Team Ai
Apppublic

radames/Text2Human-API

sourceHugging Faceupdated 3y agoView on Hugging Face
1likes
sample_from_parsing.yml94 linesDownload Raw Back to configs
1name: sample_from_parsing2use_tb_logger: true3set_CUDA_VISIBLE_DEVICES: ~4gpu_ids: [3]5 6# dataset configs7batch_size: 48num_workers: 49test_img_dir: ./datasets/test_images10segm_dir: ./datasets/segm11pose_dir: ./datasets/densepose12test_ann_file: ./datasets/texture_ann/test13downsample_factor: 214 15model_type: SampleFromParsingModel16# network configs17embed_dim: 25618n_embed: 102419codebook_spatial_size: 220 21# bottom level vqvae22bot_n_embed: 51223bot_codebook_spatial_size: 224bot_double_z: false25bot_z_channels: 25626bot_resolution: 51227bot_in_channels: 328bot_out_ch: 329bot_ch: 12830bot_ch_mult: [1, 1, 2, 4]31bot_num_res_blocks: 232bot_attn_resolutions: [64]33bot_dropout: 0.034bot_vae_path: ./pretrained_models/vqvae_bottom.pth35 36# top level vqgan37top_double_z: false38top_z_channels: 25639top_resolution: 51240top_in_channels: 341top_out_ch: 342top_ch: 12843top_ch_mult: [1, 1, 2, 2, 4]44top_num_res_blocks: 245top_attn_resolutions: [32]46top_dropout: 0.047top_vae_path: ./pretrained_models/vqvae_top.pth48 49# unet configs50index_pred_encoder_in_channels: 25651index_pred_fc_in_channels: 6452index_pred_fc_in_index: 453index_pred_fc_channels: 6454index_pred_fc_num_convs: 155index_pred_fc_concat_input: False56index_pred_fc_dropout_ratio: 0.157index_pred_fc_num_classes: 51258index_pred_fc_align_corners: False59pretrained_index_network: ./pretrained_models/index_pred_net.pth60 61# segmentation tokenization62segm_double_z: false63segm_z_channels: 3264segm_resolution: 51265segm_in_channels: 2466segm_out_ch: 2467segm_ch: 6468segm_ch_mult: [1, 1, 2, 2, 4]69segm_num_res_blocks: 170segm_attn_resolutions: [16]71segm_dropout: 0.072segm_num_segm_classes: 2473segm_n_embed: 102474segm_embed_dim: 3275segm_token_path: ./pretrained_models/parsing_token.pth76 77# sampler configs78codebook_size: 1843279segm_codebook_size: 102480texture_codebook_size: 1881bert_n_emb: 51282bert_n_layers: 2483bert_n_head: 884block_size: 512 # 32 x 1685latent_shape: [32, 16]86embd_pdrop: 0.087resid_pdrop: 0.088attn_pdrop: 0.089num_head: 1890pretrained_sampler: ./pretrained_models/sampler.pth91 92manual_seed: 202193sample_steps: 25694