def update_cfg(self, cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
name: 'DenseDiscriminator_v1'
ch: 512
init_type: 'orth'
cfg_downsample:
name: "AvgPool2d"
num_cells: 3
cfg_cell:
name: "DenseBlock"
n_nodes: 4
cfg_mix_layer:
name: "MixedLayer"
cfg_ops:
None:
name: "D2None"
Identity:
name: "Identity"
Conv2d_3x3:
name: "Conv2dAct"
cfg_conv:
name: "SNConv2d"
kernel_size: 3
padding: 1
cfg_act:
name: "ReLU"
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
def update_cfg(self, cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
name: "DenseCell"
n_nodes: 3
cfg_mix_layer:
name: "MixedLayer"
cfg_ops:
Identity:
name: "Identity"
Conv2d_3x3:
name: "ActConv2d"
cfg_act:
name: "ReLU"
cfg_conv:
name: "Conv2d"
kernel_size: 3
padding: 1
None:
name: "D2None"
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
def update_cfg(self, cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
name: 'BigGANDisc'
img_size: "kwargs['img_size']"
n_classes: "kwargs['n_classes']"
ch: 8
use_cdisc: true
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
def update_cfg(self, cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
name: "StyleLayer"
z_dim: 128
n_mlp: 1
num_features: 256
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
def update_cfg(cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
GAN_metric:
name: TFFIDISScore
tf_fid_stat: "datasets/tf_fid_stat_{dataset_name}_{img_size}.npz"
tf_inception_model_dir: "datasets/tf_inception_model"
num_inception_images: 50000
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
def update_nni_config_file(nni_config_file, update_nni_cfg_str):
update_nni_cfg = yaml.safe_load(update_nni_cfg_str)
# os.makedirs(update_nni_cfg['logDir'], exist_ok=True)
with open(nni_config_file, 'r') as f:
nni_cfg = yaml.safe_load(f)
nni_cfg = update_config(nni_cfg, update_nni_cfg)
nni_cfg = convert_easydict_to_dict(nni_cfg)
logging.getLogger('tl').info('\nnni config:\n ' + get_dict_str(nni_cfg))
updated_config_file = nni_config_file.split('.')[-2] + '_updated.' + nni_config_file.split('.')[-1]
with open(updated_config_file, 'w') as f:
yaml.dump(nni_cfg, f, indent=2, sort_keys=False)
return updated_config_file
def update_cfg(self, cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
name: "StyleV2Conv"
cfg_modconv:
name: "ModulatedConv2d"
kernel_size: 3
style_dim: 192
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
def update_cfg(self, cfg):
if not getattr(cfg, 'update_cfg', False):
return cfg
cfg_str = """
name: ""
n_nodes: 4
cfg_mix_layer:
name: "MixedLayerWithArc"
cfg_ops:
Conv2d_1x1:
name: "Conv2d"
kernel_size: 1
padding: 0
Conv2d_3x3:
name: "Conv2d"
kernel_size: 3
padding: 1
"""
default_cfg = EasyDict(yaml.safe_load(cfg_str))
cfg = update_config(default_cfg, cfg)
return cfg
