radames/Text2Human-API

1name: vqvae_bottom2use_tb_logger: true3set_CUDA_VISIBLE_DEVICES: ~4gpu_ids: [3]5 6# dataset configs7batch_size: 48num_workers: 49train_img_dir: ./datasets/train_images10test_img_dir: ./datasets/test_images11segm_dir: ./datasets/segm12pose_dir: ./datasets/densepose13train_ann_file: ./datasets/texture_ann/train14val_ann_file: ./datasets/texture_ann/val15test_ann_file: ./datasets/texture_ann/test16downsample_factor: 217 18model_type: HierarchyVQSpatialTextureAwareModel19# network configs20embed_dim: 25621n_embed: 102422codebook_spatial_size: 223 24# bottom level vqvae25bot_n_embed: 51226bot_double_z: false27bot_z_channels: 25628bot_resolution: 51229bot_in_channels: 330bot_out_ch: 331bot_ch: 12832bot_ch_mult: [1, 1, 2, 4]33bot_num_res_blocks: 234bot_attn_resolutions: [64]35bot_dropout: 0.036 37# top level vqgan38top_double_z: false39top_z_channels: 25640top_resolution: 51241top_in_channels: 342top_out_ch: 343top_ch: 12844top_ch_mult: [1, 1, 2, 2, 4]45top_num_res_blocks: 246top_attn_resolutions: [32]47top_dropout: 0.048top_vae_path: ./pretrained_models/vqvae_top.pth49 50fix_decoder: false51 52disc_layers: 353disc_weight_max: 154disc_start_step: 155n_channels: 356ndf: 6457nf: 12858perceptual_weight: 1.059 60num_segm_classes: 2461 62# training configs63val_freq: 564print_freq: 10065weight_decay: 066manual_seed: 202167num_epochs: 100068lr: !!float 1.0e-0469lr_decay: step70gamma: 1.071step: 5072 73