| { |
| "geometry_checkpoint_dir": "weights/geometry_transformer.pt", |
| "matvae_checkpoint_dir": "weights/matvae.safetensors", |
| "normalization_params_path": "weights/normalization_params.json", |
| "matvae_checkpoint": "weights/matvae.safetensors", |
| "trellis_weights_path": "weights/TRELLIS-image-large", |
| "models": { |
| "geometry_encoder": { |
| "name": "geometry_encoder", |
| "args": { |
| "resolution": 64, |
| "in_channels": 1024, |
| "model_channels": 768, |
| "latent_channels": 2, |
| "num_blocks": 12, |
| "num_heads": 12, |
| "mlp_ratio": 4, |
| "attn_mode": "swin", |
| "window_size": 8, |
| "use_fp16": true |
| } |
| }, |
| "matvae": { |
| "name": "matvae", |
| "args": { |
| "width": 256, |
| "depth": 3, |
| "z_dim": 2, |
| "p_drop": 0.05, |
| "use_learned_variances": false, |
| "use_additional_losses": true |
| } |
| } |
| }, |
| "dataset": { |
| "name": "SparseVoxelMaterials", |
| "normalization_type": "log_minmax", |
| "args": { |
| "roots": "datasets/simready", |
| "image_size": 512, |
| "model": "dinov2_vitl14_reg", |
| "resolution": 64, |
| "min_aesthetic_score": 0.0, |
| "max_num_voxels": 32768, |
| "compute_material_stats": false |
| } |
| }, |
| "trainer": { |
| "name": "SLatVaeMaterialsTrainer", |
| "args": { |
| "max_steps": 1000000, |
| "batch_size_per_gpu": 16, |
| "batch_split": 1, |
| "optimizer": { |
| "name": "AdamW", |
| "args": { |
| "lr": 1e-4, |
| "weight_decay": 0.0 |
| } |
| }, |
| "ema_rate": [ |
| 0.9999 |
| ], |
| "fp16_mode": "inflat_all", |
| "fp16_scale_growth": 0.001, |
| "elastic": { |
| "name": "LinearMemoryController", |
| "args": { |
| "target_ratio": 0.75, |
| "max_mem_ratio_start": 0.5 |
| } |
| }, |
| "grad_clip": { |
| "name": "AdaptiveGradClipper", |
| "args": { |
| "max_norm": 1.0, |
| "clip_percentile": 95 |
| } |
| }, |
| "i_log": 10, |
| "i_save": 2000, |
| "i_eval": 1000, |
| "loss_type": "l1" |
| } |
| } |
| } |
