MCPcopy Create free account
hub / github.com/00why00/JoDiffusion / parse_args

Function parse_args

train_ae.py:78–261  ·  view source on GitHub ↗
()

Source from the content-addressed store, hash-verified

76
77
78def parse_args():
79 parser = argparse.ArgumentParser(description="training script of dense label vae.")
80 parser.add_argument(
81 "--dataset_name",
82 type=str,
83 default="ade20k_semantic",
84 help="The name of the dataset to use for training.",
85 )
86 parser.add_argument(
87 "--label_column",
88 type=str,
89 default="semantic",
90 help="The column name of the label in the dataset.",
91 )
92 parser.add_argument(
93 "--lightweight_label_vae",
94 action="store_true",
95 help="Whether or not to use lightweight vae.",
96 )
97 parser.add_argument(
98 "--loss_type",
99 type=str,
100 default="ce",
101 choices=['ce', 'seg'],
102 help="The type of loss to use for training.",
103 )
104 parser.add_argument(
105 "--output_dir",
106 type=str,
107 default=None,
108 help="The output directory where the model predictions and checkpoints will be written.",
109 )
110 parser.add_argument(
111 "--seed",
112 type=int,
113 default=42,
114 help="A seed for reproducible training."
115 )
116 parser.add_argument(
117 "--resolution",
118 type=int,
119 default=512,
120 help="The resolution for input images, all the images in the train/validation dataset will be resized to this"
121 " resolution",
122 )
123 parser.add_argument(
124 "--train_batch_size",
125 type=int,
126 default=16,
127 help="Batch size (per device) for the training dataloader."
128 )
129 parser.add_argument(
130 "--num_train_epochs",
131 type=int,
132 default=100
133 )
134 parser.add_argument(
135 "--max_train_steps",

Callers 1

mainFunction · 0.70

Calls

no outgoing calls

Tested by

no test coverage detected