| { |
| "dataset_name": "MultimodalUniverse/desi", |
| "max_samples": 97332, |
| "cache_dir": "/workspace/native_specz_mae/cache", |
| "hf_cache_dir": "/workspace/hf_cache/datasets", |
| "run_dir": "/workspace/runs/hybrid_specz_75m_97k_safe_native", |
| "resume_checkpoint": "/workspace/runs/hybrid_specz_75m_oodprobe_long_lowlr/20260518_030350/last.pt", |
| "allow_mismatched_checkpoint": false, |
| "refresh_data": false, |
| "epochs": 5, |
| "batch_size": 16, |
| "num_workers": 8, |
| "target_length": 8192, |
| "d_model": 640, |
| "conv_width": 320, |
| "layers": 12, |
| "heads": 10, |
| "dropout": 0.1, |
| "z_bins": 64, |
| "stem_stride": 8, |
| "rec_hidden_mult": 0, |
| "rec_refine_width": 16, |
| "rec_refine_kernel": 5, |
| "layerscale_init": 0.0, |
| "prediction_mode": "bin_residual", |
| "bin_temperature": 1.0, |
| "residual_scale": 0.06, |
| "candidate_topk": 5, |
| "lr": 1e-05, |
| "min_lr": 7e-07, |
| "warmup_steps": 200, |
| "weight_decay": 0.03, |
| "grad_clip": 1.0, |
| "grad_accum_steps": 1, |
| "eval_every": 1000, |
| "eval_max_val": 1000, |
| "eval_max_ood": 600, |
| "max_steps": 0, |
| "checkpoint_score": "combined", |
| "score_z_alpha": 0.7, |
| "objective": "joint", |
| "balance_redshift": false, |
| "train_clean_only": false, |
| "clean_sample_boost": 1.0, |
| "augment_ood": true, |
| "eval_ood": true, |
| "random_mask_ratio": 0.15, |
| "eval_mask_ratio": 0.25, |
| "mask_mode": "pixel", |
| "mask_span_min": 16, |
| "mask_span_max": 64, |
| "line_region_percentile": 90.0, |
| "crop_prob": 0.35, |
| "bad_window_prob": 0.25, |
| "throughput_prob": 0.45, |
| "noise_prob": 0.25, |
| "resolution_prob": 0.2, |
| "downsample_prob": 0.12, |
| "line_dropout_prob": 0.15, |
| "span_dropout_prob": 0.15, |
| "redshift_shift": 0.0, |
| "rec_weight": 0.5, |
| "z_weight": 1.0, |
| "z_bin_weight": 0.45, |
| "z_candidate_weight": 0.1, |
| "z_nll_weight": 0.03, |
| "zwarn_weight": 0.3, |
| "high_z_boost": 1.5, |
| "high_z_threshold": 1.0, |
| "clean_z_only": false, |
| "seed": 17 |
| } |