| { | |
| "model": { | |
| "type": "PoLAr-MAE", | |
| "arch": "vit_small", | |
| "num_channels": 4, | |
| "voxel_size": 5, | |
| "masking_ratio": 0.6, | |
| "masking_type": "rand", | |
| "transformer_kwargs": { | |
| "postnorm": false, | |
| "add_pos_at_every_layer": true, | |
| "drop_rate": 0.0, | |
| "attn_drop_rate": 0.05, | |
| "drop_path_rate": 0.25 | |
| }, | |
| "decoder_arch": "vit_small", | |
| "decoder_kwargs": { | |
| "transformer_kwargs": { | |
| "depth": 4, | |
| "postnorm": true, | |
| "add_pos_at_every_layer": true, | |
| "drop_rate": 0.0, | |
| "attn_drop_rate": 0.05, | |
| "drop_path_rate": 0.25 | |
| } | |
| }, | |
| "decoder_use_kv": false, | |
| "mae_prediction": "full", | |
| "loss_weights": { | |
| "chamfer": 1.0, | |
| "energy": 1.0 | |
| }, | |
| "tokenizer_kwargs": { | |
| "group_max_points": 64 | |
| }, | |
| "patch_encoder_kwargs": { | |
| "feature_dim": 384, | |
| "hidden1": 128, | |
| "hidden2": 256, | |
| "pos_enc_style": "original" | |
| } | |
| } | |
| } | |