recursionpharma/OpenPhenom
221.2k
1{2 "_attn_implementation_autoset": true,3 "auto_map": {4 "AutoModel": "huggingface_mae.MAEModel",5 "AutoConfig": "huggingface_mae.MAEConfig"6 },7 "apply_loss_unmasked": false,8 "architectures": [9 "MAEModel"10 ],11 "crop_size": -1,12 "decoder": {13 "_target_": "mae_modules.CAMAEDecoder",14 "depth": 8,15 "embed_dim": 512,16 "mlp_ratio": 4,17 "norm_layer": {18 "_partial_": true,19 "_target_": "torch.nn.LayerNorm",20 "eps": 1e-0621 },22 "num_heads": 16,23 "num_modalities": 6,24 "qkv_bias": true,25 "tokens_per_modality": 25626 },27 "encoder": {28 "_target_": "mae_modules.MAEEncoder",29 "channel_agnostic": true,30 "max_in_chans": 11,31 "vit_backbone": {32 "_target_": "vit.sincos_positional_encoding_vit",33 "vit_backbone": {34 "_target_": "vit.vit_small_patch16_256",35 "global_pool": "avg"36 }37 }38 },39 "fourier_loss": {40 "_target_": "loss.FourierLoss",41 "num_multimodal_modalities": 642 },43 "fourier_loss_weight": 0.0,44 "input_norm": {45 "_args_": [46 {47 "_target_": "normalizer.Normalizer"48 },49 {50 "_target_": "torch.nn.InstanceNorm2d",51 "affine": false,52 "num_features": null,53 "track_running_stats": false54 }55 ],56 "_target_": "torch.nn.Sequential"57 },58 "layernorm_unfreeze": true,59 "loss": {60 "_target_": "torch.nn.MSELoss",61 "reduction": "none"62 },63 "lr_scheduler": {64 "_partial_": true,65 "_target_": "torch.optim.lr_scheduler.OneCycleLR",66 "anneal_strategy": "cos",67 "max_lr": 0.0001,68 "pct_start": 0.169 },70 "mask_fourier_loss": true,71 "mask_ratio": 0.0,72 "model_type": "MAE",73 "norm_pix_loss": false,74 "num_blocks_to_freeze": 0,75 "optimizer": {76 "_partial_": true,77 "_target_": "timm.optim.lion.Lion",78 "betas": [79 0.9,80 0.9581 ],82 "lr": 0.0001,83 "weight_decay": 0.0584 },85 "torch_dtype": "float32",86 "transformers_version": "4.46.1",87 "trim_encoder_blocks": null,88 "use_MAE_weight_init": false89}90 