radames/Text2Human-API

1name: parsing_tokenization2use_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: VQSegmentationModel19# network configs20embed_dim: 3221n_embed: 102422image_key: "segmentation"23n_labels: 2424double_z: false25z_channels: 3226resolution: 51227in_channels: 2428out_ch: 2429ch: 6430ch_mult: [1, 1, 2, 2, 4]31num_res_blocks: 132attn_resolutions: [16]33dropout: 0.034 35num_segm_classes: 2436 37 38# training configs39val_freq: 540print_freq: 10041weight_decay: 042manual_seed: 202143num_epochs: 10044lr: !!float 4.5e-0545lr_decay: step46gamma: 0.147step: 5048