        type=str,
+        default='1.0',
+        help='Number used for versioning.')
+    parser.add_argument(
+        '-f',
+        '--force',
+        action='store_true',
+        help='overwrite the existing `{model_name}.mar`')
+    args = parser.parse_args()
+
+    return args
+
+
+if __name__ == '__main__':
+    args = parse_args()
+
+    if package_model is None:
+        raise ImportError('`torch-model-archiver` is required.'
+                          'Try: pip install torch-model-archiver')
+
+    mmgen2torchserver(args.config, args.checkpoint, args.output_folder,
+                      args.model_name, args.model_version, args.model_type,
+                      args.force)

build/lib/mmgen/.mim/tools/deployment/mmgen_unconditional_handler.py

+# Copyright (c) OpenMMLab. All rights reserved.
+import os
+
+import numpy as np
+import torch
+from ts.torch_handler.base_handler import BaseHandler
+
+from mmgen.apis import init_model
+
+
+class MMGenUnconditionalHandler(BaseHandler):
+
+    def initialize(self, context):
+        properties = context.system_properties
+        self.map_location = 'cuda' if torch.cuda.is_available() else 'cpu'
+        self.device = torch.device(self.map_location + ':' +
+                                   str(properties.get('gpu_id')) if torch.cuda.
+                                   is_available() else self.map_location)
+        self.manifest = context.manifest
+
+        model_dir = properties.get('model_dir')
+        serialized_file = self.manifest['model']['serializedFile']
+        checkpoint = os.path.join(model_dir, serialized_file)
+        self.config_file = os.path.join(model_dir, 'config.py')
+
+        self.model = init_model(self.config_file, checkpoint, self.device)
+        self.initialized = True
+
+    def preprocess(self, data, *args, **kwargs):
+        data_decode = dict()
+        # `data` type is `list[dict]`
+        for k, v in data[0].items():
+            # decode strings
+            if isinstance(v, bytearray):
+                data_decode[k] = v.decode()
+        return data_decode
+
+    def inference(self, data, *args, **kwargs):
+        sample_model = data['sample_model']
+        print(sample_model)
+        results = self.model.sample_from_noise(
+            None, num_batches=1, sample_model=sample_model, **kwargs)
+        return results
+
+    def postprocess(self, data):
+        # convert torch tensor to numpy and then convert to bytes
+        output_list = []
+        for data_ in data:
+            data_ = (data_ + 1) / 2
+            data_ = data_[[2, 1, 0], ...]
+            data_ = data_.clamp_(0, 1)
+            data_ = (data_ * 255).permute(1, 2, 0)
+            data_np = data_.detach().cpu().numpy().astype(np.uint8)
+            data_byte = data_np.tobytes()
+            output_list.append(data_byte)
+
+        return output_list

build/lib/mmgen/.mim/tools/deployment/test_torchserver.py

+# Copyright (c) OpenMMLab. All rights reserved.
+from argparse import ArgumentParser
+
+import numpy as np
+import requests
+from PIL import Image
+
+
+def parse_args():
+    parser = ArgumentParser()
+    parser.add_argument('model_name', help='The model name in the server')
+    parser.add_argument(
+        '--inference-addr',
+        default='127.0.0.1:8080',
+        help='Address and port of the inference server')
+    parser.add_argument(
+        '--img-path',
+        type=str,
+        default='demo.png',
+        help='Path to save generated image.')
+    parser.add_argument(
+        '--img-size', type=int, default=128, help='Size of the output image.')
+    parser.add_argument(
+        '--sample-model',
+        type=str,
+        default='ema/orig',
+        help='Which model you want to use.')
+    args = parser.parse_args()
+    return args
+
+
+def save_results(contents, img_path, img_size):
+    if not isinstance(contents, list):
+        Image.frombytes('RGB', (img_size, img_size), contents).save(img_path)
+        return
+
+    imgs = []
+    for content in contents:
+        imgs.append(
+            np.array(Image.frombytes('RGB', (img_size, img_size), content)))
+    Image.fromarray(np.concatenate(imgs, axis=1)).save(img_path)
+
+
+def main(args):
+    url = 'http://' + args.inference_addr + '/predictions/' + args.model_name
+
+    if args.sample_model == 'ema/orig':
+        cont_ema = requests.post(url, {'sample_model': 'ema'}).content
+        cont_orig = requests.post(url, {'sample_model': 'orig'}).content
+        save_results([cont_ema, cont_orig], args.img_path, args.img_size)
+        return
+
+    response = requests.post(url, {'sample_model': args.sample_model})
+    save_results(response.content, args.img_path, args.img_size)
+
+
+if __name__ == '__main__':
+    args = parse_args()
+    main(args)

build/lib/mmgen/.mim/tools/dist_eval.sh

+#!/usr/bin/env bash
+
+CONFIG=$1
+CHECKPOINT=$2
+GPUS=$3
+NNODES=${NNODES:-1}
+NODE_RANK=${NODE_RANK:-0}
+PORT=${PORT:-29500}
+MASTER_ADDR=${MASTER_ADDR:-"127.0.0.1"}
+
+PYTHONPATH="$(dirname $0)/..":$PYTHONPATH \
+python -m torch.distributed.launch \
+    --nnodes=$NNODES \
+    --node_rank=$NODE_RANK \
+    --master_addr=$MASTER_ADDR \
+    --nproc_per_node=$GPUS \
+    --master_port=$PORT \
+    $(dirname "$0")/test.py \
+    $CONFIG \
+    $CHECKPOINT \
+    --launcher pytorch \
+    ${@:4}