{ "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/conv300m_97k_wavegrid_gentle", "resume_checkpoint": "/workspace/runs/hybrid_specz_300m_oodprobe_joint/20260518_035753/best.pt", "allow_mismatched_checkpoint": false, "refresh_data": false, "epochs": 3, "batch_size": 8, "num_workers": 4, "target_length": 8192, "architecture": "conv", "d_model": 1024, "conv_width": 512, "layers": 20, "heads": 16, "dropout": 0.1, "z_bins": 64, "stem_stride": 8, "token_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": 2e-06, "min_lr": 5e-07, "warmup_steps": 50, "weight_decay": 0.03, "grad_clip": 1.0, "grad_accum_steps": 1, "eval_every": 250, "eval_max_val": 1200, "eval_max_ood": 700, "max_steps": 1500, "checkpoint_score": "combined", "score_z_alpha": 0.7, "objective": "joint", "balance_redshift": false, "train_clean_only": false, "clean_sample_boost": 3.0, "augment_ood": true, "eval_ood": true, "random_mask_ratio": 0.3, "eval_mask_ratio": 0.3, "mask_mode": "mixed_span", "mask_span_min": 16, "mask_span_max": 96, "line_region_percentile": 90.0, "crop_prob": 0.25, "bad_window_prob": 0.18, "throughput_prob": 0.35, "noise_prob": 0.2, "resolution_prob": 0.25, "downsample_prob": 0.25, "line_dropout_prob": 0.12, "span_dropout_prob": 0.18, "grid_jitter_prob": 0.35, "grid_shift_frac": 0.015, "grid_scale_frac": 0.04, "redshift_shift": 0.0, "rec_weight": 0.4, "z_weight": 1.0, "z_bin_weight": 0.35, "z_candidate_weight": 0.1, "z_nll_weight": 0.02, "zwarn_weight": 0.3, "high_z_boost": 2.0, "high_z_threshold": 1.0, "clean_z_only": false, "seed": 17 }