Text2Human/configs/vqvae_top.yml

name: vqvae_top
use_tb_logger: true
set_CUDA_VISIBLE_DEVICES: ~
gpu_ids: [3]

# dataset configs
batch_size: 4
num_workers: 4
train_img_dir: ./datasets/train_images
test_img_dir: ./datasets/test_images
segm_dir: ./datasets/segm
pose_dir: ./datasets/densepose
train_ann_file: ./datasets/texture_ann/train
val_ann_file: ./datasets/texture_ann/val
test_ann_file: ./datasets/texture_ann/test
downsample_factor: 2

model_type: VQImageSegmTextureModel
# network configs
embed_dim: 256
n_embed: 1024
double_z: false
z_channels: 256
resolution: 512
in_channels: 3
out_ch: 3
ch: 128
ch_mult: [1, 1, 2, 2, 4]
num_res_blocks: 2
attn_resolutions: [32]
dropout: 0.0

disc_layers: 3
disc_weight_max: 0
disc_start_step: 3000000000000000000000000001
n_channels: 3
ndf: 64
nf: 128
perceptual_weight: 1.0

num_segm_classes: 24


# training configs
val_freq: 5
print_freq: 100
weight_decay: 0
manual_seed: 2021
num_epochs: 1000
lr: !!float 1.0e-04
lr_decay: step
gamma: 1.0
step: 50