{ "training": { "patch_sizes": [ 1, 2, 3, 4, 5, 6, 7, 8 ], "max_lr": 0.003, "num_epochs": 500, "batch_size": 128, "effective_batch_size": 512, "warmup_epochs": 30, "final_lr": 1e-06, "weight_decay": 0.01, "grad_clip": true, "betas": [ 0.9, 0.999 ], "ema": [ 0.996, 1.0 ], "shape_time_combinations": [ { "size": 4, "timesteps": 12 }, { "size": 5, "timesteps": 6 }, { "size": 6, "timesteps": 4 }, { "size": 7, "timesteps": 3 }, { "size": 9, "timesteps": 3 }, { "size": 12, "timesteps": 3 } ], "masking_probabilities": [ 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0 ], "augmentation": { "flip+rotate": true }, "encode_ratio": 0.1, "decode_ratio": 0.5, "max_unmasking_channels": 17, "loss_type": "patch_disc", "tau": 0.1, "pred2unit": false, "loss_mask_other_samples": true, "normalization": "std", "eval_eurosat_every_n_epochs": 10, "random_masking": "half", "unmasking_channels_combo": "all", "double_loss": true, "double_predictors": true, "target_exit_after": 0, "token_exit_cfg": { "S1": 12, "S2_RGB": 12, "S2_Red_Edge": 12, "S2_NIR_10m": 12, "S2_NIR_20m": 12, "S2_SWIR": 12, "NDVI": 6, "ERA5": 6, "TC": 6, "VIIRS": 12, "SRTM": 6, "DW": 0, "WC": 0, "LS": 0, "location": 12, "DW_static": 0, "WC_static": 0 }, "target_masking": "all" }, "model": { "encoder": { "embedding_size": 192, "depth": 12, "num_heads": 3, "mlp_ratio": 4, "max_sequence_length": 24, "freeze_projections": false, "drop_path": 0.1, "max_patch_size": 8 }, "decoder": { "depth": 4, "num_heads": 3, "mlp_ratio": 4, "max_sequence_length": 24, "learnable_channel_embeddings": true, "max_patch_size": 8, "embedding_size": 192 } } }