| { |
| "models": { |
| "encoder": { |
| "name": "AniGenSparseStructureEncoder", |
| "args": { |
| "in_channels": 1, |
| "in_channels_skl": 1, |
| "latent_channels": 8, |
| "latent_channels_skl": 4, |
| "num_res_blocks": 2, |
| "num_res_blocks_middle": 2, |
| "channels": [ |
| 32, |
| 128, |
| 512 |
| ], |
| "use_fp16": true, |
| "encode_global": false, |
| "encode_global_skl": false, |
| "use_pretrain_branch": true, |
| "freeze_pretrain_branch": true, |
| "modules_to_freeze": [ |
| "blocks", |
| "input_layer", |
| "middle_block", |
| "out_layer" |
| ], |
| "latent_denoising": false, |
| "latent_denoising_skl": true, |
| "normalize_z": false, |
| "normalize_z_skl": true, |
| "normalize_scale": 1.0 |
| } |
| }, |
| "decoder": { |
| "name": "AniGenSparseStructureDecoder", |
| "args": { |
| "out_channels": 1, |
| "out_channels_skl": 1, |
| "latent_channels": 8, |
| "latent_channels_skl": 4, |
| "num_res_blocks": 2, |
| "num_res_blocks_middle": 2, |
| "channels": [ |
| 512, |
| 128, |
| 32 |
| ], |
| "use_fp16": true, |
| "encode_global": false, |
| "encode_global_skl": false, |
| "use_pretrain_branch": true, |
| "freeze_pretrain_branch": true, |
| "modules_to_freeze": [ |
| "blocks", |
| "input_layer", |
| "middle_block", |
| "out_layer" |
| ], |
| "normalize_z": false, |
| "normalize_z_skl": false, |
| "normalize_scale": 1.0 |
| } |
| } |
| }, |
| "dataset": { |
| "name": "AniGenSparseStructure", |
| "args": { |
| "resolution": 64, |
| "min_aesthetic_score": 4.5, |
| "skl_dilation_iter": 2 |
| } |
| }, |
| "trainer": { |
| "name": "AniGenSparseStructureVaeTrainer", |
| "args": { |
| "latent_denoising": false, |
| "latent_denoising_gamma": 1.0, |
| "latent_denoising_skl": true, |
| "latent_denoising_gamma_skl": 1.0, |
| "latent_time_max": 0.75, |
| "latent_time_max_skl": 0.25, |
| "latent_achive_max_step": 10000, |
| "latent_achive_max_step_skl": 10000, |
| "max_steps": 1000000, |
| "batch_size_per_gpu": 32, |
| "batch_split": 1, |
| "optimizer": { |
| "name": "AdamW", |
| "args": { |
| "lr": 0.0001, |
| "weight_decay": 0.0 |
| } |
| }, |
| "log_param_stats": false, |
| "lr_scheduler": { |
| "name": "SequentialLR", |
| "schedulers": [ |
| { |
| "name": "LinearLR", |
| "args": { |
| "start_factor": 1e-06, |
| "end_factor": 1.0, |
| "total_iters": 1000 |
| } |
| }, |
| { |
| "name": "CosineAnnealingLR", |
| "args": { |
| "T_max": 1000000, |
| "eta_min": 0.0 |
| } |
| } |
| ], |
| "args": { |
| "milestones": [ |
| 1000 |
| ] |
| } |
| }, |
| "ema_rate": [ |
| 0.9999 |
| ], |
| "fp16_mode": "inflat_all", |
| "fp16_scale_growth": 0.001, |
| "grad_clip": { |
| "name": "AdaptiveGradClipper", |
| "args": { |
| "max_norm": 1.0, |
| "clip_percentile": 95 |
| } |
| }, |
| "i_log": 500, |
| "i_sample": 5000, |
| "i_save": 5000, |
| "loss_type": "dice", |
| "lambda_kl": 0.001, |
| "lambda_kl_skl": 0.001 |
| } |
| } |
| } |
