|
{ |
|
"_attn_implementation_autoset": true, |
|
"auto_map": { |
|
"AutoModel": "huggingface_mae.MAEModel", |
|
"AutoConfig": "huggingface_mae.MAEConfig" |
|
}, |
|
"apply_loss_unmasked": false, |
|
"architectures": [ |
|
"MAEModel" |
|
], |
|
"crop_size": -1, |
|
"decoder": { |
|
"_target_": "mae_modules.CAMAEDecoder", |
|
"depth": 8, |
|
"embed_dim": 512, |
|
"mlp_ratio": 4, |
|
"norm_layer": { |
|
"_partial_": true, |
|
"_target_": "torch.nn.LayerNorm", |
|
"eps": 1e-06 |
|
}, |
|
"num_heads": 16, |
|
"num_modalities": 6, |
|
"qkv_bias": true, |
|
"tokens_per_modality": 256 |
|
}, |
|
"encoder": { |
|
"_target_": "mae_modules.MAEEncoder", |
|
"channel_agnostic": true, |
|
"max_in_chans": 11, |
|
"vit_backbone": { |
|
"_target_": "vit.sincos_positional_encoding_vit", |
|
"vit_backbone": { |
|
"_target_": "vit.vit_small_patch16_256", |
|
"global_pool": "avg" |
|
} |
|
} |
|
}, |
|
"fourier_loss": { |
|
"_target_": "loss.FourierLoss", |
|
"num_multimodal_modalities": 6 |
|
}, |
|
"fourier_loss_weight": 0.0, |
|
"input_norm": { |
|
"_args_": [ |
|
{ |
|
"_target_": "normalizer.Normalizer" |
|
}, |
|
{ |
|
"_target_": "torch.nn.InstanceNorm2d", |
|
"affine": false, |
|
"num_features": null, |
|
"track_running_stats": false |
|
} |
|
], |
|
"_target_": "torch.nn.Sequential" |
|
}, |
|
"layernorm_unfreeze": true, |
|
"loss": { |
|
"_target_": "torch.nn.MSELoss", |
|
"reduction": "none" |
|
}, |
|
"lr_scheduler": { |
|
"_partial_": true, |
|
"_target_": "torch.optim.lr_scheduler.OneCycleLR", |
|
"anneal_strategy": "cos", |
|
"max_lr": 0.0001, |
|
"pct_start": 0.1 |
|
}, |
|
"mask_fourier_loss": true, |
|
"mask_ratio": 0.0, |
|
"model_type": "MAE", |
|
"norm_pix_loss": false, |
|
"num_blocks_to_freeze": 0, |
|
"optimizer": { |
|
"_partial_": true, |
|
"_target_": "timm.optim.lion.Lion", |
|
"betas": [ |
|
0.9, |
|
0.95 |
|
], |
|
"lr": 0.0001, |
|
"weight_decay": 0.05 |
|
}, |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.46.1", |
|
"trim_encoder_blocks": null, |
|
"use_MAE_weight_init": false |
|
} |
|
|