macs_v2 runs from pc2 2026-10-24_22:51

macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt

@@ -0,0 +1,108 @@
+[05:16:39][      0s][START] target=dhanyasi-P14-mouse-cerebellum variant=macs-v11-norestart-acq-laplace budget=5.0% seed=2 | {"task": "membrane", "target": "dhanyasi-P14-mouse-cerebellum", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "laplace", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[05:16:39][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[05:16:39][      0s][DATA] dhanyasi-P14-mouse-cerebellum: train=8649 val=1922 test=2883 patches
+[05:16:41][      3s][CALIB] source-val temperatures: c-elegans=2.58, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, snemi=3.55, mouse-brain=4.08
+[05:16:42][      4s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 1s
+[05:16:48][      9s][WEIGHTS] mixture: c-elegans=0.051, flywire=0.156, berghia=0.010, microns=0.417, oct-glia=0.040, oct-sfl=0.054, snemi=0.107, mouse-brain=0.165 | eff.sources=5.34 | MMD2 mix=0.0662 uniform=0.1039
+[05:16:48][      9s][WEIGHTS] per-source: c-elegans=0.183, flywire=0.001, berghia=0.071, microns=0.290, oct-glia=0.195, oct-sfl=0.144, snemi=0.112, mouse-brain=0.004 | min-MMD source (NeuroADDA rule) = microns
+[05:17:12][     34s][TEACHERS] target-val Dice: c-elegans=0.413, flywire=0.107, berghia=0.158, microns=0.223, oct-glia=0.110, oct-sfl=0.291, snemi=0.297, mouse-brain=0.448
+[05:17:12][     34s][TEACHERS] ensemble target-val Dice: mixture=0.210, mixture@raw=0.220, mixture@logit=0.189, per_source=0.193, per_source@raw=0.209, per_source@logit=0.156, uniform=0.192, uniform@raw=0.210, uniform@logit=0.154
+[05:17:12][     34s][WEIGHTS] using mixture: c-elegans=0.051, flywire=0.156, berghia=0.010, microns=0.417, oct-glia=0.040, oct-sfl=0.054, snemi=0.107, mouse-brain=0.165
+[05:17:12][     34s][INIT] student initialised from teacher microns (weight 0.417)
+[05:17:15][     36s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[05:17:15][     36s][KD] distilling the mixture ensemble (prob space) into the student
+[05:18:23][    105s][KD] soft labels for 6000 patches from 8 teachers in 69s
+[05:19:20][    161s][KD] epoch 1/5 | patches=6000 | loss=0.5836 | 56s
+[05:20:15][    217s][KD] epoch 2/5 | patches=6000 | loss=0.4512 | 56s
+[05:21:10][    272s][KD] epoch 3/5 | patches=6000 | loss=0.4330 | 55s
+[05:22:06][    327s][KD] epoch 4/5 | patches=6000 | loss=0.4247 | 55s
+[05:23:01][    382s][KD] epoch 5/5 | patches=6000 | loss=0.4224 | 55s
+[05:23:03][    384s][ROUND0] before any target label: val_dice=0.1795 val_vi=0.5071
+[05:23:03][    384s][AL] patches=8649 (pool 8647 after dropping blank; free blank negatives: True) | budget=432 (5.0%) | rounds=6 | per_round=72 | acquisition=laplace | first_round=stratified
+[05:23:29][    410s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': -0.3018138618955013, 'auroc_top20_error': 0.24228212890625}
+[05:23:52][    434s][AL] round 1: stratified cold start: 36 patches from the membrane-rich half (ensemble prevalence >= 0.184) + 36 from the membrane-poor half of 2000 candidates
+[05:24:00][    441s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.449, uniform=0.257, top1=0.413, top2=0.428, top3=0.558 -> using top3
+[05:24:02][    443s][WEIGHTS] round 1: student kept (labeled Dice 0.228 vs best teacher mouse-brain 0.413); distilling the corrected ensemble into it
+[05:24:02][    443s][WEIGHTS] round 1: teacher Dice on 72 labeled patches + 2 blank (891480 fg px): c-elegans=0.347, flywire=0.154, berghia=0.171, microns=0.295, oct-glia=0.191, oct-sfl=0.403, snemi=0.403, mouse-brain=0.413 | weights: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.157, snemi=0.312, mouse-brain=0.531
+[05:24:28][    470s][KD] soft labels for 6000 patches from 8 teachers in 26s
+[05:25:23][    525s][KD] epoch 1/2 | patches=6000 | loss=0.5636 | 55s
+[05:26:18][    579s][KD] epoch 2/2 | patches=6000 | loss=0.5549 | 55s
+[05:26:27][    589s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 9s
+[05:26:30][    592s][FT] round 1/6 epoch 1/5 | patches=74 | loss=1.4207
+[05:26:33][    595s][FT] round 1/6 epoch 2/5 | patches=74 | loss=1.3712
+[05:26:36][    598s][FT] round 1/6 epoch 3/5 | patches=74 | loss=1.3298
+[05:26:39][    601s][FT] round 1/6 epoch 4/5 | patches=74 | loss=1.3434
+[05:26:43][    604s][FT] round 1/6 epoch 5/5 | patches=74 | loss=1.3357
+[05:26:45][    606s][AL] round 1/6 | labeled=72 | selection_seconds=49.4000 round_seconds=219.7000 restart_student_dice=0.2279 restart_teacher_dice=0.4127 val_dice=0.6071 val_vi=0.6185 val_ece=0.0518
+[05:27:11][    632s][AL] round 2: 1218/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:27:21][    643s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.353, uniform=0.177, top1=0.350, top2=0.374, top3=0.407 -> using top3
+[05:27:21][    643s][WEIGHTS] round 2: teacher Dice on 144 labeled patches + 2 blank (1729464 fg px): c-elegans=0.303, flywire=0.120, berghia=0.147, microns=0.225, oct-glia=0.119, oct-sfl=0.317, snemi=0.337, mouse-brain=0.350 | weights: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.130, snemi=0.317, mouse-brain=0.553
+[05:27:30][    652s][CONS] round 2: on the 72 new patches student Dice=0.510 vs ensemble 0.313 -> pseudo-labels from the student | consistency x0.03
+[05:27:35][    656s][FT] round 2/6 epoch 1/5 | patches=146 | loss=1.1291
+[05:27:39][    660s][FT] round 2/6 epoch 2/5 | patches=146 | loss=1.0857
+[05:27:43][    665s][FT] round 2/6 epoch 3/5 | patches=146 | loss=1.0509
+[05:27:48][    669s][FT] round 2/6 epoch 4/5 | patches=146 | loss=1.0033
+[05:27:52][    673s][FT] round 2/6 epoch 5/5 | patches=146 | loss=0.9764
+[05:27:54][    676s][AL] round 2/6 | labeled=144 | selection_seconds=26.0000 round_seconds=67.2000 pseudo_student_dice=0.5102 pseudo_ensemble_dice=0.3126 consistency_scale=0.0340 val_dice=0.6110 val_vi=0.6381 val_ece=0.0065
+[05:28:20][    701s][AL] round 3: 1237/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:29:14][    755s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.424, flywire=0.151, berghia=0.175, microns=0.280, oct-glia=0.177, oct-sfl=0.437, snemi=0.449, mouse-brain=0.492 | blended with labeled Dice at 0.50
+[05:30:04][    805s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.503, uniform=0.227, top1=0.450, top2=0.464, top3=0.477 -> using mixture
+[05:30:04][    805s][WEIGHTS] round 3: teacher Dice on 216 labeled patches + 2 blank (2913575 fg px): c-elegans=0.382, flywire=0.141, berghia=0.168, microns=0.254, oct-glia=0.139, oct-sfl=0.397, snemi=0.401, mouse-brain=0.450 | weights: c-elegans=0.072, flywire=0.020, berghia=0.002, microns=0.165, oct-glia=0.005, oct-sfl=0.089, snemi=0.185, mouse-brain=0.462
+[05:30:14][    815s][CONS] round 3: on the 72 new patches student Dice=0.592 vs ensemble 0.554 -> pseudo-labels from the student | consistency x0.31
+[05:30:19][    821s][FT] round 3/6 epoch 1/5 | patches=218 | loss=1.0665
+[05:30:25][    826s][FT] round 3/6 epoch 2/5 | patches=218 | loss=1.0386
+[05:30:30][    831s][FT] round 3/6 epoch 3/5 | patches=218 | loss=1.0204
+[05:30:35][    837s][FT] round 3/6 epoch 4/5 | patches=218 | loss=0.9218
+[05:30:41][    843s][FT] round 3/6 epoch 5/5 | patches=218 | loss=0.9563
+[05:30:44][    845s][AL] round 3/6 | labeled=216 | selection_seconds=25.9000 round_seconds=167.2000 pseudo_student_dice=0.5916 pseudo_ensemble_dice=0.5539 consistency_scale=0.3050 val_dice=0.5930 val_vi=0.5589 val_ece=0.0208
+[05:31:09][    871s][AL] round 4: 1527/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:32:05][    926s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.348, flywire=0.192, berghia=0.172, microns=0.361, oct-glia=0.240, oct-sfl=0.485, snemi=0.415, mouse-brain=0.414 | blended with labeled Dice at 0.50
+[05:32:56][    978s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.487, uniform=0.262, top1=0.432, top2=0.445, top3=0.491 -> using mixture
+[05:32:56][    978s][WEIGHTS] round 4: teacher Dice on 288 labeled patches + 2 blank (4568760 fg px): c-elegans=0.366, flywire=0.154, berghia=0.168, microns=0.291, oct-glia=0.158, oct-sfl=0.405, snemi=0.397, mouse-brain=0.432 | weights: c-elegans=0.062, flywire=0.023, berghia=0.002, microns=0.241, oct-glia=0.006, oct-sfl=0.097, snemi=0.178, mouse-brain=0.391
+[05:33:06][    988s][CONS] round 4: on the 72 new patches student Dice=0.600 vs ensemble 0.428 -> pseudo-labels from the student | consistency x0.33
+[05:33:13][    994s][FT] round 4/6 epoch 1/5 | patches=290 | loss=0.9763
+[05:33:19][   1001s][FT] round 4/6 epoch 2/5 | patches=290 | loss=0.9413
+[05:33:26][   1007s][FT] round 4/6 epoch 3/5 | patches=290 | loss=0.9665
+[05:33:33][   1014s][FT] round 4/6 epoch 4/5 | patches=290 | loss=0.9434
+[05:33:39][   1021s][FT] round 4/6 epoch 5/5 | patches=290 | loss=0.8966
+[05:33:41][   1023s][AL] round 4/6 | labeled=288 | selection_seconds=26.0000 round_seconds=175.6000 pseudo_student_dice=0.5996 pseudo_ensemble_dice=0.4281 consistency_scale=0.3320 val_dice=0.6412 val_vi=0.5584 val_ece=0.0317
+[05:34:07][   1049s][AL] round 5: 1370/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:35:04][   1105s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.399, flywire=0.168, berghia=0.184, microns=0.321, oct-glia=0.217, oct-sfl=0.479, snemi=0.434, mouse-brain=0.464 | blended with labeled Dice at 0.50
+[05:35:57][   1159s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.505, uniform=0.229, top1=0.466, top2=0.482, top3=0.485 -> using mixture
+[05:35:57][   1159s][WEIGHTS] round 5: teacher Dice on 360 labeled patches + 2 blank (5969708 fg px): c-elegans=0.395, flywire=0.136, berghia=0.176, microns=0.263, oct-glia=0.142, oct-sfl=0.388, snemi=0.414, mouse-brain=0.466 | weights: c-elegans=0.073, flywire=0.017, berghia=0.002, microns=0.160, oct-glia=0.005, oct-sfl=0.072, snemi=0.187, mouse-brain=0.484
+[05:36:07][   1168s][CONS] round 5: on the 72 new patches student Dice=0.643 vs ensemble 0.444 -> pseudo-labels from the student | consistency x0.48
+[05:36:15][   1176s][FT] round 5/6 epoch 1/5 | patches=362 | loss=1.0087
+[05:36:23][   1184s][FT] round 5/6 epoch 2/5 | patches=362 | loss=0.9617
+[05:36:31][   1192s][FT] round 5/6 epoch 3/5 | patches=362 | loss=0.9284
+[05:36:39][   1200s][FT] round 5/6 epoch 4/5 | patches=362 | loss=0.9311
+[05:36:46][   1208s][FT] round 5/6 epoch 5/5 | patches=362 | loss=0.9370
+[05:36:49][   1210s][AL] round 5/6 | labeled=360 | selection_seconds=26.0000 round_seconds=184.9000 pseudo_student_dice=0.6431 pseudo_ensemble_dice=0.4439 consistency_scale=0.4770 val_dice=0.6439 val_vi=0.5519 val_ece=0.0053
+[05:37:14][   1236s][AL] round 6: 1343/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:38:13][   1294s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.395, flywire=0.160, berghia=0.185, microns=0.301, oct-glia=0.199, oct-sfl=0.445, snemi=0.422, mouse-brain=0.477 | blended with labeled Dice at 0.50
+[05:39:07][   1349s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.491, uniform=0.218, top1=0.478, top2=0.491, top3=0.493 -> using mixture
+[05:39:07][   1349s][WEIGHTS] round 6: teacher Dice on 432 labeled patches + 2 blank (7439104 fg px): c-elegans=0.402, flywire=0.134, berghia=0.183, microns=0.251, oct-glia=0.130, oct-sfl=0.375, snemi=0.409, mouse-brain=0.478 | weights: c-elegans=0.076, flywire=0.016, berghia=0.002, microns=0.138, oct-glia=0.004, oct-sfl=0.062, snemi=0.173, mouse-brain=0.530
+[05:39:18][   1359s][CONS] round 6: on the 72 new patches student Dice=0.681 vs ensemble 0.499 -> pseudo-labels from the student | consistency x0.60
+[05:39:27][   1368s][FT] round 6/6 epoch 1/5 | patches=434 | loss=0.9291
+[05:39:36][   1377s][FT] round 6/6 epoch 2/5 | patches=434 | loss=0.9589
+[05:39:45][   1386s][FT] round 6/6 epoch 3/5 | patches=434 | loss=0.9226
+[05:39:54][   1395s][FT] round 6/6 epoch 4/5 | patches=434 | loss=0.8989
+[05:40:03][   1404s][FT] round 6/6 epoch 5/5 | patches=434 | loss=0.8698
+[05:40:05][   1406s][AL] round 6/6 | labeled=432 | selection_seconds=26.0000 round_seconds=194.0000 pseudo_student_dice=0.6814 pseudo_ensemble_dice=0.4993 consistency_scale=0.6050 val_dice=0.6561 val_vi=0.5270 val_ece=0.0070
+[05:40:08][   1409s][CONS] final consolidation: pseudo-labels refreshed from the student
+[05:40:17][   1418s][FT] final 1/10 epoch 1/1 | patches=434 | loss=0.8707
+[05:40:26][   1427s][FT] final 2/10 epoch 1/1 | patches=434 | loss=0.8387
+[05:40:35][   1437s][FT] final 3/10 epoch 1/1 | patches=434 | loss=0.8128
+[05:40:44][   1446s][FT] final 4/10 epoch 1/1 | patches=434 | loss=0.7855
+[05:40:54][   1455s][FT] final 5/10 epoch 1/1 | patches=434 | loss=0.7703
+[05:41:03][   1464s][FT] final 6/10 epoch 1/1 | patches=434 | loss=0.7591
+[05:41:12][   1473s][FT] final 7/10 epoch 1/1 | patches=434 | loss=0.7416
+[05:41:21][   1482s][FT] final 8/10 epoch 1/1 | patches=434 | loss=0.7372
+[05:41:30][   1491s][FT] final 9/10 epoch 1/1 | patches=434 | loss=0.7155
+[05:41:39][   1501s][FT] final 10/10 epoch 1/1 | patches=434 | loss=0.7190
+[05:41:41][   1503s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=94.4000 val_dice=0.6658 val_vi=0.5143 val_ece=0.0212
+[05:41:41][   1503s][WEIGHTS] final: c-elegans=0.076, flywire=0.016, berghia=0.002, microns=0.138, oct-glia=0.004, oct-sfl=0.062, snemi=0.173, mouse-brain=0.530
+[05:42:21][   1542s][TEST] raw student (no EMA, no TTA): patch_dice=0.6969 patch_vi=0.4345
+[05:42:26][   1548s][TEST] final model without TTA: patch_dice=0.6969 patch_vi=0.4345
+[05:42:26][   1548s][RESOURCES] {"phases": {"setup": {"seconds": 2.22, "peak_cpu_rss_mb": 1028.3, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 33.69, "peak_cpu_rss_mb": 1800.9, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 346.12, "peak_cpu_rss_mb": 3693.7, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 1118.32, "peak_cpu_rss_mb": 3693.7, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4642.0, "rounds": 6}, "inference": {"seconds": 44.73, "peak_cpu_rss_mb": 3693.7, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4644.0, "patches": 2883, "images": 3, "patches_per_second": 64.5, "seconds_per_image": 14.91}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 3693.7, "total_seconds": 1545.1}
+[05:42:26][   1548s][TEST] patch_dice=0.7515 patch_iou=0.6503 patch_precision=0.8460 patch_recall=0.7459 patch_vi=0.4085 image_dice=0.8127 image_iou=0.6879 image_precision=0.8866 image_recall=0.7541 image_vi=0.4586 cal_ece=0.0015 cal_nll=0.1819 cal_brier=0.0545 cal_mean_confidence=0.9268 cal_accuracy=0.9255 cal_predicted_prevalence=0.2046 cal_true_prevalence=0.2301 cal_mean_probability=0.2107
+[05:42:26][   1548s][DONE] dhanyasi-P14-mouse-cerebellum macs-v11-norestart-acq-laplace b5 s2 | test patch Dice=0.7515 VI=0.4085 | image Dice=0.8127 VI=0.4586 | 1548s

macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json

@@ -0,0 +1,785 @@
+{
+  "target": "dhanyasi-P14-mouse-cerebellum",
+  "variant": "macs-v11-norestart-acq-laplace",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "dhanyasi-P14-mouse-cerebellum",
+    "variant": "macs-v11-norestart-acq-laplace",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "laplace",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "dhanyasi-P14-mouse-cerebellum",
+      "--variants",
+      "macs-v11-norestart-acq-laplace",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.05083387092965836,
+    0.15580245740591697,
+    0.010243446832791585,
+    0.41728942804507835,
+    0.03964191869924105,
+    0.053950781313640436,
+    0.10715830600341246,
+    0.16507979077026083
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.05083387092965836,
+      0.15580245740591697,
+      0.010243446832791585,
+      0.41728942804507835,
+      0.03964191869924105,
+      0.053950781313640436,
+      0.10715830600341246,
+      0.16507979077026083
+    ],
+    "per_source": [
+      0.1831792908635141,
+      0.0008117602305311557,
+      0.07130039492253423,
+      0.2901333759297942,
+      0.19473854254781,
+      0.14427929446206111,
+      0.11165168345469406,
+      0.0039056575890611817
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.04944674080252824,
+      "mmd2_mixture": 0.06619580859842833,
+      "mmd2_uniform": 0.10385692805437269,
+      "mmd2_single_source": [
+        0.35157092449944216,
+        0.1914113737092883,
+        0.2296477365874025,
+        0.08478080568303437,
+        0.272622477181822,
+        0.2463089382682188,
+        0.15004801624266395,
+        0.17311813097471462
+      ],
+      "median_squared_distance": 0.18230687630230707,
+      "gammas": [
+        1.3713141548508683,
+        2.7426283097017365,
+        5.485256619403473,
+        10.970513238806946,
+        21.941026477613892
+      ],
+      "max_weight": 0.41728942804507835,
+      "entropy": 1.6749067795900086,
+      "effective_sources": 5.338297487596983,
+      "min_weight": 0.010243446832791585
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.24983749105579034,
+        0.6076273510548802,
+        0.31213616057205607,
+        0.2194742775498013,
+        0.24579726709659655,
+        0.26559858325874997,
+        0.28252517774848496,
+        0.5039038205278614
+      ],
+      "standardized": [
+        -0.6514591536335823,
+        2.0580486812336294,
+        -0.17967738442258652,
+        -0.8813968184025964,
+        -0.6820553769924043,
+        -0.5321019342588011,
+        -0.4039184793457027,
+        1.2725604658220449
+      ],
+      "tau": 0.5,
+      "max_weight": 0.2901333759297942,
+      "entropy": 1.7283708993213025,
+      "effective_sources": 5.631472196315185,
+      "min_weight": 0.0008117602305311557
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "microns_basil",
+    "seconds": 7.0
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.41286983026930635,
+      0.10651095633970442,
+      0.15770527050491268,
+      0.22267858774357308,
+      0.11006956393332956,
+      0.2911632103623596,
+      0.2970295092802324,
+      0.447529566690438
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.20976331913409396,
+      "mixture@raw": 0.2201044706093569,
+      "mixture@logit": 0.1889936769494014,
+      "per_source": 0.19317883340253061,
+      "per_source@raw": 0.20883607926566583,
+      "per_source@logit": 0.15551849586998046,
+      "uniform": 0.19238025487255683,
+      "uniform@raw": 0.21034175472151873,
+      "uniform@logit": 0.15367464892063687
+    },
+    "val_patches": 1024
+  },
+  "temperatures": [
+    2.575807571411133,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.41728942804507835,
+    "entropy": 1.6749067795900086,
+    "effective_sources": 5.338297487596983,
+    "min_weight": 0.010243446832791585
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.2380952380952381,
+    "spearman_weight_taxonomy": 0.7324670207647144,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.00023124060585038732,
+  "kd_seconds": 346.1,
+  "round0_val": {
+    "val_dice": 0.1795085898180082,
+    "val_vi": 0.5071205839268854,
+    "val_ece": 0.1431383671563685
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 72,
+      "selection_seconds": 49.4,
+      "round_seconds": 219.7,
+      "teacher_dice_labeled": [
+        0.34699131258491106,
+        0.15400938239211243,
+        0.17063725881990668,
+        0.29518215496468825,
+        0.1907363491430849,
+        0.40286997571879657,
+        0.402560670886263,
+        0.4126936934594237
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 891480,
+      "evidence_nonempty_patches": 58,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.22785475732277613,
+      "restart_teacher_dice": 0.4126936934594237,
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.1573376001938336,
+        0.311542520327699,
+        0.5311198794784675
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.4486622376468838,
+        "uniform": 0.257123442242594,
+        "top1": 0.4126936934594237,
+        "top2": 0.427720538104064,
+        "top3": 0.5579440810273272
+      },
+      "val_dice": 0.6070700831590716,
+      "val_vi": 0.6184691251569843,
+      "val_ece": 0.05175704877036804
+    },
+    {
+      "round": 2,
+      "n_labeled": 144,
+      "selection_seconds": 26.0,
+      "round_seconds": 67.2,
+      "teacher_dice_labeled": [
+        0.30315355038755776,
+        0.12001301230858169,
+        0.1465265228793779,
+        0.2254534366318072,
+        0.11921934308719918,
+        0.3172824812217642,
+        0.33741894167974107,
+        0.3498583982698843
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1729464,
+      "evidence_nonempty_patches": 100,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.13042238801348177,
+        0.31683400374116205,
+        0.5527436082453562
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.3530103977499102,
+        "uniform": 0.17731911177941523,
+        "top1": 0.3498583982698843,
+        "top2": 0.37390484882696645,
+        "top3": 0.4072446824388192
+      },
+      "pseudo_student_dice": 0.5102164768161838,
+      "pseudo_ensemble_dice": 0.31257957043181106,
+      "pseudo_source": "student",
+      "consistency_scale": 0.034,
+      "val_dice": 0.6110047127116678,
+      "val_vi": 0.638117576109154,
+      "val_ece": 0.006539609706106881
+    },
+    {
+      "round": 3,
+      "n_labeled": 216,
+      "selection_seconds": 25.9,
+      "round_seconds": 167.2,
+      "teacher_dice_labeled": [
+        0.3819753440736413,
+        0.14061082822423704,
+        0.16821673752804672,
+        0.2537591878167585,
+        0.13870534762009895,
+        0.3973398975358051,
+        0.40137266311189335,
+        0.44969605142945246
+      ],
+      "teacher_dice_pool": [
+        0.42386395699475177,
+        0.15099438303350263,
+        0.17495050231510595,
+        0.279604574493834,
+        0.17744932514923112,
+        0.43695860860524693,
+        0.4493108258230955,
+        0.49185366434198335
+      ],
+      "evidence_fg_pixels": 2913575,
+      "evidence_nonempty_patches": 165,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.07229111007511609,
+        0.019827741191740342,
+        0.0017180507822275705,
+        0.1646392588259067,
+        0.0049496923673013095,
+        0.08946576052936685,
+        0.18501163344613708,
+        0.4620967527822041
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5030819805858586,
+        "uniform": 0.2265744778786027,
+        "top1": 0.44969605142945246,
+        "top2": 0.4635763198897255,
+        "top3": 0.477477686886879
+      },
+      "pseudo_student_dice": 0.591602265736447,
+      "pseudo_ensemble_dice": 0.5539129798210711,
+      "pseudo_source": "student",
+      "consistency_scale": 0.305,
+      "val_dice": 0.5930079496634324,
+      "val_vi": 0.5588676820879914,
+      "val_ece": 0.020844457318665577
+    },
+    {
+      "round": 4,
+      "n_labeled": 288,
+      "selection_seconds": 26.0,
+      "round_seconds": 175.6,
+      "teacher_dice_labeled": [
+        0.3659048985997267,
+        0.15353753119751318,
+        0.16844964751011626,
+        0.2908847684179655,
+        0.15789017264557856,
+        0.4047283674903942,
+        0.3966932197117854,
+        0.4321963382581655
+      ],
+      "teacher_dice_pool": [
+        0.3476838243870192,
+        0.19217188601799204,
+        0.172233344062994,
+        0.3611551305335358,
+        0.2396295638748975,
+        0.4852648082757493,
+        0.4149907624250994,
+        0.4142943549311604
+      ],
+      "evidence_fg_pixels": 4568760,
+      "evidence_nonempty_patches": 233,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.062100170113087695,
+        0.022762195720855558,
+        0.0017371945807277208,
+        0.24075133517549557,
+        0.0060491979750396845,
+        0.09717298181633877,
+        0.17810553945189095,
+        0.391321385166564
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.4866048628809274,
+        "uniform": 0.2619714021531271,
+        "top1": 0.4321963382581655,
+        "top2": 0.44509639703054504,
+        "top3": 0.49060202517976464
+      },
+      "pseudo_student_dice": 0.5995607250850519,
+      "pseudo_ensemble_dice": 0.4281460952256102,
+      "pseudo_source": "student",
+      "consistency_scale": 0.332,
+      "val_dice": 0.6411766760615955,
+      "val_vi": 0.5584266727212388,
+      "val_ece": 0.03166756076429024
+    },
+    {
+      "round": 5,
+      "n_labeled": 360,
+      "selection_seconds": 26.0,
+      "round_seconds": 184.9,
+      "teacher_dice_labeled": [
+        0.3953085237250244,
+        0.13647534350877416,
+        0.17612099976432438,
+        0.2628817625648954,
+        0.1419262187673558,
+        0.3879575357978311,
+        0.4141298405020889,
+        0.46600203134969875
+      ],
+      "teacher_dice_pool": [
+        0.3986317767680898,
+        0.16823970417904313,
+        0.1841090592603666,
+        0.3212210790491769,
+        0.21687275845407342,
+        0.4787500737459872,
+        0.43429163241666896,
+        0.4643731595703601
+      ],
+      "evidence_fg_pixels": 5969708,
+      "evidence_nonempty_patches": 301,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0734540416368638,
+        0.016917594038813894,
+        0.0016534433165064843,
+        0.16039018219487342,
+        0.004545606243436821,
+        0.0724328074489532,
+        0.18690817398135207,
+        0.48369815113920034
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5052101415491518,
+        "uniform": 0.22916977644195366,
+        "top1": 0.46600203134969875,
+        "top2": 0.4822501640003084,
+        "top3": 0.48496910511121183
+      },
+      "pseudo_student_dice": 0.6430858944634501,
+      "pseudo_ensemble_dice": 0.44390733840638275,
+      "pseudo_source": "student",
+      "consistency_scale": 0.477,
+      "val_dice": 0.64392883350122,
+      "val_vi": 0.5519165497753227,
+      "val_ece": 0.005270940102244961
+    },
+    {
+      "round": 6,
+      "n_labeled": 432,
+      "selection_seconds": 26.0,
+      "round_seconds": 194.0,
+      "teacher_dice_labeled": [
+        0.4015779769021923,
+        0.13371681261033305,
+        0.18321938102607802,
+        0.250881543038054,
+        0.12972555025287827,
+        0.37503542861674366,
+        0.40930616637479933,
+        0.4782651715898207
+      ],
+      "teacher_dice_pool": [
+        0.3952798389140022,
+        0.15953434509246006,
+        0.1854799560811665,
+        0.30070633663258745,
+        0.19935759134853634,
+        0.44519469823025515,
+        0.4222142629817597,
+        0.4765445169330087
+      ],
+      "evidence_fg_pixels": 7439104,
+      "evidence_nonempty_patches": 370,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0758345688067689,
+        0.01595813326405757,
+        0.001721237600976133,
+        0.13793838128455513,
+        0.0039014731144281904,
+        0.06172185298498267,
+        0.17270423199249021,
+        0.5302201209517412
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.4911926135221738,
+        "uniform": 0.21810257800847072,
+        "top1": 0.4782651715898207,
+        "top2": 0.4907837766601343,
+        "top3": 0.49343918422359895
+      },
+      "pseudo_student_dice": 0.6814277622500998,
+      "pseudo_ensemble_dice": 0.49926128889806726,
+      "pseudo_source": "student",
+      "consistency_scale": 0.605,
+      "val_dice": 0.6561008947304966,
+      "val_vi": 0.5270340585580242,
+      "val_ece": 0.00698206159380761
+    },
+    {
+      "round": 7,
+      "n_labeled": 432,
+      "selection_seconds": 0.0,
+      "round_seconds": 94.4,
+      "final_consolidation": true,
+      "val_dice": 0.6658333978014943,
+      "val_vi": 0.5143445223532903,
+      "val_ece": 0.02122605101113116
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.3018138618955013,
+    "auroc_top20_error": 0.24228212890625
+  },
+  "al_seconds": 1118.3,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.0758345688067689,
+    0.01595813326405757,
+    0.001721237600976133,
+    0.13793838128455513,
+    0.0039014731144281904,
+    0.06172185298498267,
+    0.17270423199249021,
+    0.5302201209517412
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.5302201209517412,
+    "entropy": 1.3790735363210445,
+    "effective_sources": 3.971220730881249,
+    "min_weight": 0.001721237600976133
+  },
+  "top_source_final": "whole-mouse-brain",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.6968826503970063,
+    "patch_iou": 0.5873942958945824,
+    "patch_precision": 0.7688665778942382,
+    "patch_recall": 0.7397190870694399,
+    "patch_vi": 0.434540009497679,
+    "n_patches": 2883,
+    "cal_ece": 0.013984583116174568,
+    "cal_nll": 0.20291160477716605,
+    "cal_brier": 0.05947698933410016,
+    "cal_mean_confidence": 0.933616504075245,
+    "cal_accuracy": 0.9196319209590704,
+    "cal_predicted_prevalence": 0.20512520865851544,
+    "cal_true_prevalence": 0.23013442215140478,
+    "cal_mean_probability": 0.2089750948953597
+  },
+  "test_notta": {
+    "patch_dice": 0.6968826503970063,
+    "patch_iou": 0.5873942958945824,
+    "patch_precision": 0.7688665778942382,
+    "patch_recall": 0.7397190870694399,
+    "patch_vi": 0.434540009497679,
+    "n_patches": 2883,
+    "cal_ece": 0.013984583116174568,
+    "cal_nll": 0.20291160477716605,
+    "cal_brier": 0.05947698933410016,
+    "cal_mean_confidence": 0.933616504075245,
+    "cal_accuracy": 0.9196319209590704,
+    "cal_predicted_prevalence": 0.20512520865851544,
+    "cal_true_prevalence": 0.23013442215140478,
+    "cal_mean_probability": 0.2089750948953597
+  },
+  "test": {
+    "patch_dice": 0.7515223139947016,
+    "patch_iou": 0.6502919854628868,
+    "patch_precision": 0.8459618413535717,
+    "patch_recall": 0.7458955232078299,
+    "patch_vi": 0.4085141725179584,
+    "n_patches": 2883,
+    "image_dice": 0.8126819379613308,
+    "image_iou": 0.6878766774473583,
+    "image_precision": 0.8865888050426772,
+    "image_recall": 0.7540789897163429,
+    "image_vi": 0.4586308388710216,
+    "n_images": 3,
+    "cal_ece": 0.0014887245718914603,
+    "cal_nll": 0.1819273293225221,
+    "cal_brier": 0.05446771271927853,
+    "cal_mean_confidence": 0.926763834764255,
+    "cal_accuracy": 0.9254770692421089,
+    "cal_predicted_prevalence": 0.2046103369970517,
+    "cal_true_prevalence": 0.23013442215140478,
+    "cal_mean_probability": 0.2107195412252572
+  },
+  "eval_seconds": 44.7,
+  "total_seconds": 1547.5,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 2.22,
+        "peak_cpu_rss_mb": 1028.3,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 33.69,
+        "peak_cpu_rss_mb": 1800.9,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 346.12,
+        "peak_cpu_rss_mb": 3693.7,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 1118.32,
+        "peak_cpu_rss_mb": 3693.7,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4642.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 44.73,
+        "peak_cpu_rss_mb": 3693.7,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4644.0,
+        "patches": 2883,
+        "images": 3,
+        "patches_per_second": 64.5,
+        "seconds_per_image": 14.91
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 3693.7,
+    "total_seconds": 1545.1
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-laplace_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,72,49.4,219.7,"[0.34699131258491106, 0.15400938239211243, 0.17063725881990668, 0.29518215496468825, 0.1907363491430849, 0.40286997571879657, 0.402560670886263, 0.4126936934594237]",,891480,58,True,kept,0.22785475732277613,0.4126936934594237,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.1573376001938336, 0.311542520327699, 0.5311198794784675]",top3,"{'mixture': 0.4486622376468838, 'uniform': 0.257123442242594, 'top1': 0.4126936934594237, 'top2': 0.427720538104064, 'top3': 0.5579440810273272}",0.6070700831590716,0.6184691251569843,0.05175704877036804,,,,,
+2,144,26.0,67.2,"[0.30315355038755776, 0.12001301230858169, 0.1465265228793779, 0.2254534366318072, 0.11921934308719918, 0.3172824812217642, 0.33741894167974107, 0.3498583982698843]",,1729464,100,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.13042238801348177, 0.31683400374116205, 0.5527436082453562]",top3,"{'mixture': 0.3530103977499102, 'uniform': 0.17731911177941523, 'top1': 0.3498583982698843, 'top2': 0.37390484882696645, 'top3': 0.4072446824388192}",0.6110047127116678,0.638117576109154,0.006539609706106881,0.5102164768161838,0.31257957043181106,student,0.034,
+3,216,25.9,167.2,"[0.3819753440736413, 0.14061082822423704, 0.16821673752804672, 0.2537591878167585, 0.13870534762009895, 0.3973398975358051, 0.40137266311189335, 0.44969605142945246]","[0.42386395699475177, 0.15099438303350263, 0.17495050231510595, 0.279604574493834, 0.17744932514923112, 0.43695860860524693, 0.4493108258230955, 0.49185366434198335]",2913575,165,False,,,,"[0.07229111007511609, 0.019827741191740342, 0.0017180507822275705, 0.1646392588259067, 0.0049496923673013095, 0.08946576052936685, 0.18501163344613708, 0.4620967527822041]",mixture,"{'mixture': 0.5030819805858586, 'uniform': 0.2265744778786027, 'top1': 0.44969605142945246, 'top2': 0.4635763198897255, 'top3': 0.477477686886879}",0.5930079496634324,0.5588676820879914,0.020844457318665577,0.591602265736447,0.5539129798210711,student,0.305,
+4,288,26.0,175.6,"[0.3659048985997267, 0.15353753119751318, 0.16844964751011626, 0.2908847684179655, 0.15789017264557856, 0.4047283674903942, 0.3966932197117854, 0.4321963382581655]","[0.3476838243870192, 0.19217188601799204, 0.172233344062994, 0.3611551305335358, 0.2396295638748975, 0.4852648082757493, 0.4149907624250994, 0.4142943549311604]",4568760,233,False,,,,"[0.062100170113087695, 0.022762195720855558, 0.0017371945807277208, 0.24075133517549557, 0.0060491979750396845, 0.09717298181633877, 0.17810553945189095, 0.391321385166564]",mixture,"{'mixture': 0.4866048628809274, 'uniform': 0.2619714021531271, 'top1': 0.4321963382581655, 'top2': 0.44509639703054504, 'top3': 0.49060202517976464}",0.6411766760615955,0.5584266727212388,0.03166756076429024,0.5995607250850519,0.4281460952256102,student,0.332,
+5,360,26.0,184.9,"[0.3953085237250244, 0.13647534350877416, 0.17612099976432438, 0.2628817625648954, 0.1419262187673558, 0.3879575357978311, 0.4141298405020889, 0.46600203134969875]","[0.3986317767680898, 0.16823970417904313, 0.1841090592603666, 0.3212210790491769, 0.21687275845407342, 0.4787500737459872, 0.43429163241666896, 0.4643731595703601]",5969708,301,False,,,,"[0.0734540416368638, 0.016917594038813894, 0.0016534433165064843, 0.16039018219487342, 0.004545606243436821, 0.0724328074489532, 0.18690817398135207, 0.48369815113920034]",mixture,"{'mixture': 0.5052101415491518, 'uniform': 0.22916977644195366, 'top1': 0.46600203134969875, 'top2': 0.4822501640003084, 'top3': 0.48496910511121183}",0.64392883350122,0.5519165497753227,0.005270940102244961,0.6430858944634501,0.44390733840638275,student,0.477,
+6,432,26.0,194.0,"[0.4015779769021923, 0.13371681261033305, 0.18321938102607802, 0.250881543038054, 0.12972555025287827, 0.37503542861674366, 0.40930616637479933, 0.4782651715898207]","[0.3952798389140022, 0.15953434509246006, 0.1854799560811665, 0.30070633663258745, 0.19935759134853634, 0.44519469823025515, 0.4222142629817597, 0.4765445169330087]",7439104,370,False,,,,"[0.0758345688067689, 0.01595813326405757, 0.001721237600976133, 0.13793838128455513, 0.0039014731144281904, 0.06172185298498267, 0.17270423199249021, 0.5302201209517412]",mixture,"{'mixture': 0.4911926135221738, 'uniform': 0.21810257800847072, 'top1': 0.4782651715898207, 'top2': 0.4907837766601343, 'top3': 0.49343918422359895}",0.6561008947304966,0.5270340585580242,0.00698206159380761,0.6814277622500998,0.49926128889806726,student,0.605,
+7,432,0.0,94.4,,,,,,,,,,,,0.6658333978014943,0.5143445223532903,0.02122605101113116,,,,,True

macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-random/budget-5/seed-2/log.txt

@@ -0,0 +1,103 @@
+[06:17:56][      0s][START] target=dhanyasi-P14-mouse-cerebellum variant=macs-v11-norestart-acq-random budget=5.0% seed=2 | {"task": "membrane", "target": "dhanyasi-P14-mouse-cerebellum", "variant": "macs-v11-norestart-acq-random", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "random", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 0, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[06:17:56][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[06:17:56][      0s][DATA] dhanyasi-P14-mouse-cerebellum: train=8649 val=1922 test=2883 patches
+[06:17:57][      1s][CALIB] source-val temperatures: c-elegans=2.58, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, snemi=3.55, mouse-brain=4.08
+[06:17:58][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[06:18:03][      8s][WEIGHTS] mixture: c-elegans=0.051, flywire=0.156, berghia=0.010, microns=0.417, oct-glia=0.040, oct-sfl=0.054, snemi=0.107, mouse-brain=0.165 | eff.sources=5.34 | MMD2 mix=0.0662 uniform=0.1039
+[06:18:03][      8s][WEIGHTS] per-source: c-elegans=0.183, flywire=0.001, berghia=0.071, microns=0.290, oct-glia=0.195, oct-sfl=0.144, snemi=0.112, mouse-brain=0.004 | min-MMD source (NeuroADDA rule) = microns
+[06:18:27][     32s][TEACHERS] target-val Dice: c-elegans=0.413, flywire=0.107, berghia=0.158, microns=0.223, oct-glia=0.110, oct-sfl=0.291, snemi=0.297, mouse-brain=0.448
+[06:18:27][     32s][TEACHERS] ensemble target-val Dice: mixture=0.210, mixture@raw=0.220, mixture@logit=0.189, per_source=0.193, per_source@raw=0.209, per_source@logit=0.156, uniform=0.192, uniform@raw=0.210, uniform@logit=0.154
+[06:18:27][     32s][WEIGHTS] using mixture: c-elegans=0.051, flywire=0.156, berghia=0.010, microns=0.417, oct-glia=0.040, oct-sfl=0.054, snemi=0.107, mouse-brain=0.165
+[06:18:27][     32s][INIT] student initialised from teacher microns (weight 0.417)
+[06:18:30][     34s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[06:18:30][     34s][KD] distilling the mixture ensemble (prob space) into the student
+[06:19:38][    102s][KD] soft labels for 6000 patches from 8 teachers in 69s
+[06:20:32][    157s][KD] epoch 1/5 | patches=6000 | loss=0.5835 | 54s
+[06:21:27][    212s][KD] epoch 2/5 | patches=6000 | loss=0.4531 | 55s
+[06:22:22][    267s][KD] epoch 3/5 | patches=6000 | loss=0.4309 | 55s
+[06:23:16][    321s][KD] epoch 4/5 | patches=6000 | loss=0.4249 | 54s
+[06:24:09][    374s][KD] epoch 5/5 | patches=6000 | loss=0.4230 | 53s
+[06:24:12][    376s][ROUND0] before any target label: val_dice=0.1712 val_vi=0.5021
+[06:24:12][    376s][AL] patches=8649 (pool 8647 after dropping blank; free blank negatives: True) | budget=432 (5.0%) | rounds=6 | per_round=72 | acquisition=random | first_round=stratified
+[06:24:24][    388s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': -0.01683200768757291, 'auroc_top20_error': 0.5015251953125}
+[06:24:47][    412s][AL] round 1: stratified cold start: 36 patches from the membrane-rich half (ensemble prevalence >= 0.184) + 36 from the membrane-poor half of 2000 candidates
+[06:24:55][    419s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.449, uniform=0.257, top1=0.413, top2=0.428, top3=0.558 -> using top3
+[06:24:57][    421s][WEIGHTS] round 1: student kept (labeled Dice 0.216 vs best teacher mouse-brain 0.413); distilling the corrected ensemble into it
+[06:24:57][    421s][WEIGHTS] round 1: teacher Dice on 72 labeled patches + 2 blank (891480 fg px): c-elegans=0.347, flywire=0.154, berghia=0.171, microns=0.295, oct-glia=0.191, oct-sfl=0.403, snemi=0.403, mouse-brain=0.413 | weights: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.157, snemi=0.312, mouse-brain=0.531
+[06:25:23][    447s][KD] soft labels for 6000 patches from 8 teachers in 26s
+[06:26:18][    502s][KD] epoch 1/2 | patches=6000 | loss=0.5647 | 55s
+[06:27:14][    558s][KD] epoch 2/2 | patches=6000 | loss=0.5552 | 56s
+[06:27:23][    567s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 9s
+[06:27:26][    570s][FT] round 1/6 epoch 1/5 | patches=74 | loss=1.3886
+[06:27:29][    574s][FT] round 1/6 epoch 2/5 | patches=74 | loss=1.3259
+[06:27:32][    577s][FT] round 1/6 epoch 3/5 | patches=74 | loss=1.3300
+[06:27:35][    580s][FT] round 1/6 epoch 4/5 | patches=74 | loss=1.3050
+[06:27:38][    582s][FT] round 1/6 epoch 5/5 | patches=74 | loss=1.2636
+[06:27:41][    585s][AL] round 1/6 | labeled=72 | selection_seconds=35.6000 round_seconds=206.6000 restart_student_dice=0.2157 restart_teacher_dice=0.4127 val_dice=0.5873 val_vi=0.6600 val_ece=0.0484
+[06:27:51][    595s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.511, uniform=0.254, top1=0.449, top2=0.457, top3=0.492 -> using mixture
+[06:27:51][    595s][WEIGHTS] round 2: teacher Dice on 144 labeled patches + 2 blank (2038859 fg px): c-elegans=0.388, flywire=0.149, berghia=0.170, microns=0.291, oct-glia=0.178, oct-sfl=0.419, snemi=0.405, mouse-brain=0.449 | weights: c-elegans=0.069, flywire=0.019, berghia=0.002, microns=0.216, oct-glia=0.007, oct-sfl=0.100, snemi=0.172, mouse-brain=0.415
+[06:28:02][    606s][CONS] round 2: on the 72 new patches student Dice=0.772 vs ensemble 0.499 -> pseudo-labels from the student | consistency x0.91
+[06:28:06][    611s][FT] round 2/6 epoch 1/5 | patches=146 | loss=1.0746
+[06:28:11][    615s][FT] round 2/6 epoch 2/5 | patches=146 | loss=1.0634
+[06:28:15][    620s][FT] round 2/6 epoch 3/5 | patches=146 | loss=1.0433
+[06:28:20][    624s][FT] round 2/6 epoch 4/5 | patches=146 | loss=1.0409
+[06:28:24][    629s][FT] round 2/6 epoch 5/5 | patches=146 | loss=1.0049
+[06:28:27][    631s][AL] round 2/6 | labeled=144 | selection_seconds=0.2000 round_seconds=43.8000 pseudo_student_dice=0.7723 pseudo_ensemble_dice=0.4989 consistency_scale=0.9080 val_dice=0.6586 val_vi=0.5740 val_ece=0.0054
+[06:29:21][    685s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.442, flywire=0.155, berghia=0.189, microns=0.287, oct-glia=0.186, oct-sfl=0.457, snemi=0.439, mouse-brain=0.516 | blended with labeled Dice at 0.50
+[06:30:11][    735s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.545, uniform=0.266, top1=0.485, top2=0.491, top3=0.510 -> using mixture
+[06:30:11][    735s][WEIGHTS] round 3: teacher Dice on 216 labeled patches + 2 blank (3013404 fg px): c-elegans=0.414, flywire=0.154, berghia=0.180, microns=0.291, oct-glia=0.183, oct-sfl=0.442, snemi=0.418, mouse-brain=0.485 | weights: c-elegans=0.072, flywire=0.016, berghia=0.001, microns=0.172, oct-glia=0.006, oct-sfl=0.101, snemi=0.158, mouse-brain=0.474
+[06:30:21][    746s][CONS] round 3: on the 72 new patches student Dice=0.800 vs ensemble 0.507 -> pseudo-labels from the student | consistency x1.00
+[06:30:27][    751s][FT] round 3/6 epoch 1/5 | patches=218 | loss=0.9337
+[06:30:32][    757s][FT] round 3/6 epoch 2/5 | patches=218 | loss=0.9435
+[06:30:38][    762s][FT] round 3/6 epoch 3/5 | patches=218 | loss=0.8940
+[06:30:43][    768s][FT] round 3/6 epoch 4/5 | patches=218 | loss=0.9051
+[06:30:49][    773s][FT] round 3/6 epoch 5/5 | patches=218 | loss=0.8813
+[06:30:51][    776s][AL] round 3/6 | labeled=216 | selection_seconds=0.2000 round_seconds=142.6000 pseudo_student_dice=0.7999 pseudo_ensemble_dice=0.5074 consistency_scale=1.0000 val_dice=0.6489 val_vi=0.5501 val_ece=0.0084
+[06:31:46][    831s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.393, flywire=0.183, berghia=0.188, microns=0.344, oct-glia=0.231, oct-sfl=0.497, snemi=0.420, mouse-brain=0.455 | blended with labeled Dice at 0.50
+[06:32:38][    882s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.493, uniform=0.298, top1=0.466, top2=0.469, top3=0.578 -> using top3
+[06:32:38][    882s][WEIGHTS] round 4: teacher Dice on 288 labeled patches + 2 blank (4032333 fg px): c-elegans=0.389, flywire=0.168, berghia=0.180, microns=0.323, oct-glia=0.203, oct-sfl=0.466, snemi=0.419, mouse-brain=0.450 | weights: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.206, snemi=0.255, mouse-brain=0.539
+[06:32:49][    893s][CONS] round 4: on the 72 new patches student Dice=0.783 vs ensemble 0.571 -> pseudo-labels from the student | consistency x0.94
+[06:32:56][    900s][FT] round 4/6 epoch 1/5 | patches=290 | loss=0.8305
+[06:33:03][    907s][FT] round 4/6 epoch 2/5 | patches=290 | loss=0.8281
+[06:33:10][    914s][FT] round 4/6 epoch 3/5 | patches=290 | loss=0.7818
+[06:33:16][    921s][FT] round 4/6 epoch 4/5 | patches=290 | loss=0.7958
+[06:33:23][    927s][FT] round 4/6 epoch 5/5 | patches=290 | loss=0.8050
+[06:33:25][    930s][AL] round 4/6 | labeled=288 | selection_seconds=0.2000 round_seconds=151.8000 pseudo_student_dice=0.7826 pseudo_ensemble_dice=0.5707 consistency_scale=0.9420 val_dice=0.6588 val_vi=0.5911 val_ece=0.0125
+[06:34:22][    987s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.455, flywire=0.146, berghia=0.189, microns=0.268, oct-glia=0.170, oct-sfl=0.425, snemi=0.427, mouse-brain=0.522 | blended with labeled Dice at 0.50
+[06:35:16][   1040s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.528, uniform=0.258, top1=0.480, top2=0.483, top3=0.500 -> using mixture
+[06:35:16][   1040s][WEIGHTS] round 5: teacher Dice on 360 labeled patches + 2 blank (4939842 fg px): c-elegans=0.419, flywire=0.150, berghia=0.182, microns=0.284, oct-glia=0.175, oct-sfl=0.432, snemi=0.422, mouse-brain=0.480 | weights: c-elegans=0.078, flywire=0.016, berghia=0.001, microns=0.167, oct-glia=0.005, oct-sfl=0.095, snemi=0.170, mouse-brain=0.467
+[06:35:27][   1051s][CONS] round 5: on the 72 new patches student Dice=0.803 vs ensemble 0.467 -> pseudo-labels from the student | consistency x1.00
+[06:35:35][   1059s][FT] round 5/6 epoch 1/5 | patches=362 | loss=0.8099
+[06:35:43][   1067s][FT] round 5/6 epoch 2/5 | patches=362 | loss=0.8055
+[06:35:51][   1075s][FT] round 5/6 epoch 3/5 | patches=362 | loss=0.7721
+[06:35:59][   1083s][FT] round 5/6 epoch 4/5 | patches=362 | loss=0.7884
+[06:36:07][   1091s][FT] round 5/6 epoch 5/5 | patches=362 | loss=0.7350
+[06:36:09][   1093s][AL] round 5/6 | labeled=360 | selection_seconds=0.2000 round_seconds=161.3000 pseudo_student_dice=0.8026 pseudo_ensemble_dice=0.4667 consistency_scale=1.0000 val_dice=0.6882 val_vi=0.5241 val_ece=0.0292
+[06:37:07][   1151s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.405, flywire=0.161, berghia=0.183, microns=0.297, oct-glia=0.199, oct-sfl=0.448, snemi=0.416, mouse-brain=0.481 | blended with labeled Dice at 0.50
+[06:38:02][   1206s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.516, uniform=0.277, top1=0.456, top2=0.466, top3=0.531 -> using top3
+[06:38:02][   1206s][WEIGHTS] round 6: teacher Dice on 432 labeled patches + 2 blank (5814272 fg px): c-elegans=0.392, flywire=0.159, berghia=0.179, microns=0.301, oct-glia=0.190, oct-sfl=0.449, snemi=0.417, mouse-brain=0.456 | weights: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.175, snemi=0.253, mouse-brain=0.572
+[06:38:13][   1217s][CONS] round 6: on the 72 new patches student Dice=0.804 vs ensemble 0.511 -> pseudo-labels from the student | consistency x1.00
+[06:38:22][   1226s][FT] round 6/6 epoch 1/5 | patches=434 | loss=0.6374
+[06:38:31][   1235s][FT] round 6/6 epoch 2/5 | patches=434 | loss=0.6052
+[06:38:40][   1244s][FT] round 6/6 epoch 3/5 | patches=434 | loss=0.6068
+[06:38:49][   1253s][FT] round 6/6 epoch 4/5 | patches=434 | loss=0.5832
+[06:38:58][   1262s][FT] round 6/6 epoch 5/5 | patches=434 | loss=0.6386
+[06:39:00][   1265s][AL] round 6/6 | labeled=432 | selection_seconds=0.2000 round_seconds=169.2000 pseudo_student_dice=0.8042 pseudo_ensemble_dice=0.5110 consistency_scale=1.0000 val_dice=0.6883 val_vi=0.5339 val_ece=0.0105
+[06:39:04][   1268s][CONS] final consolidation: pseudo-labels refreshed from the student
+[06:39:13][   1277s][FT] final 1/10 epoch 1/1 | patches=434 | loss=0.6010
+[06:39:22][   1286s][FT] final 2/10 epoch 1/1 | patches=434 | loss=0.5822
+[06:39:31][   1296s][FT] final 3/10 epoch 1/1 | patches=434 | loss=0.5734
+[06:39:41][   1305s][FT] final 4/10 epoch 1/1 | patches=434 | loss=0.5643
+[06:39:50][   1314s][FT] final 5/10 epoch 1/1 | patches=434 | loss=0.5516
+[06:39:59][   1323s][FT] final 6/10 epoch 1/1 | patches=434 | loss=0.5410
+[06:40:08][   1332s][FT] final 7/10 epoch 1/1 | patches=434 | loss=0.5213
+[06:40:17][   1341s][FT] final 8/10 epoch 1/1 | patches=434 | loss=0.5250
+[06:40:26][   1351s][FT] final 9/10 epoch 1/1 | patches=434 | loss=0.5119
+[06:40:36][   1360s][FT] final 10/10 epoch 1/1 | patches=434 | loss=0.5281
+[06:40:38][   1362s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=95.4000 val_dice=0.7122 val_vi=0.5155 val_ece=0.0282
+[06:40:38][   1362s][WEIGHTS] final: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.175, snemi=0.253, mouse-brain=0.572
+[06:41:17][   1401s][TEST] raw student (no EMA, no TTA): patch_dice=0.7470 patch_vi=0.4291
+[06:41:22][   1407s][TEST] final model without TTA: patch_dice=0.7470 patch_vi=0.4291
+[06:41:22][   1407s][RESOURCES] {"phases": {"setup": {"seconds": 1.22, "peak_cpu_rss_mb": 1028.3, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 32.54, "peak_cpu_rss_mb": 1801.3, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 339.71, "peak_cpu_rss_mb": 3693.9, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 986.15, "peak_cpu_rss_mb": 3693.9, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4642.0, "rounds": 6}, "inference": {"seconds": 44.7, "peak_cpu_rss_mb": 3693.9, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4644.0, "patches": 2883, "images": 3, "patches_per_second": 64.5, "seconds_per_image": 14.9}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 3693.9, "total_seconds": 1404.3}
+[06:41:22][   1407s][TEST] patch_dice=0.7784 patch_iou=0.6821 patch_precision=0.8233 patch_recall=0.8023 patch_vi=0.4046 image_dice=0.8283 image_iou=0.7096 image_precision=0.8551 image_recall=0.8058 image_vi=0.4575 cal_ece=0.0084 cal_nll=0.1759 cal_brier=0.0526 cal_mean_confidence=0.9360 cal_accuracy=0.9277 cal_predicted_prevalence=0.2215 cal_true_prevalence=0.2301 cal_mean_probability=0.2258
+[06:41:23][   1407s][DONE] dhanyasi-P14-mouse-cerebellum macs-v11-norestart-acq-random b5 s2 | test patch Dice=0.7784 VI=0.4046 | image Dice=0.8283 VI=0.4575 | 1407s

macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-random/budget-5/seed-2/result.json

@@ -0,0 +1,785 @@
+{
+  "target": "dhanyasi-P14-mouse-cerebellum",
+  "variant": "macs-v11-norestart-acq-random",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "dhanyasi-P14-mouse-cerebellum",
+    "variant": "macs-v11-norestart-acq-random",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "random",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 0,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "dhanyasi-P14-mouse-cerebellum",
+      "--variants",
+      "macs-v11-norestart-acq-random",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.05083387092965836,
+    0.15580245740591697,
+    0.010243446832791585,
+    0.41728942804507835,
+    0.03964191869924105,
+    0.053950781313640436,
+    0.10715830600341246,
+    0.16507979077026083
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.05083387092965836,
+      0.15580245740591697,
+      0.010243446832791585,
+      0.41728942804507835,
+      0.03964191869924105,
+      0.053950781313640436,
+      0.10715830600341246,
+      0.16507979077026083
+    ],
+    "per_source": [
+      0.1831792908635141,
+      0.0008117602305311557,
+      0.07130039492253423,
+      0.2901333759297942,
+      0.19473854254781,
+      0.14427929446206111,
+      0.11165168345469406,
+      0.0039056575890611817
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.04944674080252824,
+      "mmd2_mixture": 0.06619580859842833,
+      "mmd2_uniform": 0.10385692805437269,
+      "mmd2_single_source": [
+        0.35157092449944216,
+        0.1914113737092883,
+        0.2296477365874025,
+        0.08478080568303437,
+        0.272622477181822,
+        0.2463089382682188,
+        0.15004801624266395,
+        0.17311813097471462
+      ],
+      "median_squared_distance": 0.18230687630230707,
+      "gammas": [
+        1.3713141548508683,
+        2.7426283097017365,
+        5.485256619403473,
+        10.970513238806946,
+        21.941026477613892
+      ],
+      "max_weight": 0.41728942804507835,
+      "entropy": 1.6749067795900086,
+      "effective_sources": 5.338297487596983,
+      "min_weight": 0.010243446832791585
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.24983749105579034,
+        0.6076273510548802,
+        0.31213616057205607,
+        0.2194742775498013,
+        0.24579726709659655,
+        0.26559858325874997,
+        0.28252517774848496,
+        0.5039038205278614
+      ],
+      "standardized": [
+        -0.6514591536335823,
+        2.0580486812336294,
+        -0.17967738442258652,
+        -0.8813968184025964,
+        -0.6820553769924043,
+        -0.5321019342588011,
+        -0.4039184793457027,
+        1.2725604658220449
+      ],
+      "tau": 0.5,
+      "max_weight": 0.2901333759297942,
+      "entropy": 1.7283708993213025,
+      "effective_sources": 5.631472196315185,
+      "min_weight": 0.0008117602305311557
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "microns_basil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.41286983026930635,
+      0.10651095633970442,
+      0.15770527050491268,
+      0.22267858774357308,
+      0.11006956393332956,
+      0.2911632103623596,
+      0.2970295092802324,
+      0.447529566690438
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.20976331913409396,
+      "mixture@raw": 0.2201044706093569,
+      "mixture@logit": 0.1889936769494014,
+      "per_source": 0.19317883340253061,
+      "per_source@raw": 0.20883607926566583,
+      "per_source@logit": 0.15551849586998046,
+      "uniform": 0.19238025487255683,
+      "uniform@raw": 0.21034175472151873,
+      "uniform@logit": 0.15367464892063687
+    },
+    "val_patches": 1024
+  },
+  "temperatures": [
+    2.575807571411133,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.41728942804507835,
+    "entropy": 1.6749067795900086,
+    "effective_sources": 5.338297487596983,
+    "min_weight": 0.010243446832791585
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.2380952380952381,
+    "spearman_weight_taxonomy": 0.7324670207647144,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.00023124060585038732,
+  "kd_seconds": 339.7,
+  "round0_val": {
+    "val_dice": 0.1711621108784323,
+    "val_vi": 0.5020588886973317,
+    "val_ece": 0.14820936102204899
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 72,
+      "selection_seconds": 35.6,
+      "round_seconds": 206.6,
+      "teacher_dice_labeled": [
+        0.34699131258491106,
+        0.15400938239211243,
+        0.17063725881990668,
+        0.29518215496468825,
+        0.1907363491430849,
+        0.40286997571879657,
+        0.402560670886263,
+        0.4126936934594237
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 891480,
+      "evidence_nonempty_patches": 58,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.21571648537522917,
+      "restart_teacher_dice": 0.4126936934594237,
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.1573376001938336,
+        0.311542520327699,
+        0.5311198794784675
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.4486622376468838,
+        "uniform": 0.257123442242594,
+        "top1": 0.4126936934594237,
+        "top2": 0.427720538104064,
+        "top3": 0.5579440810273272
+      },
+      "val_dice": 0.587305604106643,
+      "val_vi": 0.6599535991066656,
+      "val_ece": 0.04835740491699966
+    },
+    {
+      "round": 2,
+      "n_labeled": 144,
+      "selection_seconds": 0.2,
+      "round_seconds": 43.8,
+      "teacher_dice_labeled": [
+        0.3877112948623053,
+        0.14909626611196175,
+        0.17005234419964096,
+        0.29110798999593296,
+        0.17764307442922855,
+        0.41866300287842756,
+        0.40459903528437674,
+        0.4492049901441661
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 2038859,
+      "evidence_nonempty_patches": 117,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.06909489706776745,
+        0.019479382824218353,
+        0.0015792762266002333,
+        0.21586754477923684,
+        0.006593754437182248,
+        0.09993372841912157,
+        0.17244918911249582,
+        0.4150022271333776
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5114063614291914,
+        "uniform": 0.2542809724207618,
+        "top1": 0.4492049901441661,
+        "top2": 0.4573143432435651,
+        "top3": 0.4917913428812348
+      },
+      "pseudo_student_dice": 0.7723456432892528,
+      "pseudo_ensemble_dice": 0.4989113809299943,
+      "pseudo_source": "student",
+      "consistency_scale": 0.908,
+      "val_dice": 0.6585940233579239,
+      "val_vi": 0.5740162973538123,
+      "val_ece": 0.005417245034970651
+    },
+    {
+      "round": 3,
+      "n_labeled": 216,
+      "selection_seconds": 0.2,
+      "round_seconds": 142.6,
+      "teacher_dice_labeled": [
+        0.4140682111607643,
+        0.1536399927755801,
+        0.1799205572303296,
+        0.29112126324332566,
+        0.18334845976830494,
+        0.4423851087364882,
+        0.41803091046534124,
+        0.4848902568732807
+      ],
+      "teacher_dice_pool": [
+        0.4416778556160547,
+        0.15487823535765688,
+        0.18914476821851361,
+        0.28655053365525646,
+        0.18583099402223757,
+        0.4567643200946957,
+        0.43939763997607106,
+        0.5156445163650382
+      ],
+      "evidence_fg_pixels": 3013404,
+      "evidence_nonempty_patches": 174,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0718401217611655,
+        0.016284063956197966,
+        0.001392418083738922,
+        0.17246439507692587,
+        0.0055765473665446265,
+        0.10120229003831627,
+        0.157561280540448,
+        0.4736788831766628
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5454374977215515,
+        "uniform": 0.2664851046953125,
+        "top1": 0.4848902568732807,
+        "top2": 0.49108429525345076,
+        "top3": 0.5097392015266642
+      },
+      "pseudo_student_dice": 0.7999457618489225,
+      "pseudo_ensemble_dice": 0.5073909859154929,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6488990698124532,
+      "val_vi": 0.5500686787278544,
+      "val_ece": 0.008365029822864706
+    },
+    {
+      "round": 4,
+      "n_labeled": 288,
+      "selection_seconds": 0.2,
+      "round_seconds": 151.8,
+      "teacher_dice_labeled": [
+        0.38945019463626285,
+        0.1679853583375917,
+        0.18044124015986018,
+        0.32282208176596117,
+        0.2034846994902197,
+        0.4660624113903503,
+        0.41870020084547344,
+        0.4503311467877327
+      ],
+      "teacher_dice_pool": [
+        0.3928618027754761,
+        0.18289820943264143,
+        0.1883542828074731,
+        0.3437577001051289,
+        0.2306743626695978,
+        0.4966166642733409,
+        0.4197338102857259,
+        0.45470006684976394
+      ],
+      "evidence_fg_pixels": 4032333,
+      "evidence_nonempty_patches": 233,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.2061290137522376,
+        0.25496217742297783,
+        0.5389088088247845
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.4927856038827534,
+        "uniform": 0.29765161943165275,
+        "top1": 0.4660624113903503,
+        "top2": 0.4691393882366579,
+        "top3": 0.5775226351451215
+      },
+      "pseudo_student_dice": 0.7826412650210633,
+      "pseudo_ensemble_dice": 0.5706861006865913,
+      "pseudo_source": "student",
+      "consistency_scale": 0.942,
+      "val_dice": 0.658762843387845,
+      "val_vi": 0.5911395322997408,
+      "val_ece": 0.012483891875880187
+    },
+    {
+      "round": 5,
+      "n_labeled": 360,
+      "selection_seconds": 0.2,
+      "round_seconds": 161.3,
+      "teacher_dice_labeled": [
+        0.41854435812646784,
+        0.1495941701206981,
+        0.1822344165135339,
+        0.28407492428974873,
+        0.17477511072235818,
+        0.43229461743105685,
+        0.42189655355246203,
+        0.47989858564637067
+      ],
+      "teacher_dice_pool": [
+        0.45500621878025627,
+        0.14580300142111405,
+        0.18938505602311717,
+        0.2676734414185573,
+        0.1697405400963424,
+        0.4251377383150291,
+        0.4270884498370188,
+        0.5221131621647941
+      ],
+      "evidence_fg_pixels": 4939842,
+      "evidence_nonempty_patches": 289,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0779412985155291,
+        0.01622381303412828,
+        0.001478357728619292,
+        0.1667471831433008,
+        0.005309978119330455,
+        0.09491367972415017,
+        0.16990207946327957,
+        0.4674836102716623
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5283083464616156,
+        "uniform": 0.2575204836185174,
+        "top1": 0.47989858564637067,
+        "top2": 0.4829881803552951,
+        "top3": 0.5002578227349939
+      },
+      "pseudo_student_dice": 0.8026457077424155,
+      "pseudo_ensemble_dice": 0.46667988670073507,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6882328190174084,
+      "val_vi": 0.5241002137565222,
+      "val_ece": 0.029186931568365077
+    },
+    {
+      "round": 6,
+      "n_labeled": 432,
+      "selection_seconds": 0.2,
+      "round_seconds": 169.2,
+      "teacher_dice_labeled": [
+        0.39221785990090063,
+        0.15853969100727025,
+        0.1789619795127443,
+        0.3011489393039044,
+        0.18995564470764165,
+        0.4492783861051983,
+        0.4173859453694295,
+        0.45590431446624324
+      ],
+      "teacher_dice_pool": [
+        0.4051408302204626,
+        0.16061494373172766,
+        0.1831960622696714,
+        0.2969592804592786,
+        0.19929429611917968,
+        0.44823407067395027,
+        0.41566445570090216,
+        0.4808073840817319
+      ],
+      "evidence_fg_pixels": 5814272,
+      "evidence_nonempty_patches": 348,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.17503087706727477,
+        0.2527176996415204,
+        0.5722514232912048
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.5162106450575661,
+        "uniform": 0.27684872084241097,
+        "top1": 0.45590431446624324,
+        "top2": 0.46629336580424974,
+        "top3": 0.5313853044604864
+      },
+      "pseudo_student_dice": 0.8041609915056682,
+      "pseudo_ensemble_dice": 0.5110455493001167,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6883015407537225,
+      "val_vi": 0.5339455101163537,
+      "val_ece": 0.010485531571301124
+    },
+    {
+      "round": 7,
+      "n_labeled": 432,
+      "selection_seconds": 0.0,
+      "round_seconds": 95.4,
+      "final_consolidation": true,
+      "val_dice": 0.7122290178272843,
+      "val_vi": 0.515518646087249,
+      "val_ece": 0.0282238460581894
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.01683200768757291,
+    "auroc_top20_error": 0.5015251953125
+  },
+  "al_seconds": 986.1,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.17503087706727477,
+    0.2527176996415204,
+    0.5722514232912048
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.5722514232912048,
+    "entropy": 0.9720687572709196,
+    "effective_sources": 2.643407374909532,
+    "min_weight": 0.0
+  },
+  "top_source_final": "whole-mouse-brain",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7470388735427181,
+    "patch_iou": 0.6460663585703925,
+    "patch_precision": 0.7807578668019558,
+    "patch_recall": 0.8005368189549578,
+    "patch_vi": 0.42906332242828876,
+    "n_patches": 2883,
+    "cal_ece": 0.02143620715811456,
+    "cal_nll": 0.20047653371661475,
+    "cal_brier": 0.05796857954842853,
+    "cal_mean_confidence": 0.9434937833591204,
+    "cal_accuracy": 0.9220575762010059,
+    "cal_predicted_prevalence": 0.2271959276361429,
+    "cal_true_prevalence": 0.23013442215140478,
+    "cal_mean_probability": 0.23093329968843
+  },
+  "test_notta": {
+    "patch_dice": 0.7470388735427181,
+    "patch_iou": 0.6460663585703925,
+    "patch_precision": 0.7807578668019558,
+    "patch_recall": 0.8005368189549578,
+    "patch_vi": 0.42906332242828876,
+    "n_patches": 2883,
+    "cal_ece": 0.02143620715811456,
+    "cal_nll": 0.20047653371661475,
+    "cal_brier": 0.05796857954842853,
+    "cal_mean_confidence": 0.9434937833591204,
+    "cal_accuracy": 0.9220575762010059,
+    "cal_predicted_prevalence": 0.2271959276361429,
+    "cal_true_prevalence": 0.23013442215140478,
+    "cal_mean_probability": 0.23093329968843
+  },
+  "test": {
+    "patch_dice": 0.7783869137129839,
+    "patch_iou": 0.6820751619244556,
+    "patch_precision": 0.8232589978608481,
+    "patch_recall": 0.8022611640300222,
+    "patch_vi": 0.4045632298190714,
+    "n_patches": 2883,
+    "image_dice": 0.8282705378907229,
+    "image_iou": 0.7096447795902364,
+    "image_precision": 0.8550598096304546,
+    "image_recall": 0.8057972121205764,
+    "image_vi": 0.45746639872723177,
+    "n_images": 3,
+    "cal_ece": 0.008398856496640952,
+    "cal_nll": 0.17592274256698134,
+    "cal_brier": 0.05262819081835747,
+    "cal_mean_confidence": 0.9360410674965641,
+    "cal_accuracy": 0.927740472164412,
+    "cal_predicted_prevalence": 0.22152996824054805,
+    "cal_true_prevalence": 0.23013442215140478,
+    "cal_mean_probability": 0.2258376879442032
+  },
+  "eval_seconds": 44.7,
+  "total_seconds": 1406.8,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.22,
+        "peak_cpu_rss_mb": 1028.3,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 32.54,
+        "peak_cpu_rss_mb": 1801.3,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 339.71,
+        "peak_cpu_rss_mb": 3693.9,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 986.15,
+        "peak_cpu_rss_mb": 3693.9,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4642.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 44.7,
+        "peak_cpu_rss_mb": 3693.9,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4644.0,
+        "patches": 2883,
+        "images": 3,
+        "patches_per_second": 64.5,
+        "seconds_per_image": 14.9
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 3693.9,
+    "total_seconds": 1404.3
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-random_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-random/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,72,35.6,206.6,"[0.34699131258491106, 0.15400938239211243, 0.17063725881990668, 0.29518215496468825, 0.1907363491430849, 0.40286997571879657, 0.402560670886263, 0.4126936934594237]",,891480,58,True,kept,0.21571648537522917,0.4126936934594237,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.1573376001938336, 0.311542520327699, 0.5311198794784675]",top3,"{'mixture': 0.4486622376468838, 'uniform': 0.257123442242594, 'top1': 0.4126936934594237, 'top2': 0.427720538104064, 'top3': 0.5579440810273272}",0.587305604106643,0.6599535991066656,0.04835740491699966,,,,,
+2,144,0.2,43.8,"[0.3877112948623053, 0.14909626611196175, 0.17005234419964096, 0.29110798999593296, 0.17764307442922855, 0.41866300287842756, 0.40459903528437674, 0.4492049901441661]",,2038859,117,False,,,,"[0.06909489706776745, 0.019479382824218353, 0.0015792762266002333, 0.21586754477923684, 0.006593754437182248, 0.09993372841912157, 0.17244918911249582, 0.4150022271333776]",mixture,"{'mixture': 0.5114063614291914, 'uniform': 0.2542809724207618, 'top1': 0.4492049901441661, 'top2': 0.4573143432435651, 'top3': 0.4917913428812348}",0.6585940233579239,0.5740162973538123,0.005417245034970651,0.7723456432892528,0.4989113809299943,student,0.908,
+3,216,0.2,142.6,"[0.4140682111607643, 0.1536399927755801, 0.1799205572303296, 0.29112126324332566, 0.18334845976830494, 0.4423851087364882, 0.41803091046534124, 0.4848902568732807]","[0.4416778556160547, 0.15487823535765688, 0.18914476821851361, 0.28655053365525646, 0.18583099402223757, 0.4567643200946957, 0.43939763997607106, 0.5156445163650382]",3013404,174,False,,,,"[0.0718401217611655, 0.016284063956197966, 0.001392418083738922, 0.17246439507692587, 0.0055765473665446265, 0.10120229003831627, 0.157561280540448, 0.4736788831766628]",mixture,"{'mixture': 0.5454374977215515, 'uniform': 0.2664851046953125, 'top1': 0.4848902568732807, 'top2': 0.49108429525345076, 'top3': 0.5097392015266642}",0.6488990698124532,0.5500686787278544,0.008365029822864706,0.7999457618489225,0.5073909859154929,student,1.0,
+4,288,0.2,151.8,"[0.38945019463626285, 0.1679853583375917, 0.18044124015986018, 0.32282208176596117, 0.2034846994902197, 0.4660624113903503, 0.41870020084547344, 0.4503311467877327]","[0.3928618027754761, 0.18289820943264143, 0.1883542828074731, 0.3437577001051289, 0.2306743626695978, 0.4966166642733409, 0.4197338102857259, 0.45470006684976394]",4032333,233,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.2061290137522376, 0.25496217742297783, 0.5389088088247845]",top3,"{'mixture': 0.4927856038827534, 'uniform': 0.29765161943165275, 'top1': 0.4660624113903503, 'top2': 0.4691393882366579, 'top3': 0.5775226351451215}",0.658762843387845,0.5911395322997408,0.012483891875880187,0.7826412650210633,0.5706861006865913,student,0.942,
+5,360,0.2,161.3,"[0.41854435812646784, 0.1495941701206981, 0.1822344165135339, 0.28407492428974873, 0.17477511072235818, 0.43229461743105685, 0.42189655355246203, 0.47989858564637067]","[0.45500621878025627, 0.14580300142111405, 0.18938505602311717, 0.2676734414185573, 0.1697405400963424, 0.4251377383150291, 0.4270884498370188, 0.5221131621647941]",4939842,289,False,,,,"[0.0779412985155291, 0.01622381303412828, 0.001478357728619292, 0.1667471831433008, 0.005309978119330455, 0.09491367972415017, 0.16990207946327957, 0.4674836102716623]",mixture,"{'mixture': 0.5283083464616156, 'uniform': 0.2575204836185174, 'top1': 0.47989858564637067, 'top2': 0.4829881803552951, 'top3': 0.5002578227349939}",0.6882328190174084,0.5241002137565222,0.029186931568365077,0.8026457077424155,0.46667988670073507,student,1.0,
+6,432,0.2,169.2,"[0.39221785990090063, 0.15853969100727025, 0.1789619795127443, 0.3011489393039044, 0.18995564470764165, 0.4492783861051983, 0.4173859453694295, 0.45590431446624324]","[0.4051408302204626, 0.16061494373172766, 0.1831960622696714, 0.2969592804592786, 0.19929429611917968, 0.44823407067395027, 0.41566445570090216, 0.4808073840817319]",5814272,348,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.17503087706727477, 0.2527176996415204, 0.5722514232912048]",top3,"{'mixture': 0.5162106450575661, 'uniform': 0.27684872084241097, 'top1': 0.45590431446624324, 'top2': 0.46629336580424974, 'top3': 0.5313853044604864}",0.6883015407537225,0.5339455101163537,0.010485531571301124,0.8041609915056682,0.5110455493001167,student,1.0,
+7,432,0.0,95.4,,,,,,,,,,,,0.7122290178272843,0.515518646087249,0.0282238460581894,,,,,True

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-2/log.txt

@@ -0,0 +1,107 @@
+[06:56:31][      0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-entropy budget=5.0% seed=2 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-entropy", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "entropy", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[06:56:31][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[06:56:31][      0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches
+[06:56:33][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, snemi=3.55, mouse-brain=4.08
+[06:56:33][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[06:56:39][      7s][WEIGHTS] mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031 | eff.sources=3.41 | MMD2 mix=0.0128 uniform=0.0970
+[06:56:39][      7s][WEIGHTS] per-source: c-elegans=0.320, cerebellum=0.017, flywire=0.001, berghia=0.175, microns=0.027, oct-glia=0.436, snemi=0.017, mouse-brain=0.007 | min-MMD source (NeuroADDA rule) = oct-glia
+[06:56:57][     26s][TEACHERS] target-val Dice: c-elegans=0.476, cerebellum=0.633, flywire=0.142, berghia=0.467, microns=0.708, oct-glia=0.845, snemi=0.229, mouse-brain=0.115
+[06:56:57][     26s][TEACHERS] ensemble target-val Dice: mixture=0.849, mixture@raw=0.847, mixture@logit=0.850, per_source=0.843, per_source@raw=0.619, per_source@logit=0.848, uniform=0.718, uniform@raw=0.677, uniform@logit=0.807
+[06:56:57][     26s][WEIGHTS] using mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031
+[06:56:57][     26s][INIT] student initialised from teacher oct-glia (weight 0.636)
+[06:56:58][     27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[06:56:58][     27s][KD] distilling the mixture ensemble (prob space) into the student
+[06:57:39][     68s][KD] soft labels for 3584 patches from 8 teachers in 41s
+[06:58:12][    101s][KD] epoch 1/5 | patches=3584 | loss=0.6002 | 33s
+[06:58:45][    133s][KD] epoch 2/5 | patches=3584 | loss=0.4596 | 33s
+[06:59:17][    166s][KD] epoch 3/5 | patches=3584 | loss=0.4055 | 33s
+[06:59:50][    199s][KD] epoch 4/5 | patches=3584 | loss=0.3664 | 33s
+[07:00:23][    232s][KD] epoch 5/5 | patches=3584 | loss=0.3377 | 33s
+[07:00:24][    233s][ROUND0] before any target label: val_dice=0.8442 val_vi=0.1463
+[07:00:24][    233s][AL] patches=3584 (pool 2338 after dropping blank; free blank negatives: True) | budget=179 (5.0%) | rounds=6 | per_round=29 | acquisition=entropy | first_round=stratified
+[07:00:28][    237s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.4491343304947581, 'auroc_top20_error': 0.6088995155171626}
+[07:00:51][    260s][AL] round 1: stratified cold start: 15 patches from the membrane-rich half (ensemble prevalence >= 0.192) + 14 from the membrane-poor half of 2000 candidates
+[07:00:54][    263s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.765, uniform=0.741, top1=0.765, top2=0.764, top3=0.765 -> using mixture
+[07:00:55][    263s][WEIGHTS] round 1: student kept (labeled Dice 0.748 vs best teacher oct-glia 0.765); distilling the corrected ensemble into it
+[07:00:55][    263s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (179039 fg px): c-elegans=0.257, cerebellum=0.569, flywire=0.169, berghia=0.276, microns=0.581, oct-glia=0.765, snemi=0.110, mouse-brain=0.203 | weights: c-elegans=0.001, cerebellum=0.006, flywire=0.000, berghia=0.002, microns=0.008, oct-glia=0.983, snemi=0.000, mouse-brain=0.000
+[07:01:21][    290s][KD] soft labels for 2309 patches from 8 teachers in 26s
+[07:01:42][    311s][KD] epoch 1/2 | patches=2309 | loss=0.1395 | 21s
+[07:02:03][    332s][KD] epoch 2/2 | patches=2309 | loss=0.1332 | 21s
+[07:02:26][    355s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[07:02:28][    356s][FT] round 1/6 epoch 1/5 | patches=58 | loss=0.8560
+[07:02:29][    358s][FT] round 1/6 epoch 2/5 | patches=58 | loss=0.5224
+[07:02:30][    359s][FT] round 1/6 epoch 3/5 | patches=58 | loss=0.4384
+[07:02:32][    361s][FT] round 1/6 epoch 4/5 | patches=58 | loss=0.3711
+[07:02:33][    362s][FT] round 1/6 epoch 5/5 | patches=58 | loss=0.3721
+[07:02:35][    364s][AL] round 1/6 | labeled=29 | selection_seconds=26.9000 round_seconds=129.3000 restart_student_dice=0.7483 restart_teacher_dice=0.7648 val_dice=0.8506 val_vi=0.1502 val_ece=0.0197
+[07:02:39][    367s][AL] round 2: 595/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:02:44][    373s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.578, uniform=0.590, top1=0.578, top2=0.577, top3=0.577 -> using uniform
+[07:02:44][    373s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (666295 fg px): c-elegans=0.305, cerebellum=0.438, flywire=0.166, berghia=0.275, microns=0.457, oct-glia=0.578, snemi=0.187, mouse-brain=0.290 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[07:02:47][    376s][CONS] round 2: on the 29 new patches student Dice=0.527 vs ensemble 0.520 -> pseudo-labels from the ensemble | consistency x1.00
+[07:02:49][    378s][FT] round 2/6 epoch 1/5 | patches=116 | loss=0.9634
+[07:02:52][    381s][FT] round 2/6 epoch 2/5 | patches=116 | loss=0.8654
+[07:02:54][    383s][FT] round 2/6 epoch 3/5 | patches=116 | loss=0.8580
+[07:02:56][    385s][FT] round 2/6 epoch 4/5 | patches=116 | loss=0.7631
+[07:02:59][    388s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.6561
+[07:03:00][    389s][AL] round 2/6 | labeled=58 | selection_seconds=3.7000 round_seconds=23.9000 pseudo_student_dice=0.5266 pseudo_ensemble_dice=0.5199 consistency_scale=1.0000 val_dice=0.8430 val_vi=0.1733 val_ece=0.0119
+[07:03:04][    393s][AL] round 3: 550/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:03:13][    401s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.425, uniform=0.447, top1=0.423, top2=0.424, top3=0.423 -> using uniform
+[07:03:13][    401s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (1340584 fg px): c-elegans=0.257, cerebellum=0.358, flywire=0.144, berghia=0.219, microns=0.340, oct-glia=0.423, snemi=0.249, mouse-brain=0.287 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[07:03:18][    407s][CONS] round 3: on the 29 new patches student Dice=0.484 vs ensemble 0.251 -> pseudo-labels from the student | consistency x0.00
+[07:03:20][    409s][FT] round 3/6 epoch 1/5 | patches=174 | loss=0.8398
+[07:03:22][    411s][FT] round 3/6 epoch 2/5 | patches=174 | loss=0.7916
+[07:03:24][    413s][FT] round 3/6 epoch 3/5 | patches=174 | loss=1.0183
+[07:03:26][    414s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.8409
+[07:03:27][    416s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.8439
+[07:03:29][    418s][AL] round 3/6 | labeled=87 | selection_seconds=3.9000 round_seconds=26.9000 pseudo_student_dice=0.4843 pseudo_ensemble_dice=0.2510 consistency_scale=0.0000 val_dice=0.7727 val_vi=0.1726 val_ece=0.0198
+[07:03:32][    421s][AL] round 4: 574/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:04:24][    473s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.286, cerebellum=0.521, flywire=0.234, berghia=0.315, microns=0.554, oct-glia=0.793, snemi=0.220, mouse-brain=0.190 | blended with labeled Dice at 0.50
+[07:05:13][    522s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.630, uniform=0.587, top1=0.630, top2=0.630, top3=0.630 -> using mixture
+[07:05:13][    522s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1701624 fg px): c-elegans=0.294, cerebellum=0.456, flywire=0.192, berghia=0.285, microns=0.465, oct-glia=0.630, snemi=0.225, mouse-brain=0.248 | weights: c-elegans=0.005, cerebellum=0.008, flywire=0.000, berghia=0.007, microns=0.009, oct-glia=0.969, snemi=0.001, mouse-brain=0.001
+[07:05:19][    527s][CONS] round 4: on the 29 new patches student Dice=0.674 vs ensemble 0.607 -> pseudo-labels from the student | consistency x0.58
+[07:05:23][    532s][FT] round 4/6 epoch 1/5 | patches=232 | loss=0.6445
+[07:05:27][    536s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.6586
+[07:05:32][    540s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.5856
+[07:05:36][    545s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.5562
+[07:05:40][    549s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.4736
+[07:05:42][    551s][AL] round 4/6 | labeled=116 | selection_seconds=3.8000 round_seconds=131.4000 pseudo_student_dice=0.6742 pseudo_ensemble_dice=0.6070 consistency_scale=0.5810 val_dice=0.8448 val_vi=0.1761 val_ece=0.0075
+[07:05:45][    554s][AL] round 5: 536/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:06:38][    607s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.409, cerebellum=0.640, flywire=0.238, berghia=0.380, microns=0.605, oct-glia=0.820, snemi=0.314, mouse-brain=0.283 | blended with labeled Dice at 0.50
+[07:07:28][    657s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.657, uniform=0.657, top1=0.656, top2=0.657, top3=0.656 -> using mixture
+[07:07:28][    657s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1973658 fg px): c-elegans=0.353, cerebellum=0.514, flywire=0.195, berghia=0.320, microns=0.494, oct-glia=0.656, snemi=0.263, mouse-brain=0.290 | weights: c-elegans=0.007, cerebellum=0.010, flywire=0.000, berghia=0.008, microns=0.009, oct-glia=0.963, snemi=0.001, mouse-brain=0.001
+[07:07:34][    663s][CONS] round 5: on the 29 new patches student Dice=0.689 vs ensemble 0.631 -> pseudo-labels from the student | consistency x0.63
+[07:07:39][    668s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.4823
+[07:07:44][    673s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.4024
+[07:07:50][    679s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.4317
+[07:07:55][    684s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.3859
+[07:08:00][    689s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.4092
+[07:08:02][    691s][AL] round 5/6 | labeled=145 | selection_seconds=3.8000 round_seconds=138.5000 pseudo_student_dice=0.6891 pseudo_ensemble_dice=0.6310 consistency_scale=0.6300 val_dice=0.5182 val_vi=0.1551 val_ece=0.0061
+[07:08:05][    694s][AL] round 6: 551/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:08:59][    748s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.376, cerebellum=0.625, flywire=0.243, berghia=0.368, microns=0.609, oct-glia=0.861, snemi=0.290, mouse-brain=0.253 | blended with labeled Dice at 0.50
+[07:09:51][    799s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.682, uniform=0.662, top1=0.682, top2=0.682, top3=0.682 -> using mixture
+[07:09:51][    799s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2295762 fg px): c-elegans=0.338, cerebellum=0.514, flywire=0.198, berghia=0.319, microns=0.501, oct-glia=0.682, snemi=0.248, mouse-brain=0.275 | weights: c-elegans=0.005, cerebellum=0.008, flywire=0.000, berghia=0.006, microns=0.008, oct-glia=0.972, snemi=0.000, mouse-brain=0.001
+[07:09:56][    805s][CONS] round 6: on the 29 new patches student Dice=0.729 vs ensemble 0.557 -> pseudo-labels from the student | consistency x0.76
+[07:10:02][    811s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.5120
+[07:10:09][    817s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.4880
+[07:10:15][    824s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.4572
+[07:10:21][    830s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.4679
+[07:10:27][    836s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.4644
+[07:10:29][    838s][AL] round 6/6 | labeled=174 | selection_seconds=3.6000 round_seconds=145.5000 pseudo_student_dice=0.7290 pseudo_ensemble_dice=0.5568 consistency_scale=0.7630 val_dice=0.8730 val_vi=0.1522 val_ece=0.0088
+[07:10:32][    841s][CONS] final consolidation: pseudo-labels refreshed from the student
+[07:10:38][    847s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.3857
+[07:10:45][    853s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.3635
+[07:10:51][    860s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.3861
+[07:10:57][    866s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.3872
+[07:11:03][    872s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.3739
+[07:11:10][    879s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.3606
+[07:11:16][    885s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.3436
+[07:11:23][    891s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.3332
+[07:11:29][    898s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.3277
+[07:11:35][    904s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.3268
+[07:11:37][    906s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=66.2000 val_dice=0.8848 val_vi=0.1468 val_ece=0.0087
+[07:11:37][    906s][WEIGHTS] final: c-elegans=0.005, cerebellum=0.008, flywire=0.000, berghia=0.006, microns=0.008, oct-glia=0.972, snemi=0.000, mouse-brain=0.001
+[07:11:47][    916s][TEST] raw student (no EMA, no TTA): patch_dice=0.8539 patch_vi=0.1791
+[07:11:49][    918s][TEST] final model without TTA: patch_dice=0.8539 patch_vi=0.1791
+[07:11:49][    918s][RESOURCES] {"phases": {"setup": {"seconds": 1.13, "peak_cpu_rss_mb": 1027.8, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 25.8, "peak_cpu_rss_mb": 1809.7, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 204.46, "peak_cpu_rss_mb": 2938.7, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 672.49, "peak_cpu_rss_mb": 2938.7, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 12.28, "peak_cpu_rss_mb": 2938.7, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 768, "images": 6, "patches_per_second": 62.5, "seconds_per_image": 2.047}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2938.7, "total_seconds": 916.2}
+[07:11:49][    918s][TEST] patch_dice=0.8592 patch_iou=0.7973 patch_precision=0.8997 patch_recall=0.8627 patch_vi=0.1753 image_dice=0.7942 image_iou=0.6669 image_precision=0.7997 image_recall=0.7956 image_vi=0.1990 cal_ece=0.0074 cal_nll=0.0770 cal_brier=0.0223 cal_mean_confidence=0.9774 cal_accuracy=0.9701 cal_predicted_prevalence=0.0725 cal_true_prevalence=0.0718 cal_mean_probability=0.0736
+[07:11:49][    918s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-entropy b5 s2 | test patch Dice=0.8592 VI=0.1753 | image Dice=0.7942 VI=0.1990 | 918s

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-2/result.json

@@ -0,0 +1,776 @@
+{
+  "target": "octopus-vulgaris-vertical-lobe-sfltract",
+  "variant": "macs-v11-norestart-acq-entropy",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "octopus-vulgaris-vertical-lobe-sfltract",
+    "variant": "macs-v11-norestart-acq-entropy",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "entropy",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "--variants",
+      "macs-v11-norestart-acq-entropy",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.09888773316940314,
+    0.02852238358730045,
+    0.006877111295002701,
+    0.14762496137410624,
+    0.031629415990885375,
+    0.6364113152878205,
+    0.01903121486130303,
+    0.03101586443417853
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.09888773316940314,
+      0.02852238358730045,
+      0.006877111295002701,
+      0.14762496137410624,
+      0.031629415990885375,
+      0.6364113152878205,
+      0.01903121486130303,
+      0.03101586443417853
+    ],
+    "per_source": [
+      0.3196472209911767,
+      0.016524281783989066,
+      0.0007715704214095669,
+      0.17539935780265475,
+      0.02748152710843107,
+      0.4363201963447303,
+      0.016643455180579362,
+      0.00721239036702913
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.00048684441851293454,
+      "mmd2_mixture": 0.01275567083190321,
+      "mmd2_uniform": 0.09698324592915836,
+      "mmd2_single_source": [
+        0.07244541101040552,
+        0.256897395483041,
+        0.3929088032810887,
+        0.05287172109384486,
+        0.26498470596000767,
+        0.011934168264888134,
+        0.339258942897097,
+        0.2684433232010991
+      ],
+      "median_squared_distance": 0.18344226468067681,
+      "gammas": [
+        1.3628266116054668,
+        2.7256532232109336,
+        5.451306446421867,
+        10.902612892843734,
+        21.80522578568747
+      ],
+      "max_weight": 0.6364113152878205,
+      "entropy": 1.2268826413390275,
+      "effective_sources": 3.410580943129122,
+      "min_weight": 0.006877111295002701
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.0496200334528889,
+        0.42663682749168497,
+        0.8166058321365515,
+        0.12600018979150718,
+        0.3618977926472007,
+        0.01001955344998784,
+        0.4257222632020228,
+        0.5321458020642489
+      ],
+      "standardized": [
+        -1.1548905481172125,
+        0.3263029673133632,
+        1.8583820949064265,
+        -0.8548142807860197,
+        0.07196140979373515,
+        -1.3104697560349354,
+        0.32270990006354566,
+        0.7408182128610976
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4363201963447303,
+      "entropy": 1.3076016834873658,
+      "effective_sources": 3.6972957848927845,
+      "min_weight": 0.0007715704214095669
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4757138324136084,
+      0.6329416506363946,
+      0.14191107641715917,
+      0.4669227387129098,
+      0.7079858630413108,
+      0.8450575342195966,
+      0.2290226697023551,
+      0.11473197497037783
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.8487996328814233,
+      "mixture@raw": 0.8468783370062004,
+      "mixture@logit": 0.8498165961112344,
+      "per_source": 0.8430704623597217,
+      "per_source@raw": 0.6192918885999461,
+      "per_source@logit": 0.8479512470904803,
+      "uniform": 0.7183095932451996,
+      "uniform@raw": 0.6771602818911759,
+      "uniform@logit": 0.8065076510127938
+    },
+    "val_patches": 768
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "weight_stats": {
+    "max_weight": 0.6364113152878205,
+    "entropy": 1.2268826413390275,
+    "effective_sources": 3.410580943129122,
+    "min_weight": 0.006877111295002701
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.5952380952380953,
+    "spearman_weight_taxonomy": 0.6257415691925263,
+    "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "student_init": "top-source",
+  "blank_fraction": 0.34765625,
+  "kd_seconds": 204.5,
+  "round0_val": {
+    "val_dice": 0.844188048800421,
+    "val_vi": 0.14634638944976133,
+    "val_ece": 0.06461472771169184
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 29,
+      "selection_seconds": 26.9,
+      "round_seconds": 129.3,
+      "teacher_dice_labeled": [
+        0.2571777942968238,
+        0.5690605504275248,
+        0.1686588604013407,
+        0.2758037194420113,
+        0.5811237001396865,
+        0.7648114620563284,
+        0.10985703762520847,
+        0.20342659662893037
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 179039,
+      "evidence_nonempty_patches": 21,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.7483403957093632,
+      "restart_teacher_dice": 0.7648114620563284,
+      "weights": [
+        0.0009536063313568787,
+        0.00622160049410085,
+        2.7365280978604556e-05,
+        0.0017150542671718894,
+        0.00778389682673956,
+        0.9830816831898324,
+        4.206174632514249e-05,
+        0.0001747318634946766
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7646874735949377,
+        "uniform": 0.7405359432386459,
+        "top1": 0.7648114620563284,
+        "top2": 0.7643850371191413,
+        "top3": 0.7646318754401064
+      },
+      "val_dice": 0.850588566843053,
+      "val_vi": 0.15020738136525003,
+      "val_ece": 0.019654480413019926
+    },
+    {
+      "round": 2,
+      "n_labeled": 58,
+      "selection_seconds": 3.7,
+      "round_seconds": 23.9,
+      "teacher_dice_labeled": [
+        0.3051898457957521,
+        0.43829723274180427,
+        0.1662645122120252,
+        0.2748940361871823,
+        0.4567545091150516,
+        0.5777898396180228,
+        0.18661986847273826,
+        0.28953794592991694
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 666295,
+      "evidence_nonempty_patches": 49,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.5780045228581705,
+        "uniform": 0.5900107461950028,
+        "top1": 0.5777898396180228,
+        "top2": 0.5769733017351059,
+        "top3": 0.5772074201516861
+      },
+      "pseudo_student_dice": 0.5266264913726709,
+      "pseudo_ensemble_dice": 0.5198939717868823,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.8430299891923164,
+      "val_vi": 0.17334322904515395,
+      "val_ece": 0.011880019773361832
+    },
+    {
+      "round": 3,
+      "n_labeled": 87,
+      "selection_seconds": 3.9,
+      "round_seconds": 26.9,
+      "teacher_dice_labeled": [
+        0.256554203215484,
+        0.3584625882888147,
+        0.14446923270617784,
+        0.21940902140115776,
+        0.3404308510983355,
+        0.423359355028126,
+        0.24896099647764894,
+        0.28709608031045697
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1340584,
+      "evidence_nonempty_patches": 77,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.42523293426148684,
+        "uniform": 0.4468307673264077,
+        "top1": 0.423359355028126,
+        "top2": 0.42387354326914495,
+        "top3": 0.42262162899245326
+      },
+      "pseudo_student_dice": 0.48432161207532876,
+      "pseudo_ensemble_dice": 0.2509745273916862,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.7727355398927459,
+      "val_vi": 0.17261770973435211,
+      "val_ece": 0.019825972598603598
+    },
+    {
+      "round": 4,
+      "n_labeled": 116,
+      "selection_seconds": 3.8,
+      "round_seconds": 131.4,
+      "teacher_dice_labeled": [
+        0.2936284700962486,
+        0.4555138797855507,
+        0.1916935869025312,
+        0.28483595860735256,
+        0.4645854181906192,
+        0.6302061070335462,
+        0.22524482639783452,
+        0.24801376422596666
+      ],
+      "teacher_dice_pool": [
+        0.2856535313876258,
+        0.5207219061423793,
+        0.2336147684929623,
+        0.3149336309940731,
+        0.5542851398624332,
+        0.7928285164628092,
+        0.21995953296421208,
+        0.1900572617220739
+      ],
+      "evidence_fg_pixels": 1701624,
+      "evidence_nonempty_patches": 105,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.005201196091910545,
+        0.007571920116834874,
+        0.000130517662110731,
+        0.007111074208082487,
+        0.009194086198533695,
+        0.9692522174719312,
+        0.0005051739265953853,
+        0.0010338143240010826
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6304836448276889,
+        "uniform": 0.5866893800713544,
+        "top1": 0.6302061070335462,
+        "top2": 0.6301844454387036,
+        "top3": 0.6303030425059223
+      },
+      "pseudo_student_dice": 0.6741909994732245,
+      "pseudo_ensemble_dice": 0.6069983324752639,
+      "pseudo_source": "student",
+      "consistency_scale": 0.581,
+      "val_dice": 0.8447836167150625,
+      "val_vi": 0.17608145575249626,
+      "val_ece": 0.007456402815686945
+    },
+    {
+      "round": 5,
+      "n_labeled": 145,
+      "selection_seconds": 3.8,
+      "round_seconds": 138.5,
+      "teacher_dice_labeled": [
+        0.3525403253235777,
+        0.5143178603488624,
+        0.19521558185516774,
+        0.32040272699958583,
+        0.4937034634276607,
+        0.6562794950132421,
+        0.26321708956655626,
+        0.28986782022526414
+      ],
+      "teacher_dice_pool": [
+        0.4086587040387924,
+        0.6404567254485388,
+        0.2383637366387085,
+        0.38032477471780685,
+        0.6049985309865628,
+        0.8196890401954791,
+        0.3140789070656653,
+        0.28254675899248766
+      ],
+      "evidence_fg_pixels": 1973658,
+      "evidence_nonempty_patches": 133,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.007178796284775949,
+        0.010439649829683862,
+        0.00010352913395009788,
+        0.007771362881353382,
+        0.009420286261216514,
+        0.9633177306284751,
+        0.0005655224320038362,
+        0.001203122548541244
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6567929858161429,
+        "uniform": 0.656789974783481,
+        "top1": 0.6562794950132421,
+        "top2": 0.6567521716447765,
+        "top3": 0.6561504363680574
+      },
+      "pseudo_student_dice": 0.6890727445515457,
+      "pseudo_ensemble_dice": 0.6309668156812759,
+      "pseudo_source": "student",
+      "consistency_scale": 0.63,
+      "val_dice": 0.5182098282101325,
+      "val_vi": 0.15514475495746,
+      "val_ece": 0.006062149701860885
+    },
+    {
+      "round": 6,
+      "n_labeled": 174,
+      "selection_seconds": 3.6,
+      "round_seconds": 145.5,
+      "teacher_dice_labeled": [
+        0.337566897659429,
+        0.5141171511256178,
+        0.19790145540894027,
+        0.31851610101400474,
+        0.5005018480617079,
+        0.6817950271148931,
+        0.24806949112657262,
+        0.27509876690905666
+      ],
+      "teacher_dice_pool": [
+        0.3757462025226483,
+        0.6248672855702827,
+        0.2432417622493409,
+        0.3677013462721566,
+        0.6094810118682077,
+        0.8612422762866326,
+        0.28957574781839723,
+        0.2533942182588554
+      ],
+      "evidence_fg_pixels": 2295762,
+      "evidence_nonempty_patches": 158,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.004831314409281969,
+        0.008144355165130322,
+        8.313226695241612e-05,
+        0.00596137055309223,
+        0.007881911449567177,
+        0.9719066277004534,
+        0.00037993306747879105,
+        0.0008113553880436958
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6821446395117668,
+        "uniform": 0.6616735807081089,
+        "top1": 0.6817950271148931,
+        "top2": 0.6820726476890172,
+        "top3": 0.6817917541113985
+      },
+      "pseudo_student_dice": 0.7289755472355686,
+      "pseudo_ensemble_dice": 0.5567674137146332,
+      "pseudo_source": "student",
+      "consistency_scale": 0.763,
+      "val_dice": 0.872970073630813,
+      "val_vi": 0.1522411484043612,
+      "val_ece": 0.008795662186711174
+    },
+    {
+      "round": 7,
+      "n_labeled": 174,
+      "selection_seconds": 0.0,
+      "round_seconds": 66.2,
+      "final_consolidation": true,
+      "val_dice": 0.8847641017050304,
+      "val_vi": 0.14680541323636923,
+      "val_ece": 0.008719392809047056
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.4491343304947581,
+    "auroc_top20_error": 0.6088995155171626
+  },
+  "al_seconds": 672.5,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.004831314409281969,
+    0.008144355165130322,
+    8.313226695241612e-05,
+    0.00596137055309223,
+    0.007881911449567177,
+    0.9719066277004534,
+    0.00037993306747879105,
+    0.0008113553880436958
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9719066277004534,
+    "entropy": 0.17089436807831188,
+    "effective_sources": 1.1863654243430872,
+    "min_weight": 8.313226695241612e-05
+  },
+  "top_source_final": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.8539403069986387,
+    "patch_iou": 0.7897468209479852,
+    "patch_precision": 0.8916481287479515,
+    "patch_recall": 0.8602379909823856,
+    "patch_vi": 0.17912111520513588,
+    "n_patches": 768,
+    "cal_ece": 0.010201323030848362,
+    "cal_nll": 0.08445373201829658,
+    "cal_brier": 0.023090665519235807,
+    "cal_mean_confidence": 0.9794432489222348,
+    "cal_accuracy": 0.9692433675130209,
+    "cal_predicted_prevalence": 0.07187652587890625,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.0721161566746557
+  },
+  "test_notta": {
+    "patch_dice": 0.8539403069986387,
+    "patch_iou": 0.7897468209479852,
+    "patch_precision": 0.8916481287479515,
+    "patch_recall": 0.8602379909823856,
+    "patch_vi": 0.17912111520513588,
+    "n_patches": 768,
+    "cal_ece": 0.010201323030848362,
+    "cal_nll": 0.08445373201829658,
+    "cal_brier": 0.023090665519235807,
+    "cal_mean_confidence": 0.9794432489222348,
+    "cal_accuracy": 0.9692433675130209,
+    "cal_predicted_prevalence": 0.07187652587890625,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.0721161566746557
+  },
+  "test": {
+    "patch_dice": 0.8591772266029952,
+    "patch_iou": 0.7973100976285888,
+    "patch_precision": 0.8997360376998985,
+    "patch_recall": 0.8626797510061494,
+    "patch_vi": 0.17529183800507697,
+    "n_patches": 768,
+    "image_dice": 0.794234595546598,
+    "image_iou": 0.6669425716713131,
+    "image_precision": 0.7997486828309645,
+    "image_recall": 0.7956493079051613,
+    "image_vi": 0.19896338686997997,
+    "n_images": 6,
+    "cal_ece": 0.0073982110200828165,
+    "cal_nll": 0.07701464436765242,
+    "cal_brier": 0.022278359769095674,
+    "cal_mean_confidence": 0.9774351348043228,
+    "cal_accuracy": 0.9700902303059896,
+    "cal_predicted_prevalence": 0.07253519694010417,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07355339704691839
+  },
+  "eval_seconds": 12.3,
+  "total_seconds": 917.8,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.13,
+        "peak_cpu_rss_mb": 1027.8,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 25.8,
+        "peak_cpu_rss_mb": 1809.7,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 204.46,
+        "peak_cpu_rss_mb": 2938.7,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 672.49,
+        "peak_cpu_rss_mb": 2938.7,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4626.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 12.28,
+        "peak_cpu_rss_mb": 2938.7,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4626.0,
+        "patches": 768,
+        "images": 6,
+        "patches_per_second": 62.5,
+        "seconds_per_image": 2.047
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2938.7,
+    "total_seconds": 916.2
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,29,26.9,129.3,"[0.2571777942968238, 0.5690605504275248, 0.1686588604013407, 0.2758037194420113, 0.5811237001396865, 0.7648114620563284, 0.10985703762520847, 0.20342659662893037]",,179039,21,True,kept,0.7483403957093632,0.7648114620563284,"[0.0009536063313568787, 0.00622160049410085, 2.7365280978604556e-05, 0.0017150542671718894, 0.00778389682673956, 0.9830816831898324, 4.206174632514249e-05, 0.0001747318634946766]",mixture,"{'mixture': 0.7646874735949377, 'uniform': 0.7405359432386459, 'top1': 0.7648114620563284, 'top2': 0.7643850371191413, 'top3': 0.7646318754401064}",0.850588566843053,0.15020738136525003,0.019654480413019926,,,,,
+2,58,3.7,23.9,"[0.3051898457957521, 0.43829723274180427, 0.1662645122120252, 0.2748940361871823, 0.4567545091150516, 0.5777898396180228, 0.18661986847273826, 0.28953794592991694]",,666295,49,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.5780045228581705, 'uniform': 0.5900107461950028, 'top1': 0.5777898396180228, 'top2': 0.5769733017351059, 'top3': 0.5772074201516861}",0.8430299891923164,0.17334322904515395,0.011880019773361832,0.5266264913726709,0.5198939717868823,ensemble,1.0,
+3,87,3.9,26.9,"[0.256554203215484, 0.3584625882888147, 0.14446923270617784, 0.21940902140115776, 0.3404308510983355, 0.423359355028126, 0.24896099647764894, 0.28709608031045697]",,1340584,77,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.42523293426148684, 'uniform': 0.4468307673264077, 'top1': 0.423359355028126, 'top2': 0.42387354326914495, 'top3': 0.42262162899245326}",0.7727355398927459,0.17261770973435211,0.019825972598603598,0.48432161207532876,0.2509745273916862,student,0.0,
+4,116,3.8,131.4,"[0.2936284700962486, 0.4555138797855507, 0.1916935869025312, 0.28483595860735256, 0.4645854181906192, 0.6302061070335462, 0.22524482639783452, 0.24801376422596666]","[0.2856535313876258, 0.5207219061423793, 0.2336147684929623, 0.3149336309940731, 0.5542851398624332, 0.7928285164628092, 0.21995953296421208, 0.1900572617220739]",1701624,105,False,,,,"[0.005201196091910545, 0.007571920116834874, 0.000130517662110731, 0.007111074208082487, 0.009194086198533695, 0.9692522174719312, 0.0005051739265953853, 0.0010338143240010826]",mixture,"{'mixture': 0.6304836448276889, 'uniform': 0.5866893800713544, 'top1': 0.6302061070335462, 'top2': 0.6301844454387036, 'top3': 0.6303030425059223}",0.8447836167150625,0.17608145575249626,0.007456402815686945,0.6741909994732245,0.6069983324752639,student,0.581,
+5,145,3.8,138.5,"[0.3525403253235777, 0.5143178603488624, 0.19521558185516774, 0.32040272699958583, 0.4937034634276607, 0.6562794950132421, 0.26321708956655626, 0.28986782022526414]","[0.4086587040387924, 0.6404567254485388, 0.2383637366387085, 0.38032477471780685, 0.6049985309865628, 0.8196890401954791, 0.3140789070656653, 0.28254675899248766]",1973658,133,False,,,,"[0.007178796284775949, 0.010439649829683862, 0.00010352913395009788, 0.007771362881353382, 0.009420286261216514, 0.9633177306284751, 0.0005655224320038362, 0.001203122548541244]",mixture,"{'mixture': 0.6567929858161429, 'uniform': 0.656789974783481, 'top1': 0.6562794950132421, 'top2': 0.6567521716447765, 'top3': 0.6561504363680574}",0.5182098282101325,0.15514475495746,0.006062149701860885,0.6890727445515457,0.6309668156812759,student,0.63,
+6,174,3.6,145.5,"[0.337566897659429, 0.5141171511256178, 0.19790145540894027, 0.31851610101400474, 0.5005018480617079, 0.6817950271148931, 0.24806949112657262, 0.27509876690905666]","[0.3757462025226483, 0.6248672855702827, 0.2432417622493409, 0.3677013462721566, 0.6094810118682077, 0.8612422762866326, 0.28957574781839723, 0.2533942182588554]",2295762,158,False,,,,"[0.004831314409281969, 0.008144355165130322, 8.313226695241612e-05, 0.00596137055309223, 0.007881911449567177, 0.9719066277004534, 0.00037993306747879105, 0.0008113553880436958]",mixture,"{'mixture': 0.6821446395117668, 'uniform': 0.6616735807081089, 'top1': 0.6817950271148931, 'top2': 0.6820726476890172, 'top3': 0.6817917541113985}",0.872970073630813,0.1522411484043612,0.008795662186711174,0.7289755472355686,0.5567674137146332,student,0.763,
+7,174,0.0,66.2,,,,,,,,,,,,0.8847641017050304,0.14680541323636923,0.008719392809047056,,,,,True

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt

@@ -0,0 +1,107 @@
+[05:42:29][      0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-laplace budget=5.0% seed=2 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "laplace", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[05:42:29][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[05:42:29][      0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches
+[05:42:30][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, snemi=3.55, mouse-brain=4.08
+[05:42:30][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[05:42:36][      7s][WEIGHTS] mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031 | eff.sources=3.41 | MMD2 mix=0.0128 uniform=0.0970
+[05:42:36][      7s][WEIGHTS] per-source: c-elegans=0.320, cerebellum=0.017, flywire=0.001, berghia=0.175, microns=0.027, oct-glia=0.436, snemi=0.017, mouse-brain=0.007 | min-MMD source (NeuroADDA rule) = oct-glia
+[05:42:55][     26s][TEACHERS] target-val Dice: c-elegans=0.476, cerebellum=0.633, flywire=0.142, berghia=0.467, microns=0.708, oct-glia=0.845, snemi=0.229, mouse-brain=0.115
+[05:42:55][     26s][TEACHERS] ensemble target-val Dice: mixture=0.849, mixture@raw=0.847, mixture@logit=0.850, per_source=0.843, per_source@raw=0.619, per_source@logit=0.848, uniform=0.718, uniform@raw=0.677, uniform@logit=0.807
+[05:42:55][     26s][WEIGHTS] using mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031
+[05:42:55][     26s][INIT] student initialised from teacher oct-glia (weight 0.636)
+[05:42:56][     27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[05:42:56][     27s][KD] distilling the mixture ensemble (prob space) into the student
+[05:43:37][     68s][KD] soft labels for 3584 patches from 8 teachers in 41s
+[05:44:09][    100s][KD] epoch 1/5 | patches=3584 | loss=0.5967 | 32s
+[05:44:40][    131s][KD] epoch 2/5 | patches=3584 | loss=0.4536 | 31s
+[05:45:11][    162s][KD] epoch 3/5 | patches=3584 | loss=0.4007 | 31s
+[05:45:42][    193s][KD] epoch 4/5 | patches=3584 | loss=0.3637 | 31s
+[05:46:14][    225s][KD] epoch 5/5 | patches=3584 | loss=0.3392 | 32s
+[05:46:15][    227s][ROUND0] before any target label: val_dice=0.8527 val_vi=0.1456
+[05:46:15][    227s][AL] patches=3584 (pool 2338 after dropping blank; free blank negatives: True) | budget=179 (5.0%) | rounds=6 | per_round=29 | acquisition=laplace | first_round=stratified
+[05:46:23][    235s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.4468831233044695, 'auroc_top20_error': 0.618147538735774}
+[05:46:46][    258s][AL] round 1: stratified cold start: 15 patches from the membrane-rich half (ensemble prevalence >= 0.192) + 14 from the membrane-poor half of 2000 candidates
+[05:46:49][    261s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.765, uniform=0.741, top1=0.765, top2=0.764, top3=0.765 -> using mixture
+[05:46:50][    261s][WEIGHTS] round 1: student kept (labeled Dice 0.760 vs best teacher oct-glia 0.765); distilling the corrected ensemble into it
+[05:46:50][    261s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (179039 fg px): c-elegans=0.257, cerebellum=0.569, flywire=0.169, berghia=0.276, microns=0.581, oct-glia=0.765, snemi=0.110, mouse-brain=0.203 | weights: c-elegans=0.001, cerebellum=0.006, flywire=0.000, berghia=0.002, microns=0.008, oct-glia=0.983, snemi=0.000, mouse-brain=0.000
+[05:47:16][    288s][KD] soft labels for 2309 patches from 8 teachers in 26s
+[05:47:37][    308s][KD] epoch 1/2 | patches=2309 | loss=0.1388 | 21s
+[05:47:58][    329s][KD] epoch 2/2 | patches=2309 | loss=0.1333 | 21s
+[05:48:21][    352s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[05:48:22][    354s][FT] round 1/6 epoch 1/5 | patches=58 | loss=0.8715
+[05:48:24][    355s][FT] round 1/6 epoch 2/5 | patches=58 | loss=0.5653
+[05:48:25][    357s][FT] round 1/6 epoch 3/5 | patches=58 | loss=0.5161
+[05:48:27][    358s][FT] round 1/6 epoch 4/5 | patches=58 | loss=0.5170
+[05:48:28][    360s][FT] round 1/6 epoch 5/5 | patches=58 | loss=0.4795
+[05:48:30][    361s][AL] round 1/6 | labeled=29 | selection_seconds=30.9000 round_seconds=132.8000 restart_student_dice=0.7603 restart_teacher_dice=0.7648 val_dice=0.8582 val_vi=0.1520 val_ece=0.0197
+[05:48:37][    368s][AL] round 2: 590/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:48:43][    374s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.598, uniform=0.658, top1=0.599, top2=0.597, top3=0.597 -> using uniform
+[05:48:43][    374s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (638880 fg px): c-elegans=0.359, cerebellum=0.528, flywire=0.175, berghia=0.310, microns=0.562, oct-glia=0.599, snemi=0.177, mouse-brain=0.325 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[05:48:49][    380s][CONS] round 2: on the 29 new patches student Dice=0.661 vs ensemble 0.616 -> pseudo-labels from the student | consistency x0.54
+[05:48:51][    383s][FT] round 2/6 epoch 1/5 | patches=116 | loss=0.5240
+[05:48:54][    385s][FT] round 2/6 epoch 2/5 | patches=116 | loss=0.5249
+[05:48:56][    387s][FT] round 2/6 epoch 3/5 | patches=116 | loss=0.4267
+[05:48:58][    390s][FT] round 2/6 epoch 4/5 | patches=116 | loss=0.4202
+[05:49:01][    392s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.4486
+[05:49:02][    393s][AL] round 2/6 | labeled=58 | selection_seconds=7.5000 round_seconds=30.8000 pseudo_student_dice=0.6608 pseudo_ensemble_dice=0.6160 consistency_scale=0.5360 val_dice=0.8519 val_vi=0.1695 val_ece=0.0174
+[05:49:09][    401s][AL] round 3: 567/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:50:00][    451s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.433, cerebellum=0.661, flywire=0.229, berghia=0.388, microns=0.577, oct-glia=0.797, snemi=0.336, mouse-brain=0.309 | blended with labeled Dice at 0.50
+[05:50:47][    498s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.708, uniform=0.708, top1=0.708, top2=0.709, top3=0.708 -> using mixture
+[05:50:47][    498s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (816708 fg px): c-elegans=0.355, cerebellum=0.544, flywire=0.200, berghia=0.336, microns=0.539, oct-glia=0.708, snemi=0.247, mouse-brain=0.285 | weights: c-elegans=0.004, cerebellum=0.008, flywire=0.000, berghia=0.005, microns=0.009, oct-glia=0.972, snemi=0.000, mouse-brain=0.001
+[05:50:50][    501s][CONS] round 3: on the 29 new patches student Dice=0.542 vs ensemble 0.682 -> pseudo-labels from the ensemble | consistency x1.00
+[05:50:53][    505s][FT] round 3/6 epoch 1/5 | patches=174 | loss=0.5448
+[05:50:57][    508s][FT] round 3/6 epoch 2/5 | patches=174 | loss=0.5017
+[05:51:00][    511s][FT] round 3/6 epoch 3/5 | patches=174 | loss=0.5088
+[05:51:04][    515s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.4270
+[05:51:07][    518s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.4082
+[05:51:08][    520s][AL] round 3/6 | labeled=87 | selection_seconds=7.6000 round_seconds=124.8000 pseudo_student_dice=0.5422 pseudo_ensemble_dice=0.6818 consistency_scale=1.0000 val_dice=0.8693 val_vi=0.1472 val_ece=0.0117
+[05:51:16][    527s][AL] round 4: 584/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:51:26][    538s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.472, uniform=0.491, top1=0.471, top2=0.470, top3=0.470 -> using uniform
+[05:51:26][    538s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1509201 fg px): c-elegans=0.281, cerebellum=0.374, flywire=0.150, berghia=0.245, microns=0.402, oct-glia=0.471, snemi=0.209, mouse-brain=0.290 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[05:51:32][    544s][CONS] round 4: on the 29 new patches student Dice=0.526 vs ensemble 0.265 -> pseudo-labels from the student | consistency x0.09
+[05:51:36][    548s][FT] round 4/6 epoch 1/5 | patches=232 | loss=0.6832
+[05:51:41][    552s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.9581
+[05:51:45][    556s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.5798
+[05:51:49][    561s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.3980
+[05:51:54][    565s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.4769
+[05:51:55][    567s][AL] round 4/6 | labeled=116 | selection_seconds=7.6000 round_seconds=45.4000 pseudo_student_dice=0.5262 pseudo_ensemble_dice=0.2655 consistency_scale=0.0870 val_dice=0.8838 val_vi=0.1529 val_ece=0.0111
+[05:52:03][    574s][AL] round 5: 569/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:52:55][    627s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.420, cerebellum=0.661, flywire=0.239, berghia=0.388, microns=0.594, oct-glia=0.833, snemi=0.317, mouse-brain=0.291 | blended with labeled Dice at 0.50
+[05:53:45][    676s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.657, uniform=0.667, top1=0.656, top2=0.657, top3=0.656 -> using uniform
+[05:53:45][    676s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1935317 fg px): c-elegans=0.358, cerebellum=0.529, flywire=0.197, berghia=0.324, microns=0.497, oct-glia=0.656, snemi=0.265, mouse-brain=0.294 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[05:53:51][    682s][CONS] round 5: on the 29 new patches student Dice=0.723 vs ensemble 0.558 -> pseudo-labels from the student | consistency x0.74
+[05:53:56][    687s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.5566
+[05:54:01][    693s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.5073
+[05:54:07][    698s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.4316
+[05:54:12][    703s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.4643
+[05:54:17][    709s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.6065
+[05:54:19][    710s][AL] round 5/6 | labeled=145 | selection_seconds=7.4000 round_seconds=141.9000 pseudo_student_dice=0.7225 pseudo_ensemble_dice=0.5582 consistency_scale=0.7420 val_dice=0.8797 val_vi=0.1583 val_ece=0.0099
+[05:54:26][    717s][AL] round 6: 561/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[05:55:20][    772s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.419, cerebellum=0.654, flywire=0.238, berghia=0.389, microns=0.593, oct-glia=0.826, snemi=0.320, mouse-brain=0.293 | blended with labeled Dice at 0.50
+[05:56:11][    823s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.665, uniform=0.674, top1=0.664, top2=0.665, top3=0.664 -> using uniform
+[05:56:11][    823s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2258511 fg px): c-elegans=0.360, cerebellum=0.536, flywire=0.199, berghia=0.334, microns=0.501, oct-glia=0.664, snemi=0.265, mouse-brain=0.294 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[05:56:17][    828s][CONS] round 6: on the 29 new patches student Dice=0.767 vs ensemble 0.632 -> pseudo-labels from the student | consistency x0.89
+[05:56:23][    835s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.5293
+[05:56:29][    841s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.4097
+[05:56:36][    847s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.3725
+[05:56:42][    853s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.3776
+[05:56:48][    860s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.3708
+[05:56:50][    861s][AL] round 6/6 | labeled=174 | selection_seconds=7.5000 round_seconds=149.7000 pseudo_student_dice=0.7673 pseudo_ensemble_dice=0.6316 consistency_scale=0.8910 val_dice=0.8887 val_vi=0.1513 val_ece=0.0094
+[05:56:53][    864s][CONS] final consolidation: pseudo-labels refreshed from the student
+[05:57:00][    871s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.3354
+[05:57:06][    877s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.3671
+[05:57:12][    884s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.4036
+[05:57:18][    890s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.4250
+[05:57:25][    896s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.3658
+[05:57:31][    903s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.3438
+[05:57:37][    909s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.3156
+[05:57:44][    915s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.3556
+[05:57:50][    921s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.3240
+[05:57:56][    928s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.3354
+[05:57:58][    929s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=66.3000 val_dice=0.8887 val_vi=0.1468 val_ece=0.0108
+[05:57:58][    929s][WEIGHTS] final: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[05:58:08][    940s][TEST] raw student (no EMA, no TTA): patch_dice=0.8648 patch_vi=0.1775
+[05:58:10][    941s][TEST] final model without TTA: patch_dice=0.8648 patch_vi=0.1775
+[05:58:10][    941s][RESOURCES] {"phases": {"setup": {"seconds": 1.14, "peak_cpu_rss_mb": 1027.2, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 25.95, "peak_cpu_rss_mb": 1808.4, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 197.94, "peak_cpu_rss_mb": 2938.2, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 702.31, "peak_cpu_rss_mb": 2938.2, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 12.0, "peak_cpu_rss_mb": 2938.2, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 768, "images": 6, "patches_per_second": 64.0, "seconds_per_image": 2.0}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2938.2, "total_seconds": 939.3}
+[05:58:10][    941s][TEST] patch_dice=0.8647 patch_iou=0.8033 patch_precision=0.9035 patch_recall=0.8667 patch_vi=0.1753 image_dice=0.7921 image_iou=0.6652 image_precision=0.7986 image_recall=0.7938 image_vi=0.1992 cal_ece=0.0066 cal_nll=0.0771 cal_brier=0.0219 cal_mean_confidence=0.9765 cal_accuracy=0.9701 cal_predicted_prevalence=0.0729 cal_true_prevalence=0.0718 cal_mean_probability=0.0727
+[05:58:10][    941s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-laplace b5 s2 | test patch Dice=0.8647 VI=0.1753 | image Dice=0.7921 VI=0.1992 | 941s

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json

@@ -0,0 +1,776 @@
+{
+  "target": "octopus-vulgaris-vertical-lobe-sfltract",
+  "variant": "macs-v11-norestart-acq-laplace",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "octopus-vulgaris-vertical-lobe-sfltract",
+    "variant": "macs-v11-norestart-acq-laplace",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "laplace",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "--variants",
+      "macs-v11-norestart-acq-laplace",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.09888773316940314,
+    0.02852238358730045,
+    0.006877111295002701,
+    0.14762496137410624,
+    0.031629415990885375,
+    0.6364113152878205,
+    0.01903121486130303,
+    0.03101586443417853
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.09888773316940314,
+      0.02852238358730045,
+      0.006877111295002701,
+      0.14762496137410624,
+      0.031629415990885375,
+      0.6364113152878205,
+      0.01903121486130303,
+      0.03101586443417853
+    ],
+    "per_source": [
+      0.3196472209911767,
+      0.016524281783989066,
+      0.0007715704214095669,
+      0.17539935780265475,
+      0.02748152710843107,
+      0.4363201963447303,
+      0.016643455180579362,
+      0.00721239036702913
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.00048684441851293454,
+      "mmd2_mixture": 0.01275567083190321,
+      "mmd2_uniform": 0.09698324592915836,
+      "mmd2_single_source": [
+        0.07244541101040552,
+        0.256897395483041,
+        0.3929088032810887,
+        0.05287172109384486,
+        0.26498470596000767,
+        0.011934168264888134,
+        0.339258942897097,
+        0.2684433232010991
+      ],
+      "median_squared_distance": 0.18344226468067681,
+      "gammas": [
+        1.3628266116054668,
+        2.7256532232109336,
+        5.451306446421867,
+        10.902612892843734,
+        21.80522578568747
+      ],
+      "max_weight": 0.6364113152878205,
+      "entropy": 1.2268826413390275,
+      "effective_sources": 3.410580943129122,
+      "min_weight": 0.006877111295002701
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.0496200334528889,
+        0.42663682749168497,
+        0.8166058321365515,
+        0.12600018979150718,
+        0.3618977926472007,
+        0.01001955344998784,
+        0.4257222632020228,
+        0.5321458020642489
+      ],
+      "standardized": [
+        -1.1548905481172125,
+        0.3263029673133632,
+        1.8583820949064265,
+        -0.8548142807860197,
+        0.07196140979373515,
+        -1.3104697560349354,
+        0.32270990006354566,
+        0.7408182128610976
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4363201963447303,
+      "entropy": 1.3076016834873658,
+      "effective_sources": 3.6972957848927845,
+      "min_weight": 0.0007715704214095669
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4757138324136084,
+      0.6329416506363946,
+      0.14191107641715917,
+      0.4669227387129098,
+      0.7079858630413108,
+      0.8450575342195966,
+      0.2290226697023551,
+      0.11473197497037783
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.8487996328814233,
+      "mixture@raw": 0.8468783370062004,
+      "mixture@logit": 0.8498165961112344,
+      "per_source": 0.8430704623597217,
+      "per_source@raw": 0.6192918885999461,
+      "per_source@logit": 0.8479512470904803,
+      "uniform": 0.7183095932451996,
+      "uniform@raw": 0.6771602818911759,
+      "uniform@logit": 0.8065076510127938
+    },
+    "val_patches": 768
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "weight_stats": {
+    "max_weight": 0.6364113152878205,
+    "entropy": 1.2268826413390275,
+    "effective_sources": 3.410580943129122,
+    "min_weight": 0.006877111295002701
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.5952380952380953,
+    "spearman_weight_taxonomy": 0.6257415691925263,
+    "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "student_init": "top-source",
+  "blank_fraction": 0.34765625,
+  "kd_seconds": 197.9,
+  "round0_val": {
+    "val_dice": 0.8527240517674085,
+    "val_vi": 0.1455715366868596,
+    "val_ece": 0.06194851765646773
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 29,
+      "selection_seconds": 30.9,
+      "round_seconds": 132.8,
+      "teacher_dice_labeled": [
+        0.2571777942968238,
+        0.5690605504275248,
+        0.1686588604013407,
+        0.2758037194420113,
+        0.5811237001396865,
+        0.7648114620563284,
+        0.10985703762520847,
+        0.20342659662893037
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 179039,
+      "evidence_nonempty_patches": 21,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.7603000274646353,
+      "restart_teacher_dice": 0.7648114620563284,
+      "weights": [
+        0.0009536063313568787,
+        0.00622160049410085,
+        2.7365280978604556e-05,
+        0.0017150542671718894,
+        0.00778389682673956,
+        0.9830816831898324,
+        4.206174632514249e-05,
+        0.0001747318634946766
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7646874735949377,
+        "uniform": 0.7405359432386459,
+        "top1": 0.7648114620563284,
+        "top2": 0.7643850371191413,
+        "top3": 0.7646318754401064
+      },
+      "val_dice": 0.8582032598395823,
+      "val_vi": 0.15202137350161224,
+      "val_ece": 0.01967065256522741
+    },
+    {
+      "round": 2,
+      "n_labeled": 58,
+      "selection_seconds": 7.5,
+      "round_seconds": 30.8,
+      "teacher_dice_labeled": [
+        0.3588600320416357,
+        0.5283610665271623,
+        0.1754836982934096,
+        0.30965104253978754,
+        0.5621811921846875,
+        0.5986021775475412,
+        0.17715468712343485,
+        0.3248640901664514
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 638880,
+      "evidence_nonempty_patches": 50,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.5983930359391161,
+        "uniform": 0.6576421407682509,
+        "top1": 0.5986021775475412,
+        "top2": 0.5968083368993796,
+        "top3": 0.597320828329163
+      },
+      "pseudo_student_dice": 0.6608478637558942,
+      "pseudo_ensemble_dice": 0.6159762447621395,
+      "pseudo_source": "student",
+      "consistency_scale": 0.536,
+      "val_dice": 0.8519298021852384,
+      "val_vi": 0.1695486292888474,
+      "val_ece": 0.01740237649855266
+    },
+    {
+      "round": 3,
+      "n_labeled": 87,
+      "selection_seconds": 7.6,
+      "round_seconds": 124.8,
+      "teacher_dice_labeled": [
+        0.3550508758955952,
+        0.5444899469348708,
+        0.20038206997152114,
+        0.33595390889582355,
+        0.5390476896390397,
+        0.7080628920438248,
+        0.24712166474669486,
+        0.28541810272502866
+      ],
+      "teacher_dice_pool": [
+        0.4328659013976558,
+        0.661135411122184,
+        0.22903361986981352,
+        0.38842096758112005,
+        0.5765943393419145,
+        0.797088399227804,
+        0.33584992040025263,
+        0.30945976554298166
+      ],
+      "evidence_fg_pixels": 816708,
+      "evidence_nonempty_patches": 74,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.00442401041120939,
+        0.008483654377623146,
+        6.551815847839877e-05,
+        0.005456277398952017,
+        0.008909491152141185,
+        0.9716801215408971,
+        0.00028934076331636185,
+        0.0006915861973823141
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.708058559923522,
+        "uniform": 0.7080861303326307,
+        "top1": 0.7080628920438248,
+        "top2": 0.708505089247222,
+        "top3": 0.7076425236551863
+      },
+      "pseudo_student_dice": 0.5422034726585485,
+      "pseudo_ensemble_dice": 0.6817766491064452,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.8692568562603569,
+      "val_vi": 0.14722995535203087,
+      "val_ece": 0.011732638054617456
+    },
+    {
+      "round": 4,
+      "n_labeled": 116,
+      "selection_seconds": 7.6,
+      "round_seconds": 45.4,
+      "teacher_dice_labeled": [
+        0.2812613113577766,
+        0.37428606024354677,
+        0.15015731686738024,
+        0.24517187706258378,
+        0.4022338376611055,
+        0.471145297805343,
+        0.2088359438902864,
+        0.29038033021372894
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1509201,
+      "evidence_nonempty_patches": 103,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.47207579835642194,
+        "uniform": 0.4913551341147341,
+        "top1": 0.471145297805343,
+        "top2": 0.4699058743924868,
+        "top3": 0.4702756933037702
+      },
+      "pseudo_student_dice": 0.526227817416522,
+      "pseudo_ensemble_dice": 0.26545333677506955,
+      "pseudo_source": "student",
+      "consistency_scale": 0.087,
+      "val_dice": 0.8837533492723129,
+      "val_vi": 0.15290057656135012,
+      "val_ece": 0.011086664785801689
+    },
+    {
+      "round": 5,
+      "n_labeled": 145,
+      "selection_seconds": 7.4,
+      "round_seconds": 141.9,
+      "teacher_dice_labeled": [
+        0.3575857705768084,
+        0.5288894927276441,
+        0.1972882651699285,
+        0.3237538784944006,
+        0.4972437412072611,
+        0.6563423781453516,
+        0.2654844034258358,
+        0.2937566814955266
+      ],
+      "teacher_dice_pool": [
+        0.4195271845365769,
+        0.6613145146970063,
+        0.23861478496716995,
+        0.38766042413489726,
+        0.5937290521963867,
+        0.8326240947868595,
+        0.31651714330217995,
+        0.29086484509438304
+      ],
+      "evidence_fg_pixels": 1935317,
+      "evidence_nonempty_patches": 132,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.6568639291864343,
+        "uniform": 0.6666955000371759,
+        "top1": 0.6563423781453516,
+        "top2": 0.6570021039931723,
+        "top3": 0.6561298907989814
+      },
+      "pseudo_student_dice": 0.7225415124397063,
+      "pseudo_ensemble_dice": 0.5582015873469427,
+      "pseudo_source": "student",
+      "consistency_scale": 0.742,
+      "val_dice": 0.8797230617226582,
+      "val_vi": 0.15834655247909965,
+      "val_ece": 0.00989123741765351
+    },
+    {
+      "round": 6,
+      "n_labeled": 174,
+      "selection_seconds": 7.5,
+      "round_seconds": 149.7,
+      "teacher_dice_labeled": [
+        0.3602505543526682,
+        0.5359821774533255,
+        0.19879252836890904,
+        0.333613065162099,
+        0.5009071969940995,
+        0.6643916205152369,
+        0.26497846806752806,
+        0.2937445620594825
+      ],
+      "teacher_dice_pool": [
+        0.41908704552384624,
+        0.6539098615025841,
+        0.23820759536247726,
+        0.38935192574692595,
+        0.592662873689235,
+        0.8257198302767631,
+        0.32037599926452937,
+        0.29283550908122574
+      ],
+      "evidence_fg_pixels": 2258511,
+      "evidence_nonempty_patches": 161,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.6648869229132081,
+        "uniform": 0.6739570229701966,
+        "top1": 0.6643916205152369,
+        "top2": 0.6649912132591502,
+        "top3": 0.6642192650367976
+      },
+      "pseudo_student_dice": 0.7673293550156336,
+      "pseudo_ensemble_dice": 0.6316422578928547,
+      "pseudo_source": "student",
+      "consistency_scale": 0.891,
+      "val_dice": 0.8886952685341428,
+      "val_vi": 0.1513188957973902,
+      "val_ece": 0.0094336474487976
+    },
+    {
+      "round": 7,
+      "n_labeled": 174,
+      "selection_seconds": 0.0,
+      "round_seconds": 66.3,
+      "final_consolidation": true,
+      "val_dice": 0.8886864148471417,
+      "val_vi": 0.14682279863099465,
+      "val_ece": 0.010849301665396582
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.4468831233044695,
+    "auroc_top20_error": 0.618147538735774
+  },
+  "al_seconds": 702.3,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.125,
+    "entropy": 2.0794415416798357,
+    "effective_sources": 7.999999999999998,
+    "min_weight": 0.125
+  },
+  "top_source_final": "c-elegans-dauer-stage",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.8647887142358766,
+    "patch_iou": 0.8012740206376692,
+    "patch_precision": 0.8955808545122568,
+    "patch_recall": 0.8680106101740352,
+    "patch_vi": 0.17752917596284057,
+    "n_patches": 768,
+    "cal_ece": 0.008089954957642595,
+    "cal_nll": 0.08198473665119577,
+    "cal_brier": 0.0224464631914164,
+    "cal_mean_confidence": 0.9775250897061557,
+    "cal_accuracy": 0.9694976806640625,
+    "cal_predicted_prevalence": 0.07344818115234375,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07302339563645663
+  },
+  "test_notta": {
+    "patch_dice": 0.8647887142358766,
+    "patch_iou": 0.8012740206376692,
+    "patch_precision": 0.8955808545122568,
+    "patch_recall": 0.8680106101740352,
+    "patch_vi": 0.17752917596284057,
+    "n_patches": 768,
+    "cal_ece": 0.008089954957642595,
+    "cal_nll": 0.08198473665119577,
+    "cal_brier": 0.0224464631914164,
+    "cal_mean_confidence": 0.9775250897061557,
+    "cal_accuracy": 0.9694976806640625,
+    "cal_predicted_prevalence": 0.07344818115234375,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07302339563645663
+  },
+  "test": {
+    "patch_dice": 0.8647069438389737,
+    "patch_iou": 0.8033281169756871,
+    "patch_precision": 0.9034503196996484,
+    "patch_recall": 0.8667368064369821,
+    "patch_vi": 0.17527923318371696,
+    "n_patches": 768,
+    "image_dice": 0.7920637119427608,
+    "image_iou": 0.665237657237726,
+    "image_precision": 0.7986320096435179,
+    "image_recall": 0.7938427626550224,
+    "image_vi": 0.1992288512318123,
+    "n_images": 6,
+    "cal_ece": 0.006552162411524904,
+    "cal_nll": 0.07708213635633097,
+    "cal_brier": 0.021867434336899864,
+    "cal_mean_confidence": 0.9765031516520883,
+    "cal_accuracy": 0.9700775146484375,
+    "cal_predicted_prevalence": 0.07292429606119792,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.0726615540301751
+  },
+  "eval_seconds": 12.0,
+  "total_seconds": 941.1,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.14,
+        "peak_cpu_rss_mb": 1027.2,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 25.95,
+        "peak_cpu_rss_mb": 1808.4,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 197.94,
+        "peak_cpu_rss_mb": 2938.2,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 702.31,
+        "peak_cpu_rss_mb": 2938.2,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4626.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 12.0,
+        "peak_cpu_rss_mb": 2938.2,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4626.0,
+        "patches": 768,
+        "images": 6,
+        "patches_per_second": 64.0,
+        "seconds_per_image": 2.0
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2938.2,
+    "total_seconds": 939.3
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,29,30.9,132.8,"[0.2571777942968238, 0.5690605504275248, 0.1686588604013407, 0.2758037194420113, 0.5811237001396865, 0.7648114620563284, 0.10985703762520847, 0.20342659662893037]",,179039,21,True,kept,0.7603000274646353,0.7648114620563284,"[0.0009536063313568787, 0.00622160049410085, 2.7365280978604556e-05, 0.0017150542671718894, 0.00778389682673956, 0.9830816831898324, 4.206174632514249e-05, 0.0001747318634946766]",mixture,"{'mixture': 0.7646874735949377, 'uniform': 0.7405359432386459, 'top1': 0.7648114620563284, 'top2': 0.7643850371191413, 'top3': 0.7646318754401064}",0.8582032598395823,0.15202137350161224,0.01967065256522741,,,,,
+2,58,7.5,30.8,"[0.3588600320416357, 0.5283610665271623, 0.1754836982934096, 0.30965104253978754, 0.5621811921846875, 0.5986021775475412, 0.17715468712343485, 0.3248640901664514]",,638880,50,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.5983930359391161, 'uniform': 0.6576421407682509, 'top1': 0.5986021775475412, 'top2': 0.5968083368993796, 'top3': 0.597320828329163}",0.8519298021852384,0.1695486292888474,0.01740237649855266,0.6608478637558942,0.6159762447621395,student,0.536,
+3,87,7.6,124.8,"[0.3550508758955952, 0.5444899469348708, 0.20038206997152114, 0.33595390889582355, 0.5390476896390397, 0.7080628920438248, 0.24712166474669486, 0.28541810272502866]","[0.4328659013976558, 0.661135411122184, 0.22903361986981352, 0.38842096758112005, 0.5765943393419145, 0.797088399227804, 0.33584992040025263, 0.30945976554298166]",816708,74,False,,,,"[0.00442401041120939, 0.008483654377623146, 6.551815847839877e-05, 0.005456277398952017, 0.008909491152141185, 0.9716801215408971, 0.00028934076331636185, 0.0006915861973823141]",mixture,"{'mixture': 0.708058559923522, 'uniform': 0.7080861303326307, 'top1': 0.7080628920438248, 'top2': 0.708505089247222, 'top3': 0.7076425236551863}",0.8692568562603569,0.14722995535203087,0.011732638054617456,0.5422034726585485,0.6817766491064452,ensemble,1.0,
+4,116,7.6,45.4,"[0.2812613113577766, 0.37428606024354677, 0.15015731686738024, 0.24517187706258378, 0.4022338376611055, 0.471145297805343, 0.2088359438902864, 0.29038033021372894]",,1509201,103,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.47207579835642194, 'uniform': 0.4913551341147341, 'top1': 0.471145297805343, 'top2': 0.4699058743924868, 'top3': 0.4702756933037702}",0.8837533492723129,0.15290057656135012,0.011086664785801689,0.526227817416522,0.26545333677506955,student,0.087,
+5,145,7.4,141.9,"[0.3575857705768084, 0.5288894927276441, 0.1972882651699285, 0.3237538784944006, 0.4972437412072611, 0.6563423781453516, 0.2654844034258358, 0.2937566814955266]","[0.4195271845365769, 0.6613145146970063, 0.23861478496716995, 0.38766042413489726, 0.5937290521963867, 0.8326240947868595, 0.31651714330217995, 0.29086484509438304]",1935317,132,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6568639291864343, 'uniform': 0.6666955000371759, 'top1': 0.6563423781453516, 'top2': 0.6570021039931723, 'top3': 0.6561298907989814}",0.8797230617226582,0.15834655247909965,0.00989123741765351,0.7225415124397063,0.5582015873469427,student,0.742,
+6,174,7.5,149.7,"[0.3602505543526682, 0.5359821774533255, 0.19879252836890904, 0.333613065162099, 0.5009071969940995, 0.6643916205152369, 0.26497846806752806, 0.2937445620594825]","[0.41908704552384624, 0.6539098615025841, 0.23820759536247726, 0.38935192574692595, 0.592662873689235, 0.8257198302767631, 0.32037599926452937, 0.29283550908122574]",2258511,161,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6648869229132081, 'uniform': 0.6739570229701966, 'top1': 0.6643916205152369, 'top2': 0.6649912132591502, 'top3': 0.6642192650367976}",0.8886952685341428,0.1513188957973902,0.0094336474487976,0.7673293550156336,0.6316422578928547,student,0.891,
+7,174,0.0,66.3,,,,,,,,,,,,0.8886864148471417,0.14682279863099465,0.010849301665396582,,,,,True

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-3/log.txt

@@ -0,0 +1,107 @@
+[10:02:19][      0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-laplace budget=5.0% seed=3 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 3, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "laplace", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[10:02:19][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[10:02:19][      0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches
+[10:02:20][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, snemi=3.55, mouse-brain=4.08
+[10:02:20][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[10:02:26][      7s][WEIGHTS] mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031 | eff.sources=3.41 | MMD2 mix=0.0128 uniform=0.0970
+[10:02:26][      7s][WEIGHTS] per-source: c-elegans=0.320, cerebellum=0.017, flywire=0.001, berghia=0.175, microns=0.027, oct-glia=0.436, snemi=0.017, mouse-brain=0.007 | min-MMD source (NeuroADDA rule) = oct-glia
+[10:02:45][     26s][TEACHERS] target-val Dice: c-elegans=0.476, cerebellum=0.633, flywire=0.142, berghia=0.467, microns=0.708, oct-glia=0.845, snemi=0.229, mouse-brain=0.115
+[10:02:45][     26s][TEACHERS] ensemble target-val Dice: mixture=0.849, mixture@raw=0.847, mixture@logit=0.850, per_source=0.843, per_source@raw=0.619, per_source@logit=0.848, uniform=0.718, uniform@raw=0.677, uniform@logit=0.807
+[10:02:45][     26s][WEIGHTS] using mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031
+[10:02:45][     26s][INIT] student initialised from teacher oct-glia (weight 0.636)
+[10:02:46][     27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[10:02:46][     27s][KD] distilling the mixture ensemble (prob space) into the student
+[10:03:27][     68s][KD] soft labels for 3584 patches from 8 teachers in 41s
+[10:03:59][    100s][KD] epoch 1/5 | patches=3584 | loss=0.5673 | 32s
+[10:04:30][    131s][KD] epoch 2/5 | patches=3584 | loss=0.4392 | 31s
+[10:05:03][    164s][KD] epoch 3/5 | patches=3584 | loss=0.3945 | 33s
+[10:05:36][    197s][KD] epoch 4/5 | patches=3584 | loss=0.3593 | 33s
+[10:06:07][    228s][KD] epoch 5/5 | patches=3584 | loss=0.3311 | 31s
+[10:06:08][    230s][ROUND0] before any target label: val_dice=0.8515 val_vi=0.1453
+[10:06:08][    230s][AL] patches=3584 (pool 2338 after dropping blank; free blank negatives: True) | budget=179 (5.0%) | rounds=6 | per_round=29 | acquisition=laplace | first_round=stratified
+[10:06:16][    237s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.4591574662939882, 'auroc_top20_error': 0.6181795328854153}
+[10:06:39][    260s][AL] round 1: stratified cold start: 15 patches from the membrane-rich half (ensemble prevalence >= 0.192) + 14 from the membrane-poor half of 2000 candidates
+[10:06:42][    264s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.811, uniform=0.768, top1=0.811, top2=0.811, top3=0.811 -> using mixture
+[10:06:43][    264s][WEIGHTS] round 1: student kept (labeled Dice 0.827 vs best teacher oct-glia 0.811); distilling the corrected ensemble into it
+[10:06:43][    264s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (179827 fg px): c-elegans=0.279, cerebellum=0.617, flywire=0.194, berghia=0.316, microns=0.585, oct-glia=0.811, snemi=0.111, mouse-brain=0.210 | weights: c-elegans=0.001, cerebellum=0.006, flywire=0.000, berghia=0.002, microns=0.005, oct-glia=0.986, snemi=0.000, mouse-brain=0.000
+[10:07:09][    290s][KD] soft labels for 2309 patches from 8 teachers in 26s
+[10:07:30][    311s][KD] epoch 1/2 | patches=2309 | loss=0.1381 | 21s
+[10:07:51][    332s][KD] epoch 2/2 | patches=2309 | loss=0.1323 | 21s
+[10:08:14][    355s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[10:08:16][    357s][FT] round 1/6 epoch 1/5 | patches=58 | loss=0.8383
+[10:08:17][    358s][FT] round 1/6 epoch 2/5 | patches=58 | loss=0.8501
+[10:08:18][    360s][FT] round 1/6 epoch 3/5 | patches=58 | loss=0.6314
+[10:08:20][    361s][FT] round 1/6 epoch 4/5 | patches=58 | loss=0.4879
+[10:08:21][    363s][FT] round 1/6 epoch 5/5 | patches=58 | loss=0.4561
+[10:08:23][    364s][AL] round 1/6 | labeled=29 | selection_seconds=30.8000 round_seconds=133.2000 restart_student_dice=0.8273 restart_teacher_dice=0.8106 val_dice=0.8568 val_vi=0.1554 val_ece=0.0169
+[10:08:31][    372s][AL] round 2: 586/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:08:36][    378s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.680, uniform=0.674, top1=0.680, top2=0.680, top3=0.680 -> using mixture
+[10:08:36][    378s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (549483 fg px): c-elegans=0.320, cerebellum=0.519, flywire=0.188, berghia=0.329, microns=0.518, oct-glia=0.680, snemi=0.151, mouse-brain=0.275 | weights: c-elegans=0.004, cerebellum=0.009, flywire=0.000, berghia=0.007, microns=0.010, oct-glia=0.970, snemi=0.000, mouse-brain=0.001
+[10:08:42][    383s][CONS] round 2: on the 29 new patches student Dice=0.647 vs ensemble 0.603 -> pseudo-labels from the student | consistency x0.49
+[10:08:45][    386s][FT] round 2/6 epoch 1/5 | patches=116 | loss=0.5509
+[10:08:47][    388s][FT] round 2/6 epoch 2/5 | patches=116 | loss=0.4116
+[10:08:49][    391s][FT] round 2/6 epoch 3/5 | patches=116 | loss=0.3350
+[10:08:52][    393s][FT] round 2/6 epoch 4/5 | patches=116 | loss=0.3681
+[10:08:54][    395s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.3817
+[10:08:55][    397s][AL] round 2/6 | labeled=58 | selection_seconds=7.6000 round_seconds=30.9000 pseudo_student_dice=0.6467 pseudo_ensemble_dice=0.6028 consistency_scale=0.4890 val_dice=0.8618 val_vi=0.1453 val_ece=0.0160
+[10:09:03][    404s][AL] round 3: 586/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:09:53][    454s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.363, cerebellum=0.615, flywire=0.246, berghia=0.366, microns=0.621, oct-glia=0.869, snemi=0.273, mouse-brain=0.242 | blended with labeled Dice at 0.50
+[10:10:40][    501s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.721, uniform=0.700, top1=0.721, top2=0.721, top3=0.721 -> using mixture
+[10:10:40][    501s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (1138003 fg px): c-elegans=0.338, cerebellum=0.551, flywire=0.213, berghia=0.334, microns=0.538, oct-glia=0.721, snemi=0.243, mouse-brain=0.269 | weights: c-elegans=0.003, cerebellum=0.008, flywire=0.000, berghia=0.005, microns=0.008, oct-glia=0.975, snemi=0.000, mouse-brain=0.001
+[10:10:46][    507s][CONS] round 3: on the 29 new patches student Dice=0.522 vs ensemble 0.450 -> pseudo-labels from the student | consistency x0.07
+[10:10:49][    511s][FT] round 3/6 epoch 1/5 | patches=174 | loss=0.9099
+[10:10:53][    514s][FT] round 3/6 epoch 2/5 | patches=174 | loss=0.8412
+[10:10:56][    517s][FT] round 3/6 epoch 3/5 | patches=174 | loss=0.7315
+[10:10:59][    521s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.7279
+[10:11:03][    524s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.6884
+[10:11:04][    526s][AL] round 3/6 | labeled=87 | selection_seconds=7.4000 round_seconds=127.4000 pseudo_student_dice=0.5217 pseudo_ensemble_dice=0.4502 consistency_scale=0.0720 val_dice=0.8723 val_vi=0.1533 val_ece=0.0118
+[10:11:11][    533s][AL] round 4: 572/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:12:03][    584s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.379, cerebellum=0.620, flywire=0.246, berghia=0.376, microns=0.620, oct-glia=0.857, snemi=0.282, mouse-brain=0.257 | blended with labeled Dice at 0.50
+[10:12:52][    633s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.690, uniform=0.682, top1=0.690, top2=0.690, top3=0.690 -> using mixture
+[10:12:52][    633s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1656088 fg px): c-elegans=0.333, cerebellum=0.541, flywire=0.209, berghia=0.330, microns=0.517, oct-glia=0.690, snemi=0.259, mouse-brain=0.275 | weights: c-elegans=0.004, cerebellum=0.010, flywire=0.000, berghia=0.006, microns=0.009, oct-glia=0.970, snemi=0.000, mouse-brain=0.001
+[10:12:57][    639s][CONS] round 4: on the 29 new patches student Dice=0.547 vs ensemble 0.395 -> pseudo-labels from the student | consistency x0.16
+[10:13:02][    643s][FT] round 4/6 epoch 1/5 | patches=232 | loss=0.6824
+[10:13:06][    647s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.5460
+[10:13:10][    652s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.5290
+[10:13:15][    656s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.6514
+[10:13:19][    660s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.5690
+[10:13:20][    662s][AL] round 4/6 | labeled=116 | selection_seconds=7.4000 round_seconds=134.6000 pseudo_student_dice=0.5471 pseudo_ensemble_dice=0.3954 consistency_scale=0.1570 val_dice=0.7533 val_vi=0.3211 val_ece=0.0161
+[10:13:28][    669s][AL] round 5: 528/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:14:21][    722s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.557, cerebellum=0.595, flywire=0.207, berghia=0.471, microns=0.463, oct-glia=0.556, snemi=0.429, mouse-brain=0.484 | blended with labeled Dice at 0.50
+[10:15:10][    772s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.553, uniform=0.606, top1=0.549, top2=0.551, top3=0.549 -> using uniform
+[10:15:10][    772s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1757811 fg px): c-elegans=0.417, cerebellum=0.525, flywire=0.186, berghia=0.364, microns=0.439, oct-glia=0.549, snemi=0.314, mouse-brain=0.379 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[10:15:13][    774s][CONS] round 5: on the 29 new patches student Dice=0.269 vs ensemble 0.561 -> pseudo-labels from the ensemble | consistency x1.00
+[10:15:19][    780s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.6929
+[10:15:24][    785s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.5823
+[10:15:29][    791s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.5388
+[10:15:35][    796s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.5194
+[10:15:40][    801s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.4847
+[10:15:41][    803s][AL] round 5/6 | labeled=145 | selection_seconds=7.5000 round_seconds=139.6000 pseudo_student_dice=0.2694 pseudo_ensemble_dice=0.5612 consistency_scale=1.0000 val_dice=0.8744 val_vi=0.1512 val_ece=0.0079
+[10:15:49][    810s][AL] round 6: 573/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:16:04][    826s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.537, uniform=0.558, top1=0.536, top2=0.537, top3=0.536 -> using uniform
+[10:16:04][    826s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2188227 fg px): c-elegans=0.288, cerebellum=0.456, flywire=0.166, berghia=0.274, microns=0.415, oct-glia=0.536, snemi=0.205, mouse-brain=0.281 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[10:16:10][    831s][CONS] round 6: on the 29 new patches student Dice=0.734 vs ensemble 0.563 -> pseudo-labels from the student | consistency x0.78
+[10:16:17][    838s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.4139
+[10:16:23][    844s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.4055
+[10:16:29][    850s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.3732
+[10:16:36][    857s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.3716
+[10:16:42][    863s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.4127
+[10:16:43][    865s][AL] round 6/6 | labeled=174 | selection_seconds=7.3000 round_seconds=60.4000 pseudo_student_dice=0.7344 pseudo_ensemble_dice=0.5635 consistency_scale=0.7810 val_dice=0.8617 val_vi=0.1666 val_ece=0.0074
+[10:16:46][    868s][CONS] final consolidation: pseudo-labels refreshed from the student
+[10:16:53][    874s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.4970
+[10:16:59][    881s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.4235
+[10:17:06][    887s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.4503
+[10:17:12][    893s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.3917
+[10:17:18][    900s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.3825
+[10:17:25][    906s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.3685
+[10:17:31][    912s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.3657
+[10:17:37][    919s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.3473
+[10:17:44][    925s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.3830
+[10:17:50][    931s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.3511
+[10:17:52][    933s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=66.5000 val_dice=0.8885 val_vi=0.1459 val_ece=0.0064
+[10:17:52][    933s][WEIGHTS] final: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[10:18:02][    944s][TEST] raw student (no EMA, no TTA): patch_dice=0.8706 patch_vi=0.1784
+[10:18:04][    945s][TEST] final model without TTA: patch_dice=0.8706 patch_vi=0.1784
+[10:18:04][    945s][RESOURCES] {"phases": {"setup": {"seconds": 1.14, "peak_cpu_rss_mb": 1027.7, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 25.77, "peak_cpu_rss_mb": 1809.1, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 200.91, "peak_cpu_rss_mb": 2938.6, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 703.34, "peak_cpu_rss_mb": 2938.6, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 12.3, "peak_cpu_rss_mb": 2938.6, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 768, "images": 6, "patches_per_second": 62.4, "seconds_per_image": 2.05}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2938.6, "total_seconds": 943.5}
+[10:18:04][    945s][TEST] patch_dice=0.8722 patch_iou=0.8093 patch_precision=0.8889 patch_recall=0.8824 patch_vi=0.1747 image_dice=0.8012 image_iou=0.6755 image_precision=0.7977 image_recall=0.8106 image_vi=0.1969 cal_ece=0.0051 cal_nll=0.0693 cal_brier=0.0212 cal_mean_confidence=0.9759 cal_accuracy=0.9709 cal_predicted_prevalence=0.0741 cal_true_prevalence=0.0718 cal_mean_probability=0.0749
+[10:18:04][    945s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-laplace b5 s3 | test patch Dice=0.8722 VI=0.1747 | image Dice=0.8012 VI=0.1969 | 945s

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-3/result.json

@@ -0,0 +1,776 @@
+{
+  "target": "octopus-vulgaris-vertical-lobe-sfltract",
+  "variant": "macs-v11-norestart-acq-laplace",
+  "budget": 5.0,
+  "seed": 3,
+  "config": {
+    "task": "membrane",
+    "target": "octopus-vulgaris-vertical-lobe-sfltract",
+    "variant": "macs-v11-norestart-acq-laplace",
+    "budget": 5.0,
+    "seed": 3,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "laplace",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "--variants",
+      "macs-v11-norestart-acq-laplace",
+      "--budgets",
+      "5",
+      "--seeds",
+      "3",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.09888773316940314,
+    0.02852238358730045,
+    0.006877111295002701,
+    0.14762496137410624,
+    0.031629415990885375,
+    0.6364113152878205,
+    0.01903121486130303,
+    0.03101586443417853
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.09888773316940314,
+      0.02852238358730045,
+      0.006877111295002701,
+      0.14762496137410624,
+      0.031629415990885375,
+      0.6364113152878205,
+      0.01903121486130303,
+      0.03101586443417853
+    ],
+    "per_source": [
+      0.3196472209911767,
+      0.016524281783989066,
+      0.0007715704214095669,
+      0.17539935780265475,
+      0.02748152710843107,
+      0.4363201963447303,
+      0.016643455180579362,
+      0.00721239036702913
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.00048684441851293454,
+      "mmd2_mixture": 0.01275567083190321,
+      "mmd2_uniform": 0.09698324592915836,
+      "mmd2_single_source": [
+        0.07244541101040552,
+        0.256897395483041,
+        0.3929088032810887,
+        0.05287172109384486,
+        0.26498470596000767,
+        0.011934168264888134,
+        0.339258942897097,
+        0.2684433232010991
+      ],
+      "median_squared_distance": 0.18344226468067681,
+      "gammas": [
+        1.3628266116054668,
+        2.7256532232109336,
+        5.451306446421867,
+        10.902612892843734,
+        21.80522578568747
+      ],
+      "max_weight": 0.6364113152878205,
+      "entropy": 1.2268826413390275,
+      "effective_sources": 3.410580943129122,
+      "min_weight": 0.006877111295002701
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.0496200334528889,
+        0.42663682749168497,
+        0.8166058321365515,
+        0.12600018979150718,
+        0.3618977926472007,
+        0.01001955344998784,
+        0.4257222632020228,
+        0.5321458020642489
+      ],
+      "standardized": [
+        -1.1548905481172125,
+        0.3263029673133632,
+        1.8583820949064265,
+        -0.8548142807860197,
+        0.07196140979373515,
+        -1.3104697560349354,
+        0.32270990006354566,
+        0.7408182128610976
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4363201963447303,
+      "entropy": 1.3076016834873658,
+      "effective_sources": 3.6972957848927845,
+      "min_weight": 0.0007715704214095669
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4757138324136084,
+      0.6329416506363946,
+      0.14191107641715917,
+      0.4669227387129098,
+      0.7079858630413108,
+      0.8450575342195966,
+      0.2290226697023551,
+      0.11473197497037783
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.8487996328814233,
+      "mixture@raw": 0.8468783370062004,
+      "mixture@logit": 0.8498165961112344,
+      "per_source": 0.8430704623597217,
+      "per_source@raw": 0.6192918885999461,
+      "per_source@logit": 0.8479512470904803,
+      "uniform": 0.7183095932451996,
+      "uniform@raw": 0.6771602818911759,
+      "uniform@logit": 0.8065076510127938
+    },
+    "val_patches": 768
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "weight_stats": {
+    "max_weight": 0.6364113152878205,
+    "entropy": 1.2268826413390275,
+    "effective_sources": 3.410580943129122,
+    "min_weight": 0.006877111295002701
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.5952380952380953,
+    "spearman_weight_taxonomy": 0.6257415691925263,
+    "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "student_init": "top-source",
+  "blank_fraction": 0.34765625,
+  "kd_seconds": 200.9,
+  "round0_val": {
+    "val_dice": 0.8515102797526266,
+    "val_vi": 0.14528530548020757,
+    "val_ece": 0.05870248018856551
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 29,
+      "selection_seconds": 30.8,
+      "round_seconds": 133.2,
+      "teacher_dice_labeled": [
+        0.278686973709348,
+        0.6172069054477305,
+        0.19414410467595433,
+        0.3158733635275065,
+        0.5849527168014728,
+        0.8105764798984268,
+        0.11111243697584838,
+        0.21043401064256403
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 179827,
+      "evidence_nonempty_patches": 22,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.8273144858761027,
+      "restart_teacher_dice": 0.8105764798984268,
+      "weights": [
+        0.0007503874401662406,
+        0.006390021724894883,
+        2.240702364497097e-05,
+        0.0016248054634756897,
+        0.005132506760358811,
+        0.9859339072089507,
+        2.702979059074383e-05,
+        0.0001189345879178869
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.8107891012852039,
+        "uniform": 0.7680840640525722,
+        "top1": 0.8105764798984268,
+        "top2": 0.8107881846175238,
+        "top3": 0.8107436386731551
+      },
+      "val_dice": 0.8568044591794021,
+      "val_vi": 0.15536332065951733,
+      "val_ece": 0.016875869306927235
+    },
+    {
+      "round": 2,
+      "n_labeled": 58,
+      "selection_seconds": 7.6,
+      "round_seconds": 30.9,
+      "teacher_dice_labeled": [
+        0.3197324212585647,
+        0.5186989851123923,
+        0.18769042400258,
+        0.32897712171977045,
+        0.5183921599714619,
+        0.6796743091805819,
+        0.15061281497014214,
+        0.2754290735520091
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 549483,
+      "evidence_nonempty_patches": 50,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.004119763372614849,
+        0.008689933929460799,
+        7.650411497395904e-05,
+        0.006745884040301596,
+        0.009607033535081643,
+        0.9697850851871787,
+        0.00014612322378190518,
+        0.000829672596606807
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6799376360850514,
+        "uniform": 0.6739134714897281,
+        "top1": 0.6796743091805819,
+        "top2": 0.6799873705508308,
+        "top3": 0.6796347829641398
+      },
+      "pseudo_student_dice": 0.6466831096930886,
+      "pseudo_ensemble_dice": 0.6027818783231503,
+      "pseudo_source": "student",
+      "consistency_scale": 0.489,
+      "val_dice": 0.861787266202709,
+      "val_vi": 0.14531560750334052,
+      "val_ece": 0.01596588622935088
+    },
+    {
+      "round": 3,
+      "n_labeled": 87,
+      "selection_seconds": 7.4,
+      "round_seconds": 127.4,
+      "teacher_dice_labeled": [
+        0.33816096920786365,
+        0.5508747489357715,
+        0.2127544630286278,
+        0.3342955063859918,
+        0.5383557530871647,
+        0.7209293122344551,
+        0.24296450621113985,
+        0.2688145903643604
+      ],
+      "teacher_dice_pool": [
+        0.3631213703221983,
+        0.6145786325256796,
+        0.24602950896413575,
+        0.3664223347709242,
+        0.6212679758984406,
+        0.8686868739620435,
+        0.273348192075486,
+        0.24215863231774376
+      ],
+      "evidence_fg_pixels": 1138003,
+      "evidence_nonempty_patches": 79,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0032977847478816197,
+        0.007981231489846457,
+        6.544137261997147e-05,
+        0.004736441621020206,
+        0.007809189399856895,
+        0.9753479342568299,
+        0.0002449702865546584,
+        0.0005170068253904648
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7213260143771558,
+        "uniform": 0.6999932181400559,
+        "top1": 0.7209293122344551,
+        "top2": 0.7211679464098952,
+        "top3": 0.7211597694240854
+      },
+      "pseudo_student_dice": 0.5216835236694869,
+      "pseudo_ensemble_dice": 0.4502389093759044,
+      "pseudo_source": "student",
+      "consistency_scale": 0.072,
+      "val_dice": 0.8722882299942253,
+      "val_vi": 0.153306299739384,
+      "val_ece": 0.011781420070394528
+    },
+    {
+      "round": 4,
+      "n_labeled": 116,
+      "selection_seconds": 7.4,
+      "round_seconds": 134.6,
+      "teacher_dice_labeled": [
+        0.33335654980390694,
+        0.5414478519401984,
+        0.20853802261015897,
+        0.3302917388799157,
+        0.5167268481721374,
+        0.6897356638254668,
+        0.2592273871447747,
+        0.27469722168508637
+      ],
+      "teacher_dice_pool": [
+        0.37884987150908017,
+        0.6200130681762214,
+        0.24550768821346203,
+        0.37607195076658384,
+        0.6200735224606091,
+        0.856574798895515,
+        0.28189492315590287,
+        0.25733860588598917
+      ],
+      "evidence_fg_pixels": 1656088,
+      "evidence_nonempty_patches": 108,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.004269758757935813,
+        0.009866787505957434,
+        8.522892157258763e-05,
+        0.006181735356236639,
+        0.008545139890425265,
+        0.9699149110425108,
+        0.0003915512060296217,
+        0.0007448873193318116
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6901279025681524,
+        "uniform": 0.6820152178728482,
+        "top1": 0.6897356638254668,
+        "top2": 0.6900723044431973,
+        "top3": 0.6898417583058918
+      },
+      "pseudo_student_dice": 0.5470911753511737,
+      "pseudo_ensemble_dice": 0.39544808173752155,
+      "pseudo_source": "student",
+      "consistency_scale": 0.157,
+      "val_dice": 0.7532872911825909,
+      "val_vi": 0.3210910036063804,
+      "val_ece": 0.016072281468791392
+    },
+    {
+      "round": 5,
+      "n_labeled": 145,
+      "selection_seconds": 7.5,
+      "round_seconds": 139.6,
+      "teacher_dice_labeled": [
+        0.41676594581650617,
+        0.5246918455092885,
+        0.18630663891166677,
+        0.3641756496484856,
+        0.4389208361506449,
+        0.5486765842831056,
+        0.31428819745996883,
+        0.3791764241664638
+      ],
+      "teacher_dice_pool": [
+        0.5572855135959716,
+        0.5945990793338956,
+        0.20735111987360755,
+        0.47074842385203725,
+        0.4634087060527902,
+        0.5556566786907592,
+        0.4294796873842047,
+        0.4841524752871939
+      ],
+      "evidence_fg_pixels": 1757811,
+      "evidence_nonempty_patches": 135,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.553159440532073,
+        "uniform": 0.6059717310472152,
+        "top1": 0.5486765842831056,
+        "top2": 0.5506095424125969,
+        "top3": 0.5490036599201282
+      },
+      "pseudo_student_dice": 0.2693943934850826,
+      "pseudo_ensemble_dice": 0.5612330996243244,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.874420324316172,
+      "val_vi": 0.15116773626438473,
+      "val_ece": 0.00787832594784592
+    },
+    {
+      "round": 6,
+      "n_labeled": 174,
+      "selection_seconds": 7.3,
+      "round_seconds": 60.4,
+      "teacher_dice_labeled": [
+        0.2876735695465173,
+        0.45617902944643696,
+        0.1661216846844591,
+        0.2744202625517011,
+        0.41452631979279514,
+        0.5362464651534956,
+        0.20464379152966422,
+        0.28141783931791187
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 2188227,
+      "evidence_nonempty_patches": 164,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.5372292629570921,
+        "uniform": 0.5577719248545915,
+        "top1": 0.5362464651534956,
+        "top2": 0.5369621472938708,
+        "top3": 0.536194536894178
+      },
+      "pseudo_student_dice": 0.73435671070419,
+      "pseudo_ensemble_dice": 0.56346838228734,
+      "pseudo_source": "student",
+      "consistency_scale": 0.781,
+      "val_dice": 0.8617161226240094,
+      "val_vi": 0.16661853359663176,
+      "val_ece": 0.007415384371264444
+    },
+    {
+      "round": 7,
+      "n_labeled": 174,
+      "selection_seconds": 0.0,
+      "round_seconds": 66.5,
+      "final_consolidation": true,
+      "val_dice": 0.8885012303311527,
+      "val_vi": 0.14585021657683853,
+      "val_ece": 0.006406493248002914
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.4591574662939882,
+    "auroc_top20_error": 0.6181795328854153
+  },
+  "al_seconds": 703.3,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125,
+    0.125
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.125,
+    "entropy": 2.0794415416798357,
+    "effective_sources": 7.999999999999998,
+    "min_weight": 0.125
+  },
+  "top_source_final": "c-elegans-dauer-stage",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.8706010955494335,
+    "patch_iou": 0.8062181478807697,
+    "patch_precision": 0.8878454942987101,
+    "patch_recall": 0.879627908979745,
+    "patch_vi": 0.17839662107777113,
+    "n_patches": 768,
+    "cal_ece": 0.00687900400319244,
+    "cal_nll": 0.07303922115646799,
+    "cal_brier": 0.021915210183211336,
+    "cal_mean_confidence": 0.9768012147741905,
+    "cal_accuracy": 0.970001220703125,
+    "cal_predicted_prevalence": 0.07367706298828125,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.0744689811575781
+  },
+  "test_notta": {
+    "patch_dice": 0.8706010955494335,
+    "patch_iou": 0.8062181478807697,
+    "patch_precision": 0.8878454942987101,
+    "patch_recall": 0.879627908979745,
+    "patch_vi": 0.17839662107777113,
+    "n_patches": 768,
+    "cal_ece": 0.00687900400319244,
+    "cal_nll": 0.07303922115646799,
+    "cal_brier": 0.021915210183211336,
+    "cal_mean_confidence": 0.9768012147741905,
+    "cal_accuracy": 0.970001220703125,
+    "cal_predicted_prevalence": 0.07367706298828125,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.0744689811575781
+  },
+  "test": {
+    "patch_dice": 0.8721652716705893,
+    "patch_iou": 0.8092648914040117,
+    "patch_precision": 0.8889102641225229,
+    "patch_recall": 0.8823612947835796,
+    "patch_vi": 0.17472178939575436,
+    "n_patches": 768,
+    "image_dice": 0.8011970493537958,
+    "image_iou": 0.6754771032791154,
+    "image_precision": 0.7977181274448101,
+    "image_recall": 0.8106349655697888,
+    "image_vi": 0.19691996364868994,
+    "n_images": 6,
+    "cal_ece": 0.005060413319983306,
+    "cal_nll": 0.06933385008524794,
+    "cal_brier": 0.0211606984166082,
+    "cal_mean_confidence": 0.9759225850433184,
+    "cal_accuracy": 0.9709307352701823,
+    "cal_predicted_prevalence": 0.0740521748860677,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07489795879321344
+  },
+  "eval_seconds": 12.3,
+  "total_seconds": 945.2,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.14,
+        "peak_cpu_rss_mb": 1027.7,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 25.77,
+        "peak_cpu_rss_mb": 1809.1,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 200.91,
+        "peak_cpu_rss_mb": 2938.6,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 703.34,
+        "peak_cpu_rss_mb": 2938.6,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4626.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 12.3,
+        "peak_cpu_rss_mb": 2938.6,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4626.0,
+        "patches": 768,
+        "images": 6,
+        "patches_per_second": 62.4,
+        "seconds_per_image": 2.05
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2938.6,
+    "total_seconds": 943.5
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace_b5_s3.pth"
+}

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-3/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,29,30.8,133.2,"[0.278686973709348, 0.6172069054477305, 0.19414410467595433, 0.3158733635275065, 0.5849527168014728, 0.8105764798984268, 0.11111243697584838, 0.21043401064256403]",,179827,22,True,kept,0.8273144858761027,0.8105764798984268,"[0.0007503874401662406, 0.006390021724894883, 2.240702364497097e-05, 0.0016248054634756897, 0.005132506760358811, 0.9859339072089507, 2.702979059074383e-05, 0.0001189345879178869]",mixture,"{'mixture': 0.8107891012852039, 'uniform': 0.7680840640525722, 'top1': 0.8105764798984268, 'top2': 0.8107881846175238, 'top3': 0.8107436386731551}",0.8568044591794021,0.15536332065951733,0.016875869306927235,,,,,
+2,58,7.6,30.9,"[0.3197324212585647, 0.5186989851123923, 0.18769042400258, 0.32897712171977045, 0.5183921599714619, 0.6796743091805819, 0.15061281497014214, 0.2754290735520091]",,549483,50,False,,,,"[0.004119763372614849, 0.008689933929460799, 7.650411497395904e-05, 0.006745884040301596, 0.009607033535081643, 0.9697850851871787, 0.00014612322378190518, 0.000829672596606807]",mixture,"{'mixture': 0.6799376360850514, 'uniform': 0.6739134714897281, 'top1': 0.6796743091805819, 'top2': 0.6799873705508308, 'top3': 0.6796347829641398}",0.861787266202709,0.14531560750334052,0.01596588622935088,0.6466831096930886,0.6027818783231503,student,0.489,
+3,87,7.4,127.4,"[0.33816096920786365, 0.5508747489357715, 0.2127544630286278, 0.3342955063859918, 0.5383557530871647, 0.7209293122344551, 0.24296450621113985, 0.2688145903643604]","[0.3631213703221983, 0.6145786325256796, 0.24602950896413575, 0.3664223347709242, 0.6212679758984406, 0.8686868739620435, 0.273348192075486, 0.24215863231774376]",1138003,79,False,,,,"[0.0032977847478816197, 0.007981231489846457, 6.544137261997147e-05, 0.004736441621020206, 0.007809189399856895, 0.9753479342568299, 0.0002449702865546584, 0.0005170068253904648]",mixture,"{'mixture': 0.7213260143771558, 'uniform': 0.6999932181400559, 'top1': 0.7209293122344551, 'top2': 0.7211679464098952, 'top3': 0.7211597694240854}",0.8722882299942253,0.153306299739384,0.011781420070394528,0.5216835236694869,0.4502389093759044,student,0.072,
+4,116,7.4,134.6,"[0.33335654980390694, 0.5414478519401984, 0.20853802261015897, 0.3302917388799157, 0.5167268481721374, 0.6897356638254668, 0.2592273871447747, 0.27469722168508637]","[0.37884987150908017, 0.6200130681762214, 0.24550768821346203, 0.37607195076658384, 0.6200735224606091, 0.856574798895515, 0.28189492315590287, 0.25733860588598917]",1656088,108,False,,,,"[0.004269758757935813, 0.009866787505957434, 8.522892157258763e-05, 0.006181735356236639, 0.008545139890425265, 0.9699149110425108, 0.0003915512060296217, 0.0007448873193318116]",mixture,"{'mixture': 0.6901279025681524, 'uniform': 0.6820152178728482, 'top1': 0.6897356638254668, 'top2': 0.6900723044431973, 'top3': 0.6898417583058918}",0.7532872911825909,0.3210910036063804,0.016072281468791392,0.5470911753511737,0.39544808173752155,student,0.157,
+5,145,7.5,139.6,"[0.41676594581650617, 0.5246918455092885, 0.18630663891166677, 0.3641756496484856, 0.4389208361506449, 0.5486765842831056, 0.31428819745996883, 0.3791764241664638]","[0.5572855135959716, 0.5945990793338956, 0.20735111987360755, 0.47074842385203725, 0.4634087060527902, 0.5556566786907592, 0.4294796873842047, 0.4841524752871939]",1757811,135,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.553159440532073, 'uniform': 0.6059717310472152, 'top1': 0.5486765842831056, 'top2': 0.5506095424125969, 'top3': 0.5490036599201282}",0.874420324316172,0.15116773626438473,0.00787832594784592,0.2693943934850826,0.5612330996243244,ensemble,1.0,
+6,174,7.3,60.4,"[0.2876735695465173, 0.45617902944643696, 0.1661216846844591, 0.2744202625517011, 0.41452631979279514, 0.5362464651534956, 0.20464379152966422, 0.28141783931791187]",,2188227,164,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.5372292629570921, 'uniform': 0.5577719248545915, 'top1': 0.5362464651534956, 'top2': 0.5369621472938708, 'top3': 0.536194536894178}",0.8617161226240094,0.16661853359663176,0.007415384371264444,0.73435671070419,0.56346838228734,student,0.781,
+7,174,0.0,66.5,,,,,,,,,,,,0.8885012303311527,0.14585021657683853,0.006406493248002914,,,,,True

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-margin/budget-5/seed-1/log.txt

@@ -0,0 +1,106 @@
+[03:54:15][      0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-margin budget=5.0% seed=1 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-margin", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "margin", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[03:54:15][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[03:54:15][      0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches
+[03:54:20][      5s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, snemi=3.55, mouse-brain=4.08
+[03:54:22][      6s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 2s
+[03:54:27][     12s][WEIGHTS] mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031 | eff.sources=3.41 | MMD2 mix=0.0128 uniform=0.0970
+[03:54:27][     12s][WEIGHTS] per-source: c-elegans=0.320, cerebellum=0.017, flywire=0.001, berghia=0.175, microns=0.027, oct-glia=0.436, snemi=0.017, mouse-brain=0.007 | min-MMD source (NeuroADDA rule) = oct-glia
+[03:54:46][     30s][TEACHERS] target-val Dice: c-elegans=0.476, cerebellum=0.633, flywire=0.142, berghia=0.467, microns=0.708, oct-glia=0.845, snemi=0.229, mouse-brain=0.115
+[03:54:46][     30s][TEACHERS] ensemble target-val Dice: mixture=0.849, mixture@raw=0.847, mixture@logit=0.850, per_source=0.843, per_source@raw=0.619, per_source@logit=0.848, uniform=0.718, uniform@raw=0.677, uniform@logit=0.807
+[03:54:46][     30s][WEIGHTS] using mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031
+[03:54:46][     30s][INIT] student initialised from teacher oct-glia (weight 0.636)
+[03:54:47][     32s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[03:54:47][     32s][KD] distilling the mixture ensemble (prob space) into the student
+[03:55:28][     72s][KD] soft labels for 3584 patches from 8 teachers in 41s
+[03:55:59][    104s][KD] epoch 1/5 | patches=3584 | loss=0.5719 | 31s
+[03:56:32][    136s][KD] epoch 2/5 | patches=3584 | loss=0.4386 | 32s
+[03:57:03][    168s][KD] epoch 3/5 | patches=3584 | loss=0.4023 | 31s
+[03:57:35][    199s][KD] epoch 4/5 | patches=3584 | loss=0.3566 | 32s
+[03:58:08][    232s][KD] epoch 5/5 | patches=3584 | loss=0.3329 | 33s
+[03:58:09][    234s][ROUND0] before any target label: val_dice=0.8409 val_vi=0.1472
+[03:58:09][    234s][AL] patches=3584 (pool 2338 after dropping blank; free blank negatives: True) | budget=179 (5.0%) | rounds=6 | per_round=29 | acquisition=margin | first_round=stratified
+[03:58:13][    237s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.3286775006068347, 'auroc_top20_error': 0.4926333470451118}
+[03:58:36][    260s][AL] round 1: stratified cold start: 15 patches from the membrane-rich half (ensemble prevalence >= 0.192) + 14 from the membrane-poor half of 2000 candidates
+[03:58:39][    263s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.731, uniform=0.739, top1=0.731, top2=0.731, top3=0.731 -> using uniform
+[03:58:39][    264s][WEIGHTS] round 1: student kept (labeled Dice 0.749 vs best teacher oct-glia 0.731); distilling the corrected ensemble into it
+[03:58:39][    264s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (237130 fg px): c-elegans=0.296, cerebellum=0.620, flywire=0.196, berghia=0.330, microns=0.551, oct-glia=0.731, snemi=0.151, mouse-brain=0.234 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[03:59:06][    290s][KD] soft labels for 2309 patches from 8 teachers in 26s
+[03:59:27][    311s][KD] epoch 1/2 | patches=2309 | loss=0.7236 | 21s
+[03:59:47][    332s][KD] epoch 2/2 | patches=2309 | loss=0.6172 | 20s
+[04:00:10][    355s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[04:00:12][    356s][FT] round 1/6 epoch 1/5 | patches=58 | loss=1.4217
+[04:00:13][    358s][FT] round 1/6 epoch 2/5 | patches=58 | loss=1.3267
+[04:00:15][    359s][FT] round 1/6 epoch 3/5 | patches=58 | loss=1.3169
+[04:00:16][    361s][FT] round 1/6 epoch 4/5 | patches=58 | loss=1.2685
+[04:00:18][    362s][FT] round 1/6 epoch 5/5 | patches=58 | loss=1.3223
+[04:00:19][    364s][AL] round 1/6 | labeled=29 | selection_seconds=26.7000 round_seconds=128.6000 restart_student_dice=0.7494 restart_teacher_dice=0.7308 val_dice=0.4001 val_vi=0.1954 val_ece=0.0596
+[04:00:23][    367s][AL] round 2: 536/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:00:29][    373s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.655, uniform=0.552, top1=0.655, top2=0.655, top3=0.655 -> using mixture
+[04:00:29][    373s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (430675 fg px): c-elegans=0.226, cerebellum=0.352, flywire=0.158, berghia=0.255, microns=0.437, oct-glia=0.655, snemi=0.128, mouse-brain=0.211 | weights: c-elegans=0.002, cerebellum=0.002, flywire=0.000, berghia=0.004, microns=0.006, oct-glia=0.985, snemi=0.000, mouse-brain=0.001
+[04:00:31][    376s][CONS] round 2: on the 29 new patches student Dice=0.495 vs ensemble 0.550 -> pseudo-labels from the ensemble | consistency x1.00
+[04:00:34][    378s][FT] round 2/6 epoch 1/5 | patches=116 | loss=1.3258
+[04:00:36][    381s][FT] round 2/6 epoch 2/5 | patches=116 | loss=1.3389
+[04:00:38][    383s][FT] round 2/6 epoch 3/5 | patches=116 | loss=1.3048
+[04:00:41][    385s][FT] round 2/6 epoch 4/5 | patches=116 | loss=1.2951
+[04:00:43][    388s][FT] round 2/6 epoch 5/5 | patches=116 | loss=1.2841
+[04:00:45][    389s][AL] round 2/6 | labeled=58 | selection_seconds=3.9000 round_seconds=23.9000 pseudo_student_dice=0.4952 pseudo_ensemble_dice=0.5501 consistency_scale=1.0000 val_dice=0.4091 val_vi=0.1649 val_ece=0.0362
+[04:00:48][    393s][AL] round 3: 575/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:00:56][    401s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.419, uniform=0.435, top1=0.417, top2=0.415, top3=0.416 -> using uniform
+[04:00:56][    401s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (1337552 fg px): c-elegans=0.283, cerebellum=0.333, flywire=0.143, berghia=0.223, microns=0.357, oct-glia=0.417, snemi=0.238, mouse-brain=0.314 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[04:01:02][    406s][CONS] round 3: on the 29 new patches student Dice=0.379 vs ensemble 0.330 -> pseudo-labels from the student | consistency x0.00
+[04:01:04][    408s][FT] round 3/6 epoch 1/5 | patches=174 | loss=1.1071
+[04:01:05][    410s][FT] round 3/6 epoch 2/5 | patches=174 | loss=1.0589
+[04:01:07][    412s][FT] round 3/6 epoch 3/5 | patches=174 | loss=0.9893
+[04:01:09][    414s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.9455
+[04:01:11][    415s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.8964
+[04:01:12][    417s][AL] round 3/6 | labeled=87 | selection_seconds=3.6000 round_seconds=26.3000 pseudo_student_dice=0.3792 pseudo_ensemble_dice=0.3304 consistency_scale=0.0000 val_dice=0.4166 val_vi=0.4205 val_ece=0.2388
+[04:01:16][    420s][AL] round 4: 83/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:02:07][    472s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.328, cerebellum=0.460, flywire=0.164, berghia=0.272, microns=0.372, oct-glia=0.534, snemi=0.562, mouse-brain=0.246 | blended with labeled Dice at 0.50
+[04:02:56][    521s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.495, uniform=0.510, top1=0.492, top2=0.493, top3=0.494 -> using uniform
+[04:02:56][    521s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1499750 fg px): c-elegans=0.297, cerebellum=0.407, flywire=0.154, berghia=0.252, microns=0.373, oct-glia=0.492, snemi=0.384, mouse-brain=0.266 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[04:02:59][    523s][CONS] round 4: on the 29 new patches student Dice=0.123 vs ensemble 0.671 -> pseudo-labels from the ensemble | consistency x1.00
+[04:03:03][    528s][FT] round 4/6 epoch 1/5 | patches=232 | loss=1.3456
+[04:03:08][    532s][FT] round 4/6 epoch 2/5 | patches=232 | loss=1.2946
+[04:03:12][    536s][FT] round 4/6 epoch 3/5 | patches=232 | loss=1.2901
+[04:03:16][    541s][FT] round 4/6 epoch 4/5 | patches=232 | loss=1.2260
+[04:03:21][    545s][FT] round 4/6 epoch 5/5 | patches=232 | loss=1.2357
+[04:03:22][    547s][AL] round 4/6 | labeled=116 | selection_seconds=3.7000 round_seconds=128.2000 pseudo_student_dice=0.1227 pseudo_ensemble_dice=0.6706 consistency_scale=1.0000 val_dice=0.4529 val_vi=0.1619 val_ece=0.0398
+[04:03:26][    550s][AL] round 5: 561/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:03:39][    564s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.451, uniform=0.479, top1=0.450, top2=0.450, top3=0.449 -> using uniform
+[04:03:39][    564s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1979991 fg px): c-elegans=0.286, cerebellum=0.376, flywire=0.149, berghia=0.242, microns=0.375, oct-glia=0.450, snemi=0.212, mouse-brain=0.304 | weights: c-elegans=0.125, cerebellum=0.125, flywire=0.125, berghia=0.125, microns=0.125, oct-glia=0.125, snemi=0.125, mouse-brain=0.125
+[04:03:45][    569s][CONS] round 5: on the 29 new patches student Dice=0.657 vs ensemble 0.519 -> pseudo-labels from the student | consistency x0.52
+[04:03:50][    575s][FT] round 5/6 epoch 1/5 | patches=290 | loss=1.0593
+[04:03:56][    580s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.9525
+[04:04:01][    586s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.9657
+[04:04:06][    591s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.9394
+[04:04:12][    596s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.9132
+[04:04:13][    598s][AL] round 5/6 | labeled=145 | selection_seconds=3.8000 round_seconds=49.6000 pseudo_student_dice=0.6574 pseudo_ensemble_dice=0.5187 consistency_scale=0.5250 val_dice=0.8724 val_vi=0.1473 val_ece=0.0059
+[04:04:17][    601s][AL] round 6: 569/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:05:11][    655s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.410, cerebellum=0.662, flywire=0.240, berghia=0.384, microns=0.599, oct-glia=0.865, snemi=0.303, mouse-brain=0.278 | blended with labeled Dice at 0.50
+[04:06:02][    706s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.656, uniform=0.660, top1=0.656, top2=0.657, top3=0.656 -> using mixture
+[04:06:02][    706s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2503796 fg px): c-elegans=0.350, cerebellum=0.522, flywire=0.196, berghia=0.317, microns=0.489, oct-glia=0.656, snemi=0.265, mouse-brain=0.290 | weights: c-elegans=0.007, cerebellum=0.011, flywire=0.000, berghia=0.008, microns=0.009, oct-glia=0.963, snemi=0.001, mouse-brain=0.001
+[04:06:08][    712s][CONS] round 6: on the 29 new patches student Dice=0.618 vs ensemble 0.436 -> pseudo-labels from the student | consistency x0.39
+[04:06:14][    719s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.8625
+[04:06:21][    725s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.8528
+[04:06:27][    731s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.8994
+[04:06:33][    738s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.8433
+[04:06:39][    744s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.7899
+[04:06:41][    746s][AL] round 6/6 | labeled=174 | selection_seconds=3.6000 round_seconds=146.1000 pseudo_student_dice=0.6182 pseudo_ensemble_dice=0.4361 consistency_scale=0.3940 val_dice=0.8941 val_vi=0.1443 val_ece=0.0098
+[04:06:44][    749s][CONS] final consolidation: pseudo-labels refreshed from the student
+[04:06:50][    755s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.7530
+[04:06:57][    761s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.7795
+[04:07:03][    768s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.7870
+[04:07:09][    774s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.7724
+[04:07:16][    780s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.7437
+[04:07:22][    787s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.7325
+[04:07:28][    793s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.7313
+[04:07:35][    799s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.7160
+[04:07:41][    806s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.7257
+[04:07:47][    812s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.7021
+[04:07:49][    813s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=66.4000 val_dice=0.8996 val_vi=0.1415 val_ece=0.0054
+[04:07:49][    813s][WEIGHTS] final: c-elegans=0.007, cerebellum=0.011, flywire=0.000, berghia=0.008, microns=0.009, oct-glia=0.963, snemi=0.001, mouse-brain=0.001
+[04:08:00][    824s][TEST] raw student (no EMA, no TTA): patch_dice=0.8708 patch_vi=0.1813
+[04:08:01][    826s][TEST] final model without TTA: patch_dice=0.8708 patch_vi=0.1813
+[04:08:01][    826s][RESOURCES] {"phases": {"setup": {"seconds": 4.33, "peak_cpu_rss_mb": 1027.1, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 27.04, "peak_cpu_rss_mb": 1808.3, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 200.49, "peak_cpu_rss_mb": 2937.4, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 579.87, "peak_cpu_rss_mb": 2937.4, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 12.25, "peak_cpu_rss_mb": 2937.4, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 768, "images": 6, "patches_per_second": 62.7, "seconds_per_image": 2.042}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2937.4, "total_seconds": 824.0}
+[04:08:01][    826s][TEST] patch_dice=0.8774 patch_iou=0.8156 patch_precision=0.8777 patch_recall=0.9054 patch_vi=0.1781 image_dice=0.8013 image_iou=0.6749 image_precision=0.7567 image_recall=0.8591 image_vi=0.2061 cal_ece=0.0078 cal_nll=0.0752 cal_brier=0.0227 cal_mean_confidence=0.9765 cal_accuracy=0.9687 cal_predicted_prevalence=0.0818 cal_true_prevalence=0.0718 cal_mean_probability=0.0830
+[04:08:01][    826s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-margin b5 s1 | test patch Dice=0.8774 VI=0.1781 | image Dice=0.8013 VI=0.2061 | 826s

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-margin/budget-5/seed-1/result.json

@@ -0,0 +1,767 @@
+{
+  "target": "octopus-vulgaris-vertical-lobe-sfltract",
+  "variant": "macs-v11-norestart-acq-margin",
+  "budget": 5.0,
+  "seed": 1,
+  "config": {
+    "task": "membrane",
+    "target": "octopus-vulgaris-vertical-lobe-sfltract",
+    "variant": "macs-v11-norestart-acq-margin",
+    "budget": 5.0,
+    "seed": 1,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "margin",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "--variants",
+      "macs-v11-norestart-acq-margin",
+      "--budgets",
+      "5",
+      "--seeds",
+      "1",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.09888773316940314,
+    0.02852238358730045,
+    0.006877111295002701,
+    0.14762496137410624,
+    0.031629415990885375,
+    0.6364113152878205,
+    0.01903121486130303,
+    0.03101586443417853
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.09888773316940314,
+      0.02852238358730045,
+      0.006877111295002701,
+      0.14762496137410624,
+      0.031629415990885375,
+      0.6364113152878205,
+      0.01903121486130303,
+      0.03101586443417853
+    ],
+    "per_source": [
+      0.3196472209911767,
+      0.016524281783989066,
+      0.0007715704214095669,
+      0.17539935780265475,
+      0.02748152710843107,
+      0.4363201963447303,
+      0.016643455180579362,
+      0.00721239036702913
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.00048684441851293454,
+      "mmd2_mixture": 0.01275567083190321,
+      "mmd2_uniform": 0.09698324592915836,
+      "mmd2_single_source": [
+        0.07244541101040552,
+        0.256897395483041,
+        0.3929088032810887,
+        0.05287172109384486,
+        0.26498470596000767,
+        0.011934168264888134,
+        0.339258942897097,
+        0.2684433232010991
+      ],
+      "median_squared_distance": 0.18344226468067681,
+      "gammas": [
+        1.3628266116054668,
+        2.7256532232109336,
+        5.451306446421867,
+        10.902612892843734,
+        21.80522578568747
+      ],
+      "max_weight": 0.6364113152878205,
+      "entropy": 1.2268826413390275,
+      "effective_sources": 3.410580943129122,
+      "min_weight": 0.006877111295002701
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.0496200334528889,
+        0.42663682749168497,
+        0.8166058321365515,
+        0.12600018979150718,
+        0.3618977926472007,
+        0.01001955344998784,
+        0.4257222632020228,
+        0.5321458020642489
+      ],
+      "standardized": [
+        -1.1548905481172125,
+        0.3263029673133632,
+        1.8583820949064265,
+        -0.8548142807860197,
+        0.07196140979373515,
+        -1.3104697560349354,
+        0.32270990006354566,
+        0.7408182128610976
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4363201963447303,
+      "entropy": 1.3076016834873658,
+      "effective_sources": 3.6972957848927845,
+      "min_weight": 0.0007715704214095669
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 7.4
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4757138324136084,
+      0.6329416506363946,
+      0.14191107641715917,
+      0.4669227387129098,
+      0.7079858630413108,
+      0.8450575342195966,
+      0.2290226697023551,
+      0.11473197497037783
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.8487996328814233,
+      "mixture@raw": 0.8468783370062004,
+      "mixture@logit": 0.8498165961112344,
+      "per_source": 0.8430704623597217,
+      "per_source@raw": 0.6192918885999461,
+      "per_source@logit": 0.8479512470904803,
+      "uniform": 0.7183095932451996,
+      "uniform@raw": 0.6771602818911759,
+      "uniform@logit": 0.8065076510127938
+    },
+    "val_patches": 768
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "weight_stats": {
+    "max_weight": 0.6364113152878205,
+    "entropy": 1.2268826413390275,
+    "effective_sources": 3.410580943129122,
+    "min_weight": 0.006877111295002701
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.5952380952380953,
+    "spearman_weight_taxonomy": 0.6257415691925263,
+    "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "student_init": "top-source",
+  "blank_fraction": 0.34765625,
+  "kd_seconds": 200.5,
+  "round0_val": {
+    "val_dice": 0.8408797944320378,
+    "val_vi": 0.1471621710815322,
+    "val_ece": 0.06254463788501025
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 29,
+      "selection_seconds": 26.7,
+      "round_seconds": 128.6,
+      "teacher_dice_labeled": [
+        0.29569636245422026,
+        0.6197098722923,
+        0.1958949614829199,
+        0.3302973386383346,
+        0.5508880100223538,
+        0.7308314626395732,
+        0.1505046807462434,
+        0.23392840178802965
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 237130,
+      "evidence_nonempty_patches": 23,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.749368381186149,
+      "restart_teacher_dice": 0.7308314626395732,
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.730833809182471,
+        "uniform": 0.7387444226325794,
+        "top1": 0.7308314626395732,
+        "top2": 0.731135280471351,
+        "top3": 0.7306861293999558
+      },
+      "val_dice": 0.40014016389188084,
+      "val_vi": 0.19540568040446374,
+      "val_ece": 0.05964780378486368
+    },
+    {
+      "round": 2,
+      "n_labeled": 58,
+      "selection_seconds": 3.9,
+      "round_seconds": 23.9,
+      "teacher_dice_labeled": [
+        0.22591072500584414,
+        0.3517937333074959,
+        0.15802432610526804,
+        0.25522763821882105,
+        0.43711356544332,
+        0.6554180053154208,
+        0.127741343329143,
+        0.2108749243984569
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 430675,
+      "evidence_nonempty_patches": 49,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0020876127760991476,
+        0.0021202950500547766,
+        7.36354072139125e-05,
+        0.0041781983943484004,
+        0.005518753282945707,
+        0.9853076052792571,
+        0.00015053234578362758,
+        0.0005633674642973701
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6553196975148962,
+        "uniform": 0.551515749726868,
+        "top1": 0.6554180053154208,
+        "top2": 0.6551663541812812,
+        "top3": 0.6551893202438965
+      },
+      "pseudo_student_dice": 0.4951777392874907,
+      "pseudo_ensemble_dice": 0.5500707948375794,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.4091470433278109,
+      "val_vi": 0.16486421629670236,
+      "val_ece": 0.03624841428508624
+    },
+    {
+      "round": 3,
+      "n_labeled": 87,
+      "selection_seconds": 3.6,
+      "round_seconds": 26.3,
+      "teacher_dice_labeled": [
+        0.2825503726481719,
+        0.33345789577819174,
+        0.14333429919434593,
+        0.22282084470182645,
+        0.35654819022203804,
+        0.4168395243709581,
+        0.23813570786869134,
+        0.3137797196004813
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1337552,
+      "evidence_nonempty_patches": 78,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.418985243294828,
+        "uniform": 0.4346946450266119,
+        "top1": 0.4168395243709581,
+        "top2": 0.41525322405442944,
+        "top3": 0.41554170758199205
+      },
+      "pseudo_student_dice": 0.3791860455265543,
+      "pseudo_ensemble_dice": 0.3304222018760689,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.41663767109360444,
+      "val_vi": 0.4204583851205233,
+      "val_ece": 0.2387630061823596
+    },
+    {
+      "round": 4,
+      "n_labeled": 116,
+      "selection_seconds": 3.7,
+      "round_seconds": 128.2,
+      "teacher_dice_labeled": [
+        0.2969813743939188,
+        0.4072742803649082,
+        0.15403333647361286,
+        0.25234984003272964,
+        0.37322180151283557,
+        0.4923612267773314,
+        0.38435115895613686,
+        0.2663083992428939
+      ],
+      "teacher_dice_pool": [
+        0.32769681668831147,
+        0.4600198472689253,
+        0.16416407488008553,
+        0.2719897206667647,
+        0.37222389220911345,
+        0.5336886450915855,
+        0.5621863582223808,
+        0.24645560702223043
+      ],
+      "evidence_fg_pixels": 1499750,
+      "evidence_nonempty_patches": 104,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.49466062043967657,
+        "uniform": 0.5095499189370184,
+        "top1": 0.4923612267773314,
+        "top2": 0.49321355675721384,
+        "top3": 0.494243065939525
+      },
+      "pseudo_student_dice": 0.12269247283674595,
+      "pseudo_ensemble_dice": 0.67057551441357,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.4529003077444256,
+      "val_vi": 0.1618501742375552,
+      "val_ece": 0.03984077059544547
+    },
+    {
+      "round": 5,
+      "n_labeled": 145,
+      "selection_seconds": 3.8,
+      "round_seconds": 49.6,
+      "teacher_dice_labeled": [
+        0.28635443682328976,
+        0.3761179026020876,
+        0.14867873534616236,
+        0.2422393842583288,
+        0.37512147747731395,
+        0.4498236328623156,
+        0.21215624038179035,
+        0.30371129831917815
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1979991,
+      "evidence_nonempty_patches": 133,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125,
+        0.125
+      ],
+      "ensemble_choice": "uniform",
+      "candidate_dice": {
+        "mixture": 0.4514495482765114,
+        "uniform": 0.478574732040192,
+        "top1": 0.4498236328623156,
+        "top2": 0.4503073352358141,
+        "top3": 0.4489798877982555
+      },
+      "pseudo_student_dice": 0.6573682579006355,
+      "pseudo_ensemble_dice": 0.5186765316630713,
+      "pseudo_source": "student",
+      "consistency_scale": 0.525,
+      "val_dice": 0.872368749411296,
+      "val_vi": 0.14727879594978444,
+      "val_ece": 0.005879442031278861
+    },
+    {
+      "round": 6,
+      "n_labeled": 174,
+      "selection_seconds": 3.6,
+      "round_seconds": 146.1,
+      "teacher_dice_labeled": [
+        0.34966348004284153,
+        0.5224991181605332,
+        0.1961675156290587,
+        0.3170608008365486,
+        0.48857806541319204,
+        0.6559357726159281,
+        0.26461644923199584,
+        0.29042007013582627
+      ],
+      "teacher_dice_pool": [
+        0.4100624776609388,
+        0.6622054020576088,
+        0.24016586508634488,
+        0.3841922238786771,
+        0.5988061724159435,
+        0.8649623049709401,
+        0.3026851799005647,
+        0.27770776236942235
+      ],
+      "evidence_fg_pixels": 2503796,
+      "evidence_nonempty_patches": 162,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.006998517317108208,
+        0.011367502944338007,
+        0.00010486852735956787,
+        0.007541049303920208,
+        0.008979521030995922,
+        0.9632193067092194,
+        0.0005754075693240222,
+        0.0012138265977347025
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6564190895903512,
+        "uniform": 0.6601919630013362,
+        "top1": 0.6559357726159281,
+        "top2": 0.6565601036059059,
+        "top3": 0.6557210754470408
+      },
+      "pseudo_student_dice": 0.618239192860665,
+      "pseudo_ensemble_dice": 0.4361382280324493,
+      "pseudo_source": "student",
+      "consistency_scale": 0.394,
+      "val_dice": 0.8941350213311284,
+      "val_vi": 0.14431789469130596,
+      "val_ece": 0.009760565396891685
+    },
+    {
+      "round": 7,
+      "n_labeled": 174,
+      "selection_seconds": 0.0,
+      "round_seconds": 66.4,
+      "final_consolidation": true,
+      "val_dice": 0.8995945889022355,
+      "val_vi": 0.14149496184449994,
+      "val_ece": 0.0054247581622388205
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.3286775006068347,
+    "auroc_top20_error": 0.4926333470451118
+  },
+  "al_seconds": 579.9,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.006998517317108208,
+    0.011367502944338007,
+    0.00010486852735956787,
+    0.007541049303920208,
+    0.008979521030995922,
+    0.9632193067092194,
+    0.0005754075693240222,
+    0.0012138265977347025
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9632193067092194,
+    "entropy": 0.2142932753715035,
+    "effective_sources": 1.2389859655851392,
+    "min_weight": 0.00010486852735956787
+  },
+  "top_source_final": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.8708351518479791,
+    "patch_iou": 0.8067548801798852,
+    "patch_precision": 0.8696299699165779,
+    "patch_recall": 0.9032768737050612,
+    "patch_vi": 0.1812900673335056,
+    "n_patches": 768,
+    "cal_ece": 0.009750707704115536,
+    "cal_nll": 0.08071878577839842,
+    "cal_brier": 0.02364779874333658,
+    "cal_mean_confidence": 0.9777733720921362,
+    "cal_accuracy": 0.9680226643880209,
+    "cal_predicted_prevalence": 0.08139546712239583,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.08214897919586299
+  },
+  "test_notta": {
+    "patch_dice": 0.8708351518479791,
+    "patch_iou": 0.8067548801798852,
+    "patch_precision": 0.8696299699165779,
+    "patch_recall": 0.9032768737050612,
+    "patch_vi": 0.1812900673335056,
+    "n_patches": 768,
+    "cal_ece": 0.009750707704115536,
+    "cal_nll": 0.08071878577839842,
+    "cal_brier": 0.02364779874333658,
+    "cal_mean_confidence": 0.9777733720921362,
+    "cal_accuracy": 0.9680226643880209,
+    "cal_predicted_prevalence": 0.08139546712239583,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.08214897919586299
+  },
+  "test": {
+    "patch_dice": 0.8773587173152141,
+    "patch_iou": 0.8155625933215656,
+    "patch_precision": 0.8777242031910709,
+    "patch_recall": 0.9053688680403686,
+    "patch_vi": 0.1780580080486999,
+    "n_patches": 768,
+    "image_dice": 0.8013153406391118,
+    "image_iou": 0.6749187689784778,
+    "image_precision": 0.756699242658336,
+    "image_recall": 0.8591299849043321,
+    "image_vi": 0.2061392803206973,
+    "n_images": 6,
+    "cal_ece": 0.007827613516782296,
+    "cal_nll": 0.07517227874221732,
+    "cal_brier": 0.02271224630967771,
+    "cal_mean_confidence": 0.9764888849823414,
+    "cal_accuracy": 0.9687309265136719,
+    "cal_predicted_prevalence": 0.08178329467773438,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.08299390230094521
+  },
+  "eval_seconds": 12.2,
+  "total_seconds": 825.7,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 4.33,
+        "peak_cpu_rss_mb": 1027.1,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 27.04,
+        "peak_cpu_rss_mb": 1808.3,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 200.49,
+        "peak_cpu_rss_mb": 2937.4,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 579.87,
+        "peak_cpu_rss_mb": 2937.4,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4626.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 12.25,
+        "peak_cpu_rss_mb": 2937.4,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4626.0,
+        "patches": 768,
+        "images": 6,
+        "patches_per_second": 62.7,
+        "seconds_per_image": 2.042
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2937.4,
+    "total_seconds": 824.0
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-margin_b5_s1.pth"
+}

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-margin/budget-5/seed-1/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,29,26.7,128.6,"[0.29569636245422026, 0.6197098722923, 0.1958949614829199, 0.3302973386383346, 0.5508880100223538, 0.7308314626395732, 0.1505046807462434, 0.23392840178802965]",,237130,23,True,kept,0.749368381186149,0.7308314626395732,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.730833809182471, 'uniform': 0.7387444226325794, 'top1': 0.7308314626395732, 'top2': 0.731135280471351, 'top3': 0.7306861293999558}",0.40014016389188084,0.19540568040446374,0.05964780378486368,,,,,
+2,58,3.9,23.9,"[0.22591072500584414, 0.3517937333074959, 0.15802432610526804, 0.25522763821882105, 0.43711356544332, 0.6554180053154208, 0.127741343329143, 0.2108749243984569]",,430675,49,False,,,,"[0.0020876127760991476, 0.0021202950500547766, 7.36354072139125e-05, 0.0041781983943484004, 0.005518753282945707, 0.9853076052792571, 0.00015053234578362758, 0.0005633674642973701]",mixture,"{'mixture': 0.6553196975148962, 'uniform': 0.551515749726868, 'top1': 0.6554180053154208, 'top2': 0.6551663541812812, 'top3': 0.6551893202438965}",0.4091470433278109,0.16486421629670236,0.03624841428508624,0.4951777392874907,0.5500707948375794,ensemble,1.0,
+3,87,3.6,26.3,"[0.2825503726481719, 0.33345789577819174, 0.14333429919434593, 0.22282084470182645, 0.35654819022203804, 0.4168395243709581, 0.23813570786869134, 0.3137797196004813]",,1337552,78,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.418985243294828, 'uniform': 0.4346946450266119, 'top1': 0.4168395243709581, 'top2': 0.41525322405442944, 'top3': 0.41554170758199205}",0.41663767109360444,0.4204583851205233,0.2387630061823596,0.3791860455265543,0.3304222018760689,student,0.0,
+4,116,3.7,128.2,"[0.2969813743939188, 0.4072742803649082, 0.15403333647361286, 0.25234984003272964, 0.37322180151283557, 0.4923612267773314, 0.38435115895613686, 0.2663083992428939]","[0.32769681668831147, 0.4600198472689253, 0.16416407488008553, 0.2719897206667647, 0.37222389220911345, 0.5336886450915855, 0.5621863582223808, 0.24645560702223043]",1499750,104,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.49466062043967657, 'uniform': 0.5095499189370184, 'top1': 0.4923612267773314, 'top2': 0.49321355675721384, 'top3': 0.494243065939525}",0.4529003077444256,0.1618501742375552,0.03984077059544547,0.12269247283674595,0.67057551441357,ensemble,1.0,
+5,145,3.8,49.6,"[0.28635443682328976, 0.3761179026020876, 0.14867873534616236, 0.2422393842583288, 0.37512147747731395, 0.4498236328623156, 0.21215624038179035, 0.30371129831917815]",,1979991,133,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.4514495482765114, 'uniform': 0.478574732040192, 'top1': 0.4498236328623156, 'top2': 0.4503073352358141, 'top3': 0.4489798877982555}",0.872368749411296,0.14727879594978444,0.005879442031278861,0.6573682579006355,0.5186765316630713,student,0.525,
+6,174,3.6,146.1,"[0.34966348004284153, 0.5224991181605332, 0.1961675156290587, 0.3170608008365486, 0.48857806541319204, 0.6559357726159281, 0.26461644923199584, 0.29042007013582627]","[0.4100624776609388, 0.6622054020576088, 0.24016586508634488, 0.3841922238786771, 0.5988061724159435, 0.8649623049709401, 0.3026851799005647, 0.27770776236942235]",2503796,162,False,,,,"[0.006998517317108208, 0.011367502944338007, 0.00010486852735956787, 0.007541049303920208, 0.008979521030995922, 0.9632193067092194, 0.0005754075693240222, 0.0012138265977347025]",mixture,"{'mixture': 0.6564190895903512, 'uniform': 0.6601919630013362, 'top1': 0.6559357726159281, 'top2': 0.6565601036059059, 'top3': 0.6557210754470408}",0.8941350213311284,0.14431789469130596,0.009760565396891685,0.618239192860665,0.4361382280324493,student,0.394,
+7,174,0.0,66.4,,,,,,,,,,,,0.8995945889022355,0.14149496184449994,0.0054247581622388205,,,,,True

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-random/budget-5/seed-2/log.txt

@@ -0,0 +1,102 @@
+[06:41:25][      0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-random budget=5.0% seed=2 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-random", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "random", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 0, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[06:41:25][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[06:41:25][      0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches
+[06:41:26][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, snemi=3.55, mouse-brain=4.08
+[06:41:27][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[06:41:32][      8s][WEIGHTS] mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031 | eff.sources=3.41 | MMD2 mix=0.0128 uniform=0.0970
+[06:41:32][      8s][WEIGHTS] per-source: c-elegans=0.320, cerebellum=0.017, flywire=0.001, berghia=0.175, microns=0.027, oct-glia=0.436, snemi=0.017, mouse-brain=0.007 | min-MMD source (NeuroADDA rule) = oct-glia
+[06:41:51][     26s][TEACHERS] target-val Dice: c-elegans=0.476, cerebellum=0.633, flywire=0.142, berghia=0.467, microns=0.708, oct-glia=0.845, snemi=0.229, mouse-brain=0.115
+[06:41:51][     26s][TEACHERS] ensemble target-val Dice: mixture=0.849, mixture@raw=0.847, mixture@logit=0.850, per_source=0.843, per_source@raw=0.619, per_source@logit=0.848, uniform=0.718, uniform@raw=0.677, uniform@logit=0.807
+[06:41:51][     26s][WEIGHTS] using mixture: c-elegans=0.099, cerebellum=0.029, flywire=0.007, berghia=0.148, microns=0.032, oct-glia=0.636, snemi=0.019, mouse-brain=0.031
+[06:41:51][     26s][INIT] student initialised from teacher oct-glia (weight 0.636)
+[06:41:52][     27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[06:41:52][     27s][KD] distilling the mixture ensemble (prob space) into the student
+[06:42:33][     68s][KD] soft labels for 3584 patches from 8 teachers in 41s
+[06:43:05][    100s][KD] epoch 1/5 | patches=3584 | loss=0.5953 | 31s
+[06:43:38][    133s][KD] epoch 2/5 | patches=3584 | loss=0.4565 | 33s
+[06:44:09][    165s][KD] epoch 3/5 | patches=3584 | loss=0.4018 | 32s
+[06:44:41][    196s][KD] epoch 4/5 | patches=3584 | loss=0.3660 | 32s
+[06:45:13][    228s][KD] epoch 5/5 | patches=3584 | loss=0.3395 | 31s
+[06:45:14][    229s][ROUND0] before any target label: val_dice=0.8459 val_vi=0.1456
+[06:45:14][    229s][AL] patches=3584 (pool 2338 after dropping blank; free blank negatives: True) | budget=179 (5.0%) | rounds=6 | per_round=29 | acquisition=random | first_round=stratified
+[06:45:18][    233s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.00029523523234020787, 'auroc_top20_error': 0.4987465149229856}
+[06:45:41][    256s][AL] round 1: stratified cold start: 15 patches from the membrane-rich half (ensemble prevalence >= 0.192) + 14 from the membrane-poor half of 2000 candidates
+[06:45:44][    259s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.765, uniform=0.741, top1=0.765, top2=0.764, top3=0.765 -> using mixture
+[06:45:44][    260s][WEIGHTS] round 1: student kept (labeled Dice 0.743 vs best teacher oct-glia 0.765); distilling the corrected ensemble into it
+[06:45:44][    260s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (179039 fg px): c-elegans=0.257, cerebellum=0.569, flywire=0.169, berghia=0.276, microns=0.581, oct-glia=0.765, snemi=0.110, mouse-brain=0.203 | weights: c-elegans=0.001, cerebellum=0.006, flywire=0.000, berghia=0.002, microns=0.008, oct-glia=0.983, snemi=0.000, mouse-brain=0.000
+[06:46:11][    286s][KD] soft labels for 2309 patches from 8 teachers in 26s
+[06:46:32][    307s][KD] epoch 1/2 | patches=2309 | loss=0.1391 | 21s
+[06:46:53][    328s][KD] epoch 2/2 | patches=2309 | loss=0.1328 | 21s
+[06:47:16][    351s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[06:47:17][    352s][FT] round 1/6 epoch 1/5 | patches=58 | loss=0.8723
+[06:47:19][    354s][FT] round 1/6 epoch 2/5 | patches=58 | loss=0.6747
+[06:47:20][    355s][FT] round 1/6 epoch 3/5 | patches=58 | loss=0.6562
+[06:47:21][    357s][FT] round 1/6 epoch 4/5 | patches=58 | loss=0.4788
+[06:47:23][    358s][FT] round 1/6 epoch 5/5 | patches=58 | loss=0.4930
+[06:47:24][    359s][AL] round 1/6 | labeled=29 | selection_seconds=26.6000 round_seconds=128.7000 restart_student_dice=0.7435 restart_teacher_dice=0.7648 val_dice=0.3714 val_vi=0.3523 val_ece=0.0381
+[06:47:30][    365s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.790, uniform=0.745, top1=0.790, top2=0.790, top3=0.790 -> using mixture
+[06:47:30][    365s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (357977 fg px): c-elegans=0.272, cerebellum=0.565, flywire=0.185, berghia=0.295, microns=0.569, oct-glia=0.790, snemi=0.110, mouse-brain=0.208 | weights: c-elegans=0.001, cerebellum=0.005, flywire=0.000, berghia=0.002, microns=0.005, oct-glia=0.987, snemi=0.000, mouse-brain=0.000
+[06:47:33][    368s][CONS] round 2: on the 29 new patches student Dice=0.502 vs ensemble 0.814 -> pseudo-labels from the ensemble | consistency x1.00
+[06:47:36][    371s][FT] round 2/6 epoch 1/5 | patches=116 | loss=0.4292
+[06:47:38][    373s][FT] round 2/6 epoch 2/5 | patches=116 | loss=0.4555
+[06:47:40][    375s][FT] round 2/6 epoch 3/5 | patches=116 | loss=0.4129
+[06:47:43][    378s][FT] round 2/6 epoch 4/5 | patches=116 | loss=0.3810
+[06:47:45][    380s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.3797
+[06:47:46][    382s][AL] round 2/6 | labeled=58 | selection_seconds=0.2000 round_seconds=20.7000 pseudo_student_dice=0.5021 pseudo_ensemble_dice=0.8137 consistency_scale=1.0000 val_dice=0.8627 val_vi=0.1482 val_ece=0.0201
+[06:47:55][    390s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.777, uniform=0.749, top1=0.777, top2=0.777, top3=0.777 -> using mixture
+[06:47:55][    390s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (538912 fg px): c-elegans=0.268, cerebellum=0.578, flywire=0.187, berghia=0.302, microns=0.575, oct-glia=0.777, snemi=0.110, mouse-brain=0.206 | weights: c-elegans=0.001, cerebellum=0.006, flywire=0.000, berghia=0.002, microns=0.006, oct-glia=0.984, snemi=0.000, mouse-brain=0.000
+[06:48:01][    396s][CONS] round 3: on the 29 new patches student Dice=0.818 vs ensemble 0.751 -> pseudo-labels from the student | consistency x1.00
+[06:48:04][    399s][FT] round 3/6 epoch 1/5 | patches=174 | loss=0.3224
+[06:48:07][    402s][FT] round 3/6 epoch 2/5 | patches=174 | loss=0.2616
+[06:48:11][    406s][FT] round 3/6 epoch 3/5 | patches=174 | loss=0.2470
+[06:48:14][    409s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.2579
+[06:48:17][    412s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.2570
+[06:48:19][    414s][AL] round 3/6 | labeled=87 | selection_seconds=0.2000 round_seconds=30.8000 pseudo_student_dice=0.8182 pseudo_ensemble_dice=0.7506 consistency_scale=1.0000 val_dice=0.8649 val_vi=0.1438 val_ece=0.0222
+[06:49:10][    465s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.396, cerebellum=0.643, flywire=0.241, berghia=0.380, microns=0.601, oct-glia=0.863, snemi=0.297, mouse-brain=0.267 | blended with labeled Dice at 0.50
+[06:49:59][    514s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.810, uniform=0.786, top1=0.810, top2=0.810, top3=0.810 -> using mixture
+[06:49:59][    514s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (840112 fg px): c-elegans=0.352, cerebellum=0.620, flywire=0.217, berghia=0.358, microns=0.585, oct-glia=0.810, snemi=0.211, mouse-brain=0.250 | weights: c-elegans=0.002, cerebellum=0.007, flywire=0.000, berghia=0.002, microns=0.005, oct-glia=0.984, snemi=0.000, mouse-brain=0.000
+[06:50:04][    520s][CONS] round 4: on the 29 new patches student Dice=0.766 vs ensemble 0.719 -> pseudo-labels from the student | consistency x0.89
+[06:50:09][    524s][FT] round 4/6 epoch 1/5 | patches=232 | loss=0.3260
+[06:50:13][    528s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.3081
+[06:50:18][    533s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.3103
+[06:50:22][    537s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.2948
+[06:50:26][    541s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.3194
+[06:50:28][    543s][AL] round 4/6 | labeled=116 | selection_seconds=0.2000 round_seconds=127.4000 pseudo_student_dice=0.7662 pseudo_ensemble_dice=0.7189 consistency_scale=0.8870 val_dice=0.8627 val_vi=0.1471 val_ece=0.0211
+[06:51:21][    596s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.387, cerebellum=0.638, flywire=0.245, berghia=0.382, microns=0.597, oct-glia=0.863, snemi=0.286, mouse-brain=0.260 | blended with labeled Dice at 0.50
+[06:52:10][    646s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.797, uniform=0.768, top1=0.797, top2=0.797, top3=0.797 -> using mixture
+[06:52:10][    646s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1036034 fg px): c-elegans=0.343, cerebellum=0.602, flywire=0.214, berghia=0.353, microns=0.569, oct-glia=0.797, snemi=0.204, mouse-brain=0.245 | weights: c-elegans=0.002, cerebellum=0.006, flywire=0.000, berghia=0.003, microns=0.005, oct-glia=0.984, snemi=0.000, mouse-brain=0.000
+[06:52:16][    651s][CONS] round 5: on the 29 new patches student Dice=0.649 vs ensemble 0.604 -> pseudo-labels from the student | consistency x0.50
+[06:52:22][    657s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.3937
+[06:52:27][    662s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.3946
+[06:52:32][    668s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.3277
+[06:52:38][    673s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.3587
+[06:52:43][    678s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.3612
+[06:52:45][    680s][AL] round 5/6 | labeled=145 | selection_seconds=0.2000 round_seconds=135.4000 pseudo_student_dice=0.6492 pseudo_ensemble_dice=0.6037 consistency_scale=0.4970 val_dice=0.8781 val_vi=0.1511 val_ece=0.0167
+[06:53:39][    734s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.402, cerebellum=0.635, flywire=0.239, berghia=0.375, microns=0.587, oct-glia=0.848, snemi=0.308, mouse-brain=0.276 | blended with labeled Dice at 0.50
+[06:54:30][    785s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.791, uniform=0.769, top1=0.791, top2=0.791, top3=0.791 -> using mixture
+[06:54:30][    785s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (1249546 fg px): c-elegans=0.350, cerebellum=0.605, flywire=0.213, berghia=0.355, microns=0.566, oct-glia=0.791, snemi=0.216, mouse-brain=0.253 | weights: c-elegans=0.002, cerebellum=0.007, flywire=0.000, berghia=0.003, microns=0.005, oct-glia=0.983, snemi=0.000, mouse-brain=0.000
+[06:54:36][    791s][CONS] round 6: on the 29 new patches student Dice=0.797 vs ensemble 0.744 -> pseudo-labels from the student | consistency x0.99
+[06:54:42][    797s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.3454
+[06:54:49][    804s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.3065
+[06:54:55][    810s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.2850
+[06:55:01][    816s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.3737
+[06:55:07][    823s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.2806
+[06:55:09][    824s][AL] round 6/6 | labeled=174 | selection_seconds=0.2000 round_seconds=142.8000 pseudo_student_dice=0.7974 pseudo_ensemble_dice=0.7436 consistency_scale=0.9910 val_dice=0.8896 val_vi=0.1489 val_ece=0.0137
+[06:55:12][    827s][CONS] final consolidation: pseudo-labels refreshed from the student
+[06:55:18][    833s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.3100
+[06:55:25][    840s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.2765
+[06:55:31][    846s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.3011
+[06:55:37][    852s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.2956
+[06:55:43][    859s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.2610
+[06:55:50][    865s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.2400
+[06:55:56][    871s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.2573
+[06:56:02][    878s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.2287
+[06:56:09][    884s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.2246
+[06:56:15][    890s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.2208
+[06:56:16][    892s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=66.1000 val_dice=0.8934 val_vi=0.1478 val_ece=0.0119
+[06:56:16][    892s][WEIGHTS] final: c-elegans=0.002, cerebellum=0.007, flywire=0.000, berghia=0.003, microns=0.005, oct-glia=0.983, snemi=0.000, mouse-brain=0.000
+[06:56:27][    902s][TEST] raw student (no EMA, no TTA): patch_dice=0.8718 patch_vi=0.1764
+[06:56:29][    904s][TEST] final model without TTA: patch_dice=0.8718 patch_vi=0.1764
+[06:56:29][    904s][RESOURCES] {"phases": {"setup": {"seconds": 1.13, "peak_cpu_rss_mb": 1027.6, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 26.07, "peak_cpu_rss_mb": 1808.4, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 200.36, "peak_cpu_rss_mb": 2938.1, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 662.37, "peak_cpu_rss_mb": 2938.1, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 12.25, "peak_cpu_rss_mb": 2938.1, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 768, "images": 6, "patches_per_second": 62.7, "seconds_per_image": 2.042}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2938.1, "total_seconds": 902.2}
+[06:56:29][    904s][TEST] patch_dice=0.8733 patch_iou=0.8119 patch_precision=0.8885 patch_recall=0.8888 patch_vi=0.1735 image_dice=0.8026 image_iou=0.6775 image_precision=0.7886 image_recall=0.8228 image_vi=0.1982 cal_ece=0.0085 cal_nll=0.0749 cal_brier=0.0216 cal_mean_confidence=0.9795 cal_accuracy=0.9710 cal_predicted_prevalence=0.0761 cal_true_prevalence=0.0718 cal_mean_probability=0.0759
+[06:56:29][    904s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-random b5 s2 | test patch Dice=0.8733 VI=0.1735 | image Dice=0.8026 VI=0.1982 | 904s

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-random/budget-5/seed-2/result.json

@@ -0,0 +1,776 @@
+{
+  "target": "octopus-vulgaris-vertical-lobe-sfltract",
+  "variant": "macs-v11-norestart-acq-random",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "octopus-vulgaris-vertical-lobe-sfltract",
+    "variant": "macs-v11-norestart-acq-random",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "random",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 0,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "--variants",
+      "macs-v11-norestart-acq-random",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "snemi3d",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.09888773316940314,
+    0.02852238358730045,
+    0.006877111295002701,
+    0.14762496137410624,
+    0.031629415990885375,
+    0.6364113152878205,
+    0.01903121486130303,
+    0.03101586443417853
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "snemi3d",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.09888773316940314,
+      0.02852238358730045,
+      0.006877111295002701,
+      0.14762496137410624,
+      0.031629415990885375,
+      0.6364113152878205,
+      0.01903121486130303,
+      0.03101586443417853
+    ],
+    "per_source": [
+      0.3196472209911767,
+      0.016524281783989066,
+      0.0007715704214095669,
+      0.17539935780265475,
+      0.02748152710843107,
+      0.4363201963447303,
+      0.016643455180579362,
+      0.00721239036702913
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.00048684441851293454,
+      "mmd2_mixture": 0.01275567083190321,
+      "mmd2_uniform": 0.09698324592915836,
+      "mmd2_single_source": [
+        0.07244541101040552,
+        0.256897395483041,
+        0.3929088032810887,
+        0.05287172109384486,
+        0.26498470596000767,
+        0.011934168264888134,
+        0.339258942897097,
+        0.2684433232010991
+      ],
+      "median_squared_distance": 0.18344226468067681,
+      "gammas": [
+        1.3628266116054668,
+        2.7256532232109336,
+        5.451306446421867,
+        10.902612892843734,
+        21.80522578568747
+      ],
+      "max_weight": 0.6364113152878205,
+      "entropy": 1.2268826413390275,
+      "effective_sources": 3.410580943129122,
+      "min_weight": 0.006877111295002701
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.0496200334528889,
+        0.42663682749168497,
+        0.8166058321365515,
+        0.12600018979150718,
+        0.3618977926472007,
+        0.01001955344998784,
+        0.4257222632020228,
+        0.5321458020642489
+      ],
+      "standardized": [
+        -1.1548905481172125,
+        0.3263029673133632,
+        1.8583820949064265,
+        -0.8548142807860197,
+        0.07196140979373515,
+        -1.3104697560349354,
+        0.32270990006354566,
+        0.7408182128610976
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4363201963447303,
+      "entropy": 1.3076016834873658,
+      "effective_sources": 3.6972957848927845,
+      "min_weight": 0.0007715704214095669
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.2
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4757138324136084,
+      0.6329416506363946,
+      0.14191107641715917,
+      0.4669227387129098,
+      0.7079858630413108,
+      0.8450575342195966,
+      0.2290226697023551,
+      0.11473197497037783
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.8487996328814233,
+      "mixture@raw": 0.8468783370062004,
+      "mixture@logit": 0.8498165961112344,
+      "per_source": 0.8430704623597217,
+      "per_source@raw": 0.6192918885999461,
+      "per_source@logit": 0.8479512470904803,
+      "uniform": 0.7183095932451996,
+      "uniform@raw": 0.6771602818911759,
+      "uniform@logit": 0.8065076510127938
+    },
+    "val_patches": 768
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    3.5461363792419434,
+    4.075904369354248
+  ],
+  "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "weight_stats": {
+    "max_weight": 0.6364113152878205,
+    "entropy": 1.2268826413390275,
+    "effective_sources": 3.410580943129122,
+    "min_weight": 0.006877111295002701
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.5952380952380953,
+    "spearman_weight_taxonomy": 0.6257415691925263,
+    "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "student_init": "top-source",
+  "blank_fraction": 0.34765625,
+  "kd_seconds": 200.4,
+  "round0_val": {
+    "val_dice": 0.8459001660840654,
+    "val_vi": 0.14561681257014922,
+    "val_ece": 0.06096620951525809
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 29,
+      "selection_seconds": 26.6,
+      "round_seconds": 128.7,
+      "teacher_dice_labeled": [
+        0.2571777942968238,
+        0.5690605504275248,
+        0.1686588604013407,
+        0.2758037194420113,
+        0.5811237001396865,
+        0.7648114620563284,
+        0.10985703762520847,
+        0.20342659662893037
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 179039,
+      "evidence_nonempty_patches": 21,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.7434767776207417,
+      "restart_teacher_dice": 0.7648114620563284,
+      "weights": [
+        0.0009536063313568787,
+        0.00622160049410085,
+        2.7365280978604556e-05,
+        0.0017150542671718894,
+        0.00778389682673956,
+        0.9830816831898324,
+        4.206174632514249e-05,
+        0.0001747318634946766
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7646874735949377,
+        "uniform": 0.7405359432386459,
+        "top1": 0.7648114620563284,
+        "top2": 0.7643850371191413,
+        "top3": 0.7646318754401064
+      },
+      "val_dice": 0.3714069295182818,
+      "val_vi": 0.3523276532410457,
+      "val_ece": 0.03809157325395153
+    },
+    {
+      "round": 2,
+      "n_labeled": 58,
+      "selection_seconds": 0.2,
+      "round_seconds": 20.7,
+      "teacher_dice_labeled": [
+        0.27233939627165826,
+        0.5648992002411293,
+        0.18494126905476696,
+        0.2950788044938515,
+        0.5694094632225994,
+        0.7901282813438308,
+        0.10961051669571481,
+        0.20814065934578377
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 357977,
+      "evidence_nonempty_patches": 40,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0008651839537822588,
+        0.004652891610929437,
+        2.5107694955276247e-05,
+        0.0016213688563193685,
+        0.005397791634150245,
+        0.9872621406952384,
+        3.27122188146369e-05,
+        0.00014280333581031675
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7900516454534338,
+        "uniform": 0.744561558923848,
+        "top1": 0.7901282813438308,
+        "top2": 0.7899526344089269,
+        "top3": 0.7900473160877718
+      },
+      "pseudo_student_dice": 0.5020917211634688,
+      "pseudo_ensemble_dice": 0.8136604464983836,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.8627172784603788,
+      "val_vi": 0.14816620428021832,
+      "val_ece": 0.020074298774123814
+    },
+    {
+      "round": 3,
+      "n_labeled": 87,
+      "selection_seconds": 0.2,
+      "round_seconds": 30.8,
+      "teacher_dice_labeled": [
+        0.2683289214714117,
+        0.577856181422308,
+        0.18673087992781356,
+        0.3018313761883449,
+        0.5746961679042322,
+        0.777125265724869,
+        0.11037391096234828,
+        0.20553076412792937
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 538912,
+      "evidence_nonempty_patches": 61,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0009438330908227715,
+        0.006014484421225186,
+        2.9025722514262033e-05,
+        0.0019697535289780604,
+        0.006462194076105672,
+        0.9843852962889938,
+        3.7430824711988296e-05,
+        0.00015798204664822644
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7771597736566964,
+        "uniform": 0.7494724702493365,
+        "top1": 0.777125265724869,
+        "top2": 0.7773284324921306,
+        "top3": 0.7771113848505545
+      },
+      "pseudo_student_dice": 0.8181573209193873,
+      "pseudo_ensemble_dice": 0.7506062527980898,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.8649044370034481,
+      "val_vi": 0.1438270705034725,
+      "val_ece": 0.0221553722447091
+    },
+    {
+      "round": 4,
+      "n_labeled": 116,
+      "selection_seconds": 0.2,
+      "round_seconds": 127.4,
+      "teacher_dice_labeled": [
+        0.3523551657918085,
+        0.6195955910426016,
+        0.21702358946896538,
+        0.35817812539551996,
+        0.5849809963185317,
+        0.8099093730446942,
+        0.21149852979362696,
+        0.249944372842695
+      ],
+      "teacher_dice_pool": [
+        0.3963038925311522,
+        0.643366416580547,
+        0.2412226943807228,
+        0.3803305564344355,
+        0.6006591180223585,
+        0.8628850136387899,
+        0.2966821474784586,
+        0.26739609073689796
+      ],
+      "evidence_fg_pixels": 840112,
+      "evidence_nonempty_patches": 85,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0015748239507074387,
+        0.0065748496506743736,
+        2.8298151940109134e-05,
+        0.0024919434282833783,
+        0.00515776924456544,
+        0.9839208328705071,
+        7.410090370459648e-05,
+        0.0001773817996174603
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.8100702315447105,
+        "uniform": 0.7858644717533334,
+        "top1": 0.8099093730446942,
+        "top2": 0.8102884590298836,
+        "top3": 0.8099587844188316
+      },
+      "pseudo_student_dice": 0.7662125214904288,
+      "pseudo_ensemble_dice": 0.7188745455139126,
+      "pseudo_source": "student",
+      "consistency_scale": 0.887,
+      "val_dice": 0.8627058602543363,
+      "val_vi": 0.14709296460397747,
+      "val_ece": 0.02107210011131645
+    },
+    {
+      "round": 5,
+      "n_labeled": 145,
+      "selection_seconds": 0.2,
+      "round_seconds": 135.4,
+      "teacher_dice_labeled": [
+        0.3434458390283317,
+        0.6024990757453136,
+        0.2141995135026267,
+        0.3527860406324941,
+        0.5688136071369587,
+        0.7969131606271672,
+        0.20414479797548948,
+        0.24496127493124809
+      ],
+      "teacher_dice_pool": [
+        0.38700939335882084,
+        0.637700200318136,
+        0.2451388677680745,
+        0.3823338784049224,
+        0.5968550856440995,
+        0.8627161787334623,
+        0.28581341353040957,
+        0.26015229074359414
+      ],
+      "evidence_fg_pixels": 1036034,
+      "evidence_nonempty_patches": 103,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0016407488958869057,
+        0.006311599420629153,
+        3.13325753000206e-05,
+        0.0026892023139674236,
+        0.004997475263623532,
+        0.9840590206740554,
+        7.841325407623796e-05,
+        0.0001922076024614262
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7971050077151005,
+        "uniform": 0.7675423537446693,
+        "top1": 0.7969131606271672,
+        "top2": 0.7972941217041166,
+        "top3": 0.7969877830674019
+      },
+      "pseudo_student_dice": 0.649230568272293,
+      "pseudo_ensemble_dice": 0.603722645216952,
+      "pseudo_source": "student",
+      "consistency_scale": 0.497,
+      "val_dice": 0.8780933225724049,
+      "val_vi": 0.15112189309925872,
+      "val_ece": 0.016665274923035406
+    },
+    {
+      "round": 6,
+      "n_labeled": 174,
+      "selection_seconds": 0.2,
+      "round_seconds": 142.8,
+      "teacher_dice_labeled": [
+        0.3501263783945856,
+        0.6050864237945479,
+        0.21269241472651582,
+        0.35459013382319676,
+        0.5658740646511347,
+        0.7906464270163391,
+        0.2161101273220039,
+        0.25254151656279217
+      ],
+      "teacher_dice_pool": [
+        0.40178570744004233,
+        0.635485317205476,
+        0.23930938140955102,
+        0.37514848519991184,
+        0.58683289956519,
+        0.8480063602697288,
+        0.30812144608107045,
+        0.2764468271369436
+      ],
+      "evidence_fg_pixels": 1249546,
+      "evidence_nonempty_patches": 127,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0018652112324314258,
+        0.00688729721663382,
+        3.281886314378884e-05,
+        0.002911596851577237,
+        0.00516008754641597,
+        0.9828285328627763,
+        9.39781582817084e-05,
+        0.00022047726873971913
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.7907800643613214,
+        "uniform": 0.7685180324869638,
+        "top1": 0.7906464270163391,
+        "top2": 0.7910706954472992,
+        "top3": 0.7906376772800607
+      },
+      "pseudo_student_dice": 0.797396310946261,
+      "pseudo_ensemble_dice": 0.7435812984466436,
+      "pseudo_source": "student",
+      "consistency_scale": 0.991,
+      "val_dice": 0.8895698960959386,
+      "val_vi": 0.14886280621004105,
+      "val_ece": 0.013738702604143803
+    },
+    {
+      "round": 7,
+      "n_labeled": 174,
+      "selection_seconds": 0.0,
+      "round_seconds": 66.1,
+      "final_consolidation": true,
+      "val_dice": 0.8933540398949001,
+      "val_vi": 0.1477702542557411,
+      "val_ece": 0.01187762032777213
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.00029523523234020787,
+    "auroc_top20_error": 0.4987465149229856
+  },
+  "al_seconds": 662.4,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.0018652112324314258,
+    0.00688729721663382,
+    3.281886314378884e-05,
+    0.002911596851577237,
+    0.00516008754641597,
+    0.9828285328627763,
+    9.39781582817084e-05,
+    0.00022047726873971913
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9828285328627763,
+    "entropy": 0.11027510731915921,
+    "effective_sources": 1.116585208972361,
+    "min_weight": 3.281886314378884e-05
+  },
+  "top_source_final": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.8717935908930711,
+    "patch_iou": 0.8079621865404086,
+    "patch_precision": 0.8811307975775252,
+    "patch_recall": 0.8911383126144642,
+    "patch_vi": 0.1763703127715098,
+    "n_patches": 768,
+    "cal_ece": 0.010235559901710106,
+    "cal_nll": 0.07946818594560806,
+    "cal_brier": 0.022279491154928013,
+    "cal_mean_confidence": 0.9805627915840384,
+    "cal_accuracy": 0.9704310099283854,
+    "cal_predicted_prevalence": 0.07677459716796875,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07612501553469163
+  },
+  "test_notta": {
+    "patch_dice": 0.8717935908930711,
+    "patch_iou": 0.8079621865404086,
+    "patch_precision": 0.8811307975775252,
+    "patch_recall": 0.8911383126144642,
+    "patch_vi": 0.1763703127715098,
+    "n_patches": 768,
+    "cal_ece": 0.010235559901710106,
+    "cal_nll": 0.07946818594560806,
+    "cal_brier": 0.022279491154928013,
+    "cal_mean_confidence": 0.9805627915840384,
+    "cal_accuracy": 0.9704310099283854,
+    "cal_predicted_prevalence": 0.07677459716796875,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07612501553469163
+  },
+  "test": {
+    "patch_dice": 0.8733024516459945,
+    "patch_iou": 0.8119174567416984,
+    "patch_precision": 0.8885141746129198,
+    "patch_recall": 0.8888266529845109,
+    "patch_vi": 0.1734969541682323,
+    "n_patches": 768,
+    "image_dice": 0.8025776625259756,
+    "image_iou": 0.6774712731283673,
+    "image_precision": 0.7886269324253391,
+    "image_recall": 0.8227675597721614,
+    "image_vi": 0.19818246493296066,
+    "n_images": 6,
+    "cal_ece": 0.00853267805767482,
+    "cal_nll": 0.07485122535739094,
+    "cal_brier": 0.021577405959943684,
+    "cal_mean_confidence": 0.9795055914134073,
+    "cal_accuracy": 0.9710172017415365,
+    "cal_predicted_prevalence": 0.07612228393554688,
+    "cal_true_prevalence": 0.07175954182942708,
+    "cal_mean_probability": 0.07594731116540443
+  },
+  "eval_seconds": 12.2,
+  "total_seconds": 903.9,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.13,
+        "peak_cpu_rss_mb": 1027.6,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 26.07,
+        "peak_cpu_rss_mb": 1808.4,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 200.36,
+        "peak_cpu_rss_mb": 2938.1,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4618.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 662.37,
+        "peak_cpu_rss_mb": 2938.1,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4626.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 12.25,
+        "peak_cpu_rss_mb": 2938.1,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4626.0,
+        "patches": 768,
+        "images": 6,
+        "patches_per_second": 62.7,
+        "seconds_per_image": 2.042
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2938.1,
+    "total_seconds": 902.2
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-random_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-random/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,29,26.6,128.7,"[0.2571777942968238, 0.5690605504275248, 0.1686588604013407, 0.2758037194420113, 0.5811237001396865, 0.7648114620563284, 0.10985703762520847, 0.20342659662893037]",,179039,21,True,kept,0.7434767776207417,0.7648114620563284,"[0.0009536063313568787, 0.00622160049410085, 2.7365280978604556e-05, 0.0017150542671718894, 0.00778389682673956, 0.9830816831898324, 4.206174632514249e-05, 0.0001747318634946766]",mixture,"{'mixture': 0.7646874735949377, 'uniform': 0.7405359432386459, 'top1': 0.7648114620563284, 'top2': 0.7643850371191413, 'top3': 0.7646318754401064}",0.3714069295182818,0.3523276532410457,0.03809157325395153,,,,,
+2,58,0.2,20.7,"[0.27233939627165826, 0.5648992002411293, 0.18494126905476696, 0.2950788044938515, 0.5694094632225994, 0.7901282813438308, 0.10961051669571481, 0.20814065934578377]",,357977,40,False,,,,"[0.0008651839537822588, 0.004652891610929437, 2.5107694955276247e-05, 0.0016213688563193685, 0.005397791634150245, 0.9872621406952384, 3.27122188146369e-05, 0.00014280333581031675]",mixture,"{'mixture': 0.7900516454534338, 'uniform': 0.744561558923848, 'top1': 0.7901282813438308, 'top2': 0.7899526344089269, 'top3': 0.7900473160877718}",0.8627172784603788,0.14816620428021832,0.020074298774123814,0.5020917211634688,0.8136604464983836,ensemble,1.0,
+3,87,0.2,30.8,"[0.2683289214714117, 0.577856181422308, 0.18673087992781356, 0.3018313761883449, 0.5746961679042322, 0.777125265724869, 0.11037391096234828, 0.20553076412792937]",,538912,61,False,,,,"[0.0009438330908227715, 0.006014484421225186, 2.9025722514262033e-05, 0.0019697535289780604, 0.006462194076105672, 0.9843852962889938, 3.7430824711988296e-05, 0.00015798204664822644]",mixture,"{'mixture': 0.7771597736566964, 'uniform': 0.7494724702493365, 'top1': 0.777125265724869, 'top2': 0.7773284324921306, 'top3': 0.7771113848505545}",0.8649044370034481,0.1438270705034725,0.0221553722447091,0.8181573209193873,0.7506062527980898,student,1.0,
+4,116,0.2,127.4,"[0.3523551657918085, 0.6195955910426016, 0.21702358946896538, 0.35817812539551996, 0.5849809963185317, 0.8099093730446942, 0.21149852979362696, 0.249944372842695]","[0.3963038925311522, 0.643366416580547, 0.2412226943807228, 0.3803305564344355, 0.6006591180223585, 0.8628850136387899, 0.2966821474784586, 0.26739609073689796]",840112,85,False,,,,"[0.0015748239507074387, 0.0065748496506743736, 2.8298151940109134e-05, 0.0024919434282833783, 0.00515776924456544, 0.9839208328705071, 7.410090370459648e-05, 0.0001773817996174603]",mixture,"{'mixture': 0.8100702315447105, 'uniform': 0.7858644717533334, 'top1': 0.8099093730446942, 'top2': 0.8102884590298836, 'top3': 0.8099587844188316}",0.8627058602543363,0.14709296460397747,0.02107210011131645,0.7662125214904288,0.7188745455139126,student,0.887,
+5,145,0.2,135.4,"[0.3434458390283317, 0.6024990757453136, 0.2141995135026267, 0.3527860406324941, 0.5688136071369587, 0.7969131606271672, 0.20414479797548948, 0.24496127493124809]","[0.38700939335882084, 0.637700200318136, 0.2451388677680745, 0.3823338784049224, 0.5968550856440995, 0.8627161787334623, 0.28581341353040957, 0.26015229074359414]",1036034,103,False,,,,"[0.0016407488958869057, 0.006311599420629153, 3.13325753000206e-05, 0.0026892023139674236, 0.004997475263623532, 0.9840590206740554, 7.841325407623796e-05, 0.0001922076024614262]",mixture,"{'mixture': 0.7971050077151005, 'uniform': 0.7675423537446693, 'top1': 0.7969131606271672, 'top2': 0.7972941217041166, 'top3': 0.7969877830674019}",0.8780933225724049,0.15112189309925872,0.016665274923035406,0.649230568272293,0.603722645216952,student,0.497,
+6,174,0.2,142.8,"[0.3501263783945856, 0.6050864237945479, 0.21269241472651582, 0.35459013382319676, 0.5658740646511347, 0.7906464270163391, 0.2161101273220039, 0.25254151656279217]","[0.40178570744004233, 0.635485317205476, 0.23930938140955102, 0.37514848519991184, 0.58683289956519, 0.8480063602697288, 0.30812144608107045, 0.2764468271369436]",1249546,127,False,,,,"[0.0018652112324314258, 0.00688729721663382, 3.281886314378884e-05, 0.002911596851577237, 0.00516008754641597, 0.9828285328627763, 9.39781582817084e-05, 0.00022047726873971913]",mixture,"{'mixture': 0.7907800643613214, 'uniform': 0.7685180324869638, 'top1': 0.7906464270163391, 'top2': 0.7910706954472992, 'top3': 0.7906376772800607}",0.8895698960959386,0.14886280621004105,0.013738702604143803,0.797396310946261,0.7435812984466436,student,0.991,
+7,174,0.0,66.1,,,,,,,,,,,,0.8933540398949001,0.1477702542557411,0.01187762032777213,,,,,True

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-2/log.txt

@@ -0,0 +1,104 @@
+[07:11:51][      0s][START] target=snemi3d variant=macs-v11-norestart-acq-margin budget=5.0% seed=2 | {"task": "membrane", "target": "snemi3d", "variant": "macs-v11-norestart-acq-margin", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "margin", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[07:11:51][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[07:11:52][      0s][DATA] snemi3d: train=3430 val=735 test=735 patches
+[07:11:53][      2s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, mouse-brain=4.08
+[07:11:53][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[07:11:59][      8s][WEIGHTS] mixture: c-elegans=0.005, cerebellum=0.058, flywire=0.394, berghia=0.027, microns=0.477, oct-glia=0.007, oct-sfl=0.011, mouse-brain=0.021 | eff.sources=3.24 | MMD2 mix=0.0601 uniform=0.1529
+[07:11:59][      8s][WEIGHTS] per-source: c-elegans=0.003, cerebellum=0.308, flywire=0.031, berghia=0.005, microns=0.624, oct-glia=0.021, oct-sfl=0.005, mouse-brain=0.002 | min-MMD source (NeuroADDA rule) = microns
+[07:12:17][     26s][TEACHERS] target-val Dice: c-elegans=0.412, cerebellum=0.473, flywire=0.196, berghia=0.310, microns=0.589, oct-glia=0.137, oct-sfl=0.305, mouse-brain=0.350
+[07:12:17][     26s][TEACHERS] ensemble target-val Dice: mixture=0.421, mixture@raw=0.463, mixture@logit=0.367, per_source=0.589, per_source@raw=0.591, per_source@logit=0.578, uniform=0.411, uniform@raw=0.467, uniform@logit=0.228
+[07:12:17][     26s][WEIGHTS] using mixture: c-elegans=0.005, cerebellum=0.058, flywire=0.394, berghia=0.027, microns=0.477, oct-glia=0.007, oct-sfl=0.011, mouse-brain=0.021
+[07:12:17][     26s][INIT] student initialised from teacher microns (weight 0.477)
+[07:12:18][     27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[07:12:18][     27s][KD] distilling the mixture ensemble (prob space) into the student
+[07:12:57][     66s][KD] soft labels for 3430 patches from 8 teachers in 39s
+[07:13:29][     97s][KD] epoch 1/5 | patches=3430 | loss=0.4413 | 31s
+[07:14:00][    129s][KD] epoch 2/5 | patches=3430 | loss=0.3718 | 31s
+[07:14:31][    159s][KD] epoch 3/5 | patches=3430 | loss=0.3562 | 31s
+[07:15:02][    190s][KD] epoch 4/5 | patches=3430 | loss=0.3503 | 31s
+[07:15:32][    221s][KD] epoch 5/5 | patches=3430 | loss=0.3474 | 31s
+[07:15:34][    222s][ROUND0] before any target label: val_dice=0.3537 val_vi=0.6275
+[07:15:34][    222s][AL] patches=3430 (pool 3430 after dropping blank; free blank negatives: True) | budget=171 (5.0%) | rounds=6 | per_round=28 | acquisition=margin | first_round=stratified
+[07:15:39][    228s][AL] round-1 uncertainty quality on 3430 patches: {'spearman_uncertainty_error': 0.3417284798148, 'auroc_top20_error': 0.7648460675398855}
+[07:16:02][    251s][AL] round 1: stratified cold start: 14 patches from the membrane-rich half (ensemble prevalence >= 0.156) + 14 from the membrane-poor half of 2000 candidates
+[07:16:04][    253s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.512, uniform=0.372, top1=0.513, top2=0.514, top3=0.514 -> using mixture
+[07:16:04][    253s][WEIGHTS] round 1: student kept (labeled Dice 0.373 vs best teacher microns 0.513); distilling the corrected ensemble into it
+[07:16:04][    253s][WEIGHTS] round 1: teacher Dice on 28 labeled patches + 0 blank (229858 fg px): c-elegans=0.298, cerebellum=0.401, flywire=0.209, berghia=0.345, microns=0.513, oct-glia=0.175, oct-sfl=0.199, mouse-brain=0.227 | weights: c-elegans=0.001, cerebellum=0.036, flywire=0.036, berghia=0.010, microns=0.913, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002
+[07:16:43][    292s][KD] soft labels for 3402 patches from 8 teachers in 39s
+[07:17:14][    322s][KD] epoch 1/2 | patches=3402 | loss=0.2523 | 31s
+[07:17:44][    353s][KD] epoch 2/2 | patches=3402 | loss=0.2493 | 31s
+[07:18:07][    376s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[07:18:08][    377s][FT] round 1/6 epoch 1/5 | patches=28 | loss=1.0783
+[07:18:09][    378s][FT] round 1/6 epoch 2/5 | patches=28 | loss=0.9834
+[07:18:10][    379s][FT] round 1/6 epoch 3/5 | patches=28 | loss=0.9413
+[07:18:11][    379s][FT] round 1/6 epoch 4/5 | patches=28 | loss=0.9016
+[07:18:12][    380s][FT] round 1/6 epoch 5/5 | patches=28 | loss=0.8905
+[07:18:13][    382s][AL] round 1/6 | labeled=28 | selection_seconds=28.4000 round_seconds=158.1000 restart_student_dice=0.3728 restart_teacher_dice=0.5132 val_dice=0.7017 val_vi=0.5973 val_ece=0.0128
+[07:18:19][    387s][AL] round 2: 2/3402 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:18:22][    390s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.509, uniform=0.304, top1=0.510, top2=0.512, top3=0.512 -> using mixture
+[07:18:22][    390s][WEIGHTS] round 2: teacher Dice on 56 labeled patches + 0 blank (537619 fg px): c-elegans=0.343, cerebellum=0.380, flywire=0.189, berghia=0.273, microns=0.510, oct-glia=0.119, oct-sfl=0.162, mouse-brain=0.265 | weights: c-elegans=0.002, cerebellum=0.030, flywire=0.031, berghia=0.005, microns=0.927, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[07:18:27][    395s][CONS] round 2: on the 28 new patches student Dice=0.547 vs ensemble 0.506 -> pseudo-labels from the student | consistency x0.16
+[07:18:28][    396s][FT] round 2/6 epoch 1/5 | patches=56 | loss=0.8317
+[07:18:29][    398s][FT] round 2/6 epoch 2/5 | patches=56 | loss=0.7916
+[07:18:30][    399s][FT] round 2/6 epoch 3/5 | patches=56 | loss=0.7590
+[07:18:32][    400s][FT] round 2/6 epoch 4/5 | patches=56 | loss=0.7264
+[07:18:33][    402s][FT] round 2/6 epoch 5/5 | patches=56 | loss=0.7075
+[07:18:35][    403s][AL] round 2/6 | labeled=56 | selection_seconds=5.5000 round_seconds=19.8000 pseudo_student_dice=0.5472 pseudo_ensemble_dice=0.5061 consistency_scale=0.1570 val_dice=0.7407 val_vi=0.5701 val_ece=0.0149
+[07:19:28][    457s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.394, cerebellum=0.481, flywire=0.238, berghia=0.381, microns=0.664, oct-glia=0.201, oct-sfl=0.279, mouse-brain=0.316 | blended with labeled Dice at 0.50
+[07:20:14][    502s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.571, uniform=0.369, top1=0.571, top2=0.572, top3=0.573 -> using mixture
+[07:20:14][    502s][WEIGHTS] round 3: teacher Dice on 84 labeled patches + 0 blank (878752 fg px): c-elegans=0.378, cerebellum=0.438, flywire=0.208, berghia=0.332, microns=0.571, oct-glia=0.162, oct-sfl=0.216, mouse-brain=0.300 | weights: c-elegans=0.001, cerebellum=0.030, flywire=0.020, berghia=0.005, microns=0.939, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[07:20:18][    507s][CONS] round 3: on the 28 new patches student Dice=0.614 vs ensemble 0.421 -> pseudo-labels from the student | consistency x0.38
+[07:20:20][    509s][FT] round 3/6 epoch 1/5 | patches=84 | loss=0.8095
+[07:20:22][    511s][FT] round 3/6 epoch 2/5 | patches=84 | loss=0.8034
+[07:20:24][    512s][FT] round 3/6 epoch 3/5 | patches=84 | loss=0.7503
+[07:20:26][    514s][FT] round 3/6 epoch 4/5 | patches=84 | loss=0.7570
+[07:20:27][    516s][FT] round 3/6 epoch 5/5 | patches=84 | loss=0.7426
+[07:20:29][    517s][AL] round 3/6 | labeled=84 | selection_seconds=5.3000 round_seconds=112.9000 pseudo_student_dice=0.6139 pseudo_ensemble_dice=0.4207 consistency_scale=0.3800 val_dice=0.7468 val_vi=0.5417 val_ece=0.0109
+[07:21:23][    571s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.379, cerebellum=0.467, flywire=0.230, berghia=0.373, microns=0.646, oct-glia=0.192, oct-sfl=0.280, mouse-brain=0.304 | blended with labeled Dice at 0.50
+[07:22:09][    617s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.561, uniform=0.366, top1=0.561, top2=0.562, top3=0.562 -> using mixture
+[07:22:09][    617s][WEIGHTS] round 4: teacher Dice on 112 labeled patches + 0 blank (1174561 fg px): c-elegans=0.371, cerebellum=0.430, flywire=0.202, berghia=0.328, microns=0.561, oct-glia=0.162, oct-sfl=0.221, mouse-brain=0.295 | weights: c-elegans=0.002, cerebellum=0.031, flywire=0.021, berghia=0.005, microns=0.938, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[07:22:14][    622s][CONS] round 4: on the 28 new patches student Dice=0.645 vs ensemble 0.469 -> pseudo-labels from the student | consistency x0.48
+[07:22:16][    625s][FT] round 4/6 epoch 1/5 | patches=112 | loss=0.7597
+[07:22:18][    627s][FT] round 4/6 epoch 2/5 | patches=112 | loss=0.7372
+[07:22:21][    629s][FT] round 4/6 epoch 3/5 | patches=112 | loss=0.7457
+[07:22:23][    631s][FT] round 4/6 epoch 4/5 | patches=112 | loss=0.7221
+[07:22:25][    634s][FT] round 4/6 epoch 5/5 | patches=112 | loss=0.7071
+[07:22:27][    635s][AL] round 4/6 | labeled=112 | selection_seconds=5.3000 round_seconds=116.2000 pseudo_student_dice=0.6452 pseudo_ensemble_dice=0.4695 consistency_scale=0.4840 val_dice=0.7543 val_vi=0.5300 val_ece=0.0126
+[07:23:21][    690s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.382, cerebellum=0.470, flywire=0.233, berghia=0.378, microns=0.641, oct-glia=0.190, oct-sfl=0.278, mouse-brain=0.304 | blended with labeled Dice at 0.50
+[07:24:08][    736s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.557, uniform=0.364, top1=0.558, top2=0.559, top3=0.559 -> using mixture
+[07:24:08][    736s][WEIGHTS] round 5: teacher Dice on 140 labeled patches + 0 blank (1440921 fg px): c-elegans=0.369, cerebellum=0.432, flywire=0.202, berghia=0.328, microns=0.558, oct-glia=0.162, oct-sfl=0.224, mouse-brain=0.293 | weights: c-elegans=0.002, cerebellum=0.032, flywire=0.022, berghia=0.005, microns=0.935, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[07:24:13][    741s][CONS] round 5: on the 28 new patches student Dice=0.644 vs ensemble 0.466 -> pseudo-labels from the student | consistency x0.48
+[07:24:15][    744s][FT] round 5/6 epoch 1/5 | patches=140 | loss=0.7219
+[07:24:18][    747s][FT] round 5/6 epoch 2/5 | patches=140 | loss=0.7203
+[07:24:21][    749s][FT] round 5/6 epoch 3/5 | patches=140 | loss=0.7187
+[07:24:24][    752s][FT] round 5/6 epoch 4/5 | patches=140 | loss=0.7061
+[07:24:26][    755s][FT] round 5/6 epoch 5/5 | patches=140 | loss=0.7081
+[07:24:28][    756s][AL] round 5/6 | labeled=140 | selection_seconds=5.4000 round_seconds=119.8000 pseudo_student_dice=0.6441 pseudo_ensemble_dice=0.4658 consistency_scale=0.4800 val_dice=0.7540 val_vi=0.5364 val_ece=0.0150
+[07:25:23][    811s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.391, cerebellum=0.477, flywire=0.236, berghia=0.380, microns=0.644, oct-glia=0.193, oct-sfl=0.280, mouse-brain=0.311 | blended with labeled Dice at 0.50
+[07:26:10][    859s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.563, uniform=0.370, top1=0.564, top2=0.565, top3=0.565 -> using mixture
+[07:26:10][    859s][WEIGHTS] round 6: teacher Dice on 168 labeled patches + 0 blank (1701154 fg px): c-elegans=0.373, cerebellum=0.436, flywire=0.205, berghia=0.332, microns=0.564, oct-glia=0.166, oct-sfl=0.228, mouse-brain=0.294 | weights: c-elegans=0.002, cerebellum=0.031, flywire=0.021, berghia=0.005, microns=0.937, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[07:26:15][    863s][CONS] round 6: on the 28 new patches student Dice=0.666 vs ensemble 0.527 -> pseudo-labels from the student | consistency x0.55
+[07:26:18][    867s][FT] round 6/6 epoch 1/5 | patches=168 | loss=0.7158
+[07:26:21][    870s][FT] round 6/6 epoch 2/5 | patches=168 | loss=0.6940
+[07:26:24][    873s][FT] round 6/6 epoch 3/5 | patches=168 | loss=0.6996
+[07:26:28][    876s][FT] round 6/6 epoch 4/5 | patches=168 | loss=0.6985
+[07:26:31][    879s][FT] round 6/6 epoch 5/5 | patches=168 | loss=0.6847
+[07:26:32][    881s][AL] round 6/6 | labeled=168 | selection_seconds=5.2000 round_seconds=123.1000 pseudo_student_dice=0.6663 pseudo_ensemble_dice=0.5269 consistency_scale=0.5540 val_dice=0.7572 val_vi=0.5271 val_ece=0.0199
+[07:26:35][    884s][CONS] final consolidation: pseudo-labels refreshed from the student
+[07:26:39][    887s][FT] final 1/10 epoch 1/1 | patches=168 | loss=0.6821
+[07:26:42][    890s][FT] final 2/10 epoch 1/1 | patches=168 | loss=0.6618
+[07:26:45][    893s][FT] final 3/10 epoch 1/1 | patches=168 | loss=0.6616
+[07:26:48][    897s][FT] final 4/10 epoch 1/1 | patches=168 | loss=0.6538
+[07:26:51][    900s][FT] final 5/10 epoch 1/1 | patches=168 | loss=0.6508
+[07:26:54][    903s][FT] final 6/10 epoch 1/1 | patches=168 | loss=0.6438
+[07:26:58][    906s][FT] final 7/10 epoch 1/1 | patches=168 | loss=0.6371
+[07:27:01][    909s][FT] final 8/10 epoch 1/1 | patches=168 | loss=0.6338
+[07:27:04][    913s][FT] final 9/10 epoch 1/1 | patches=168 | loss=0.6283
+[07:27:07][    916s][FT] final 10/10 epoch 1/1 | patches=168 | loss=0.6262
+[07:27:09][    917s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=35.1000 val_dice=0.7544 val_vi=0.5236 val_ece=0.0221
+[07:27:09][    917s][WEIGHTS] final: c-elegans=0.002, cerebellum=0.031, flywire=0.021, berghia=0.005, microns=0.937, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[07:27:19][    928s][TEST] raw student (no EMA, no TTA): patch_dice=0.7464 patch_vi=0.5306
+[07:27:21][    929s][TEST] final model without TTA: patch_dice=0.7464 patch_vi=0.5306
+[07:27:21][    929s][RESOURCES] {"phases": {"setup": {"seconds": 1.13, "peak_cpu_rss_mb": 1027.0, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 25.21, "peak_cpu_rss_mb": 1809.0, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 194.08, "peak_cpu_rss_mb": 2895.7, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 695.12, "peak_cpu_rss_mb": 2895.7, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 11.65, "peak_cpu_rss_mb": 2895.7, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 735, "images": 15, "patches_per_second": 63.1, "seconds_per_image": 0.777}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2895.7, "total_seconds": 927.2}
+[07:27:21][    929s][TEST] patch_dice=0.7569 patch_iou=0.6183 patch_precision=0.7430 patch_recall=0.7744 patch_vi=0.5125 image_dice=0.8046 image_iou=0.6736 image_precision=0.7945 image_recall=0.8155 image_vi=0.5193 cal_ece=0.0111 cal_nll=0.1813 cal_brier=0.0572 cal_mean_confidence=0.9282 cal_accuracy=0.9172 cal_predicted_prevalence=0.2129 cal_true_prevalence=0.2059 cal_mean_probability=0.2114
+[07:27:21][    929s][DONE] snemi3d macs-v11-norestart-acq-margin b5 s2 | test patch Dice=0.7569 VI=0.5125 | image Dice=0.8046 VI=0.5193 | 929s

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-2/result.json

@@ -0,0 +1,785 @@
+{
+  "target": "snemi3d",
+  "variant": "macs-v11-norestart-acq-margin",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "snemi3d",
+    "variant": "macs-v11-norestart-acq-margin",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "margin",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "snemi3d",
+      "--variants",
+      "macs-v11-norestart-acq-margin",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.00526521078604608,
+    0.05775508166244069,
+    0.39371182687582884,
+    0.027420285915093805,
+    0.47701492758994885,
+    0.006835626005792793,
+    0.011200163178997314,
+    0.020796877985851614
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.00526521078604608,
+      0.05775508166244069,
+      0.39371182687582884,
+      0.027420285915093805,
+      0.47701492758994885,
+      0.006835626005792793,
+      0.011200163178997314,
+      0.020796877985851614
+    ],
+    "per_source": [
+      0.0030099345500300633,
+      0.30826329572645517,
+      0.030874948568488214,
+      0.005310476630139808,
+      0.6243402297379833,
+      0.021312995363178842,
+      0.004803663858606227,
+      0.0020844555651184054
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.04836304661062167,
+      "mmd2_mixture": 0.06012254011242163,
+      "mmd2_uniform": 0.15285652884935502,
+      "mmd2_single_source": [
+        0.4522663178230211,
+        0.14701205466558376,
+        0.12241636968711567,
+        0.27265250499014093,
+        0.08995147555294303,
+        0.3744252998076626,
+        0.3395783286723916,
+        0.23687871798038196
+      ],
+      "median_squared_distance": 0.18663079059841653,
+      "gammas": [
+        1.3395431654037109,
+        2.6790863308074218,
+        5.3581726616148435,
+        10.716345323229687,
+        21.432690646459374
+      ],
+      "max_weight": 0.47701492758994885,
+      "entropy": 1.1759493501799958,
+      "effective_sources": 3.241218535100111,
+      "min_weight": 0.00526521078604608
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.46244015980269426,
+        0.12211161125756897,
+        0.2912827113901817,
+        0.4206979810988831,
+        0.07022524085432302,
+        0.3185314378412357,
+        0.4280722723409719,
+        0.48945233449571823
+      ],
+      "standardized": [
+        0.9323157340081956,
+        -1.3822022358410426,
+        -0.2316976633920459,
+        0.6484341051420859,
+        -1.7350728286900672,
+        -0.0463835976280083,
+        0.6985854358714068,
+        1.1160210505294779
+      ],
+      "tau": 0.5,
+      "max_weight": 0.6243402297379833,
+      "entropy": 0.9300685916208571,
+      "effective_sources": 2.534683029672754,
+      "min_weight": 0.0020844555651184054
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "microns_basil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4124719184272143,
+      0.4732143695257148,
+      0.1963993253870582,
+      0.31013775029960944,
+      0.5894614335833763,
+      0.13744118666208724,
+      0.30544280939432356,
+      0.35012843982905756
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.4205090273501111,
+      "mixture@raw": 0.46291647296981747,
+      "mixture@logit": 0.3665771544587855,
+      "per_source": 0.5892233046139178,
+      "per_source@raw": 0.5905333874987907,
+      "per_source@logit": 0.5781025615679164,
+      "uniform": 0.41121937389276464,
+      "uniform@raw": 0.46695544697394986,
+      "uniform@logit": 0.22779528784246278
+    },
+    "val_patches": 735
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    4.075904369354248
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.47701492758994885,
+    "entropy": 1.1759493501799958,
+    "effective_sources": 3.241218535100111,
+    "min_weight": 0.00526521078604608
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.3571428571428572,
+    "spearman_weight_taxonomy": 0.50709255283711,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 194.1,
+  "round0_val": {
+    "val_dice": 0.35365409409657816,
+    "val_vi": 0.627529035005998,
+    "val_ece": 0.07640721373388315
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 28,
+      "selection_seconds": 28.4,
+      "round_seconds": 158.1,
+      "teacher_dice_labeled": [
+        0.2979926532667371,
+        0.40138560205746643,
+        0.2091063280267666,
+        0.345169021450426,
+        0.5132033840553815,
+        0.1750097104074813,
+        0.1991102608484407,
+        0.22650021516746904
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 229858,
+      "evidence_nonempty_patches": 28,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.3728225794885769,
+      "restart_teacher_dice": 0.5132033840553815,
+      "weights": [
+        0.0011717197029955054,
+        0.036143320102722025,
+        0.03602118748512524,
+        0.009780573323141589,
+        0.9132470305826798,
+        0.0004447112636304281,
+        0.0009272386944261949,
+        0.002264218845279129
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5124341043740674,
+        "uniform": 0.371832611058189,
+        "top1": 0.5132033840553815,
+        "top2": 0.5140566832535806,
+        "top3": 0.5140133725772849
+      },
+      "val_dice": 0.7016877048595026,
+      "val_vi": 0.5972861693259305,
+      "val_ece": 0.012805830910195319
+    },
+    {
+      "round": 2,
+      "n_labeled": 56,
+      "selection_seconds": 5.5,
+      "round_seconds": 19.8,
+      "teacher_dice_labeled": [
+        0.3432294869846446,
+        0.38007716816165,
+        0.1893160272069609,
+        0.27325800973621645,
+        0.5104215693825217,
+        0.11859077059313591,
+        0.16161456913007907,
+        0.2653818534052973
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 537619,
+      "evidence_nonempty_patches": 56,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.001923155375394908,
+        0.03049413082456546,
+        0.030855984596669918,
+        0.004974945473752838,
+        0.9273348235211605,
+        0.0002641090935606338,
+        0.0006653947874493764,
+        0.003487456327446416
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.509141392934992,
+        "uniform": 0.30389289820522647,
+        "top1": 0.5104215693825217,
+        "top2": 0.5115149870379297,
+        "top3": 0.5115711424860857
+      },
+      "pseudo_student_dice": 0.5472098965968067,
+      "pseudo_ensemble_dice": 0.506101874934941,
+      "pseudo_source": "student",
+      "consistency_scale": 0.157,
+      "val_dice": 0.7407435071089333,
+      "val_vi": 0.5701167948241403,
+      "val_ece": 0.014880619563059446
+    },
+    {
+      "round": 3,
+      "n_labeled": 84,
+      "selection_seconds": 5.3,
+      "round_seconds": 112.9,
+      "teacher_dice_labeled": [
+        0.37771424181735946,
+        0.4383924755772515,
+        0.20803697104167745,
+        0.3322271896816488,
+        0.5713981309510543,
+        0.1622958007666203,
+        0.21629404144420006,
+        0.30029101442837863
+      ],
+      "teacher_dice_pool": [
+        0.3939879167492122,
+        0.48133117792688046,
+        0.2378871834264335,
+        0.38115292524663485,
+        0.6639196547742888,
+        0.2009226453097915,
+        0.2786374526977216,
+        0.31603711564037745
+      ],
+      "evidence_fg_pixels": 878752,
+      "evidence_nonempty_patches": 84,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0014948062507949434,
+        0.030080258640846264,
+        0.0204855948778826,
+        0.004939617999442016,
+        0.9394194674567167,
+        0.0002251111641117732,
+        0.0006329282351913225,
+        0.002722215375014302
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.570805470117565,
+        "uniform": 0.3686947077957518,
+        "top1": 0.5713981309510543,
+        "top2": 0.5724235003227502,
+        "top3": 0.5725081953043248
+      },
+      "pseudo_student_dice": 0.6138777505811737,
+      "pseudo_ensemble_dice": 0.42072085731863107,
+      "pseudo_source": "student",
+      "consistency_scale": 0.38,
+      "val_dice": 0.7468245490384356,
+      "val_vi": 0.5417238671053858,
+      "val_ece": 0.010860615428633374
+    },
+    {
+      "round": 4,
+      "n_labeled": 112,
+      "selection_seconds": 5.3,
+      "round_seconds": 116.2,
+      "teacher_dice_labeled": [
+        0.37113155177689233,
+        0.4304098004089352,
+        0.20188035699616788,
+        0.32804902477829634,
+        0.5613490894740727,
+        0.16195989381006395,
+        0.22064107723935433,
+        0.2945484326528721
+      ],
+      "teacher_dice_pool": [
+        0.37930059726350224,
+        0.46679191904083284,
+        0.22973305931652954,
+        0.37275095341645953,
+        0.646127639407633,
+        0.19164165882853965,
+        0.2799476672005563,
+        0.3042154481257559
+      ],
+      "evidence_fg_pixels": 1174561,
+      "evidence_nonempty_patches": 112,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.00154438979007035,
+        0.030645980759803813,
+        0.021255485098501642,
+        0.005227669234419145,
+        0.9375132175239673,
+        0.0002475702667783434,
+        0.00072944689996532,
+        0.0028362404264940613
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5607148781393188,
+        "uniform": 0.3658178463120181,
+        "top1": 0.5613490894740727,
+        "top2": 0.5624084345460589,
+        "top3": 0.5624875840270268
+      },
+      "pseudo_student_dice": 0.6451945548649818,
+      "pseudo_ensemble_dice": 0.4694781440108841,
+      "pseudo_source": "student",
+      "consistency_scale": 0.484,
+      "val_dice": 0.7542581674695896,
+      "val_vi": 0.5300439451809495,
+      "val_ece": 0.012610876109984098
+    },
+    {
+      "round": 5,
+      "n_labeled": 140,
+      "selection_seconds": 5.4,
+      "round_seconds": 119.8,
+      "teacher_dice_labeled": [
+        0.3686013985558847,
+        0.4316764686766728,
+        0.2017286174831427,
+        0.32834735201508225,
+        0.5580438444411788,
+        0.16184451751433904,
+        0.22400369081247648,
+        0.292771023767665
+      ],
+      "teacher_dice_pool": [
+        0.3818583838279288,
+        0.4695261790989113,
+        0.23319827221739736,
+        0.3783978491677929,
+        0.6413947113666308,
+        0.18985071416968632,
+        0.2782163770065046,
+        0.30448063459351216
+      ],
+      "evidence_fg_pixels": 1440921,
+      "evidence_nonempty_patches": 140,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0015526486686203665,
+        0.03200215204034361,
+        0.02188349146114073,
+        0.005406401675151992,
+        0.9352495152292364,
+        0.00025497758356847564,
+        0.0007778608562707616,
+        0.0028729524856675776
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5573747664340176,
+        "uniform": 0.36380727550724445,
+        "top1": 0.5580438444411788,
+        "top2": 0.5590948972738314,
+        "top3": 0.5591800198932264
+      },
+      "pseudo_student_dice": 0.6441366353069468,
+      "pseudo_ensemble_dice": 0.4658413022561706,
+      "pseudo_source": "student",
+      "consistency_scale": 0.48,
+      "val_dice": 0.7539582315006403,
+      "val_vi": 0.536392270389954,
+      "val_ece": 0.01499025266145266
+    },
+    {
+      "round": 6,
+      "n_labeled": 168,
+      "selection_seconds": 5.2,
+      "round_seconds": 123.1,
+      "teacher_dice_labeled": [
+        0.3734322017511206,
+        0.4356543154702399,
+        0.20464087198641312,
+        0.3322555657020442,
+        0.563837343525697,
+        0.16618033768548526,
+        0.22835194680726428,
+        0.2938739092673822
+      ],
+      "teacher_dice_pool": [
+        0.39086690892472775,
+        0.4766700450989547,
+        0.23643556022313506,
+        0.3796429239120413,
+        0.6442590563252943,
+        0.1929323791472062,
+        0.28010353436049357,
+        0.3105360609364328
+      ],
+      "evidence_fg_pixels": 1701154,
+      "evidence_nonempty_patches": 168,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0015400210244406964,
+        0.03147228651218212,
+        0.021293052134217047,
+        0.005313185861618959,
+        0.9366166665504763,
+        0.00025165512057501784,
+        0.0007678204876118409,
+        0.0027453123088780123
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.563181371815908,
+        "uniform": 0.3701849952532835,
+        "top1": 0.563837343525697,
+        "top2": 0.5648706041858641,
+        "top3": 0.5649549898869545
+      },
+      "pseudo_student_dice": 0.6663378365932968,
+      "pseudo_ensemble_dice": 0.526872228608967,
+      "pseudo_source": "student",
+      "consistency_scale": 0.554,
+      "val_dice": 0.7571927107648795,
+      "val_vi": 0.5271002562939713,
+      "val_ece": 0.019880075923030854
+    },
+    {
+      "round": 7,
+      "n_labeled": 168,
+      "selection_seconds": 0.0,
+      "round_seconds": 35.1,
+      "final_consolidation": true,
+      "val_dice": 0.7543813767791763,
+      "val_vi": 0.5235559405723194,
+      "val_ece": 0.02210728108908942
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.3417284798148,
+    "auroc_top20_error": 0.7648460675398855
+  },
+  "al_seconds": 695.1,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.0015400210244406964,
+    0.03147228651218212,
+    0.021293052134217047,
+    0.005313185861618959,
+    0.9366166665504763,
+    0.00025165512057501784,
+    0.0007678204876118409,
+    0.0027453123088780123
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9366166665504763,
+    "entropy": 0.31373234511206094,
+    "effective_sources": 1.3685233955470057,
+    "min_weight": 0.00025165512057501784
+  },
+  "top_source_final": "microns_basil",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7463989317843966,
+    "patch_iou": 0.6049912614203673,
+    "patch_precision": 0.7317079484329733,
+    "patch_recall": 0.7650737901798045,
+    "patch_vi": 0.5305903317803623,
+    "n_patches": 735,
+    "cal_ece": 0.020550589690599023,
+    "cal_nll": 0.19512247212991776,
+    "cal_brier": 0.06096353383874734,
+    "cal_mean_confidence": 0.93365247106815,
+    "cal_accuracy": 0.913101881377551,
+    "cal_predicted_prevalence": 0.213156356292517,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21058684109828682
+  },
+  "test_notta": {
+    "patch_dice": 0.7463989317843966,
+    "patch_iou": 0.6049912614203673,
+    "patch_precision": 0.7317079484329733,
+    "patch_recall": 0.7650737901798045,
+    "patch_vi": 0.5305903317803623,
+    "n_patches": 735,
+    "cal_ece": 0.020550589690599023,
+    "cal_nll": 0.19512247212991776,
+    "cal_brier": 0.06096353383874734,
+    "cal_mean_confidence": 0.93365247106815,
+    "cal_accuracy": 0.913101881377551,
+    "cal_predicted_prevalence": 0.213156356292517,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21058684109828682
+  },
+  "test": {
+    "patch_dice": 0.7568723733296536,
+    "patch_iou": 0.6183452706631142,
+    "patch_precision": 0.7430031357981336,
+    "patch_recall": 0.7743687919004077,
+    "patch_vi": 0.5124967708597993,
+    "n_patches": 735,
+    "image_dice": 0.8045990482995631,
+    "image_iou": 0.6735972010434514,
+    "image_precision": 0.7944607469663268,
+    "image_recall": 0.8154881555346277,
+    "image_vi": 0.5193016730431826,
+    "n_images": 15,
+    "cal_ece": 0.011084755612734554,
+    "cal_nll": 0.1812968731235373,
+    "cal_brier": 0.05720631288259247,
+    "cal_mean_confidence": 0.9282390392011701,
+    "cal_accuracy": 0.9171542835884354,
+    "cal_predicted_prevalence": 0.2129092261904762,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.2114104320651455
+  },
+  "eval_seconds": 11.6,
+  "total_seconds": 929.1,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.13,
+        "peak_cpu_rss_mb": 1027.0,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 25.21,
+        "peak_cpu_rss_mb": 1809.0,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 194.08,
+        "peak_cpu_rss_mb": 2895.7,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4638.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 695.12,
+        "peak_cpu_rss_mb": 2895.7,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4646.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 11.65,
+        "peak_cpu_rss_mb": 2895.7,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4646.0,
+        "patches": 735,
+        "images": 15,
+        "patches_per_second": 63.1,
+        "seconds_per_image": 0.777
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2895.7,
+    "total_seconds": 927.2
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/snemi3d/macs-v11-norestart-acq-margin_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,28,28.4,158.1,"[0.2979926532667371, 0.40138560205746643, 0.2091063280267666, 0.345169021450426, 0.5132033840553815, 0.1750097104074813, 0.1991102608484407, 0.22650021516746904]",,229858,28,True,kept,0.3728225794885769,0.5132033840553815,"[0.0011717197029955054, 0.036143320102722025, 0.03602118748512524, 0.009780573323141589, 0.9132470305826798, 0.0004447112636304281, 0.0009272386944261949, 0.002264218845279129]",mixture,"{'mixture': 0.5124341043740674, 'uniform': 0.371832611058189, 'top1': 0.5132033840553815, 'top2': 0.5140566832535806, 'top3': 0.5140133725772849}",0.7016877048595026,0.5972861693259305,0.012805830910195319,,,,,
+2,56,5.5,19.8,"[0.3432294869846446, 0.38007716816165, 0.1893160272069609, 0.27325800973621645, 0.5104215693825217, 0.11859077059313591, 0.16161456913007907, 0.2653818534052973]",,537619,56,False,,,,"[0.001923155375394908, 0.03049413082456546, 0.030855984596669918, 0.004974945473752838, 0.9273348235211605, 0.0002641090935606338, 0.0006653947874493764, 0.003487456327446416]",mixture,"{'mixture': 0.509141392934992, 'uniform': 0.30389289820522647, 'top1': 0.5104215693825217, 'top2': 0.5115149870379297, 'top3': 0.5115711424860857}",0.7407435071089333,0.5701167948241403,0.014880619563059446,0.5472098965968067,0.506101874934941,student,0.157,
+3,84,5.3,112.9,"[0.37771424181735946, 0.4383924755772515, 0.20803697104167745, 0.3322271896816488, 0.5713981309510543, 0.1622958007666203, 0.21629404144420006, 0.30029101442837863]","[0.3939879167492122, 0.48133117792688046, 0.2378871834264335, 0.38115292524663485, 0.6639196547742888, 0.2009226453097915, 0.2786374526977216, 0.31603711564037745]",878752,84,False,,,,"[0.0014948062507949434, 0.030080258640846264, 0.0204855948778826, 0.004939617999442016, 0.9394194674567167, 0.0002251111641117732, 0.0006329282351913225, 0.002722215375014302]",mixture,"{'mixture': 0.570805470117565, 'uniform': 0.3686947077957518, 'top1': 0.5713981309510543, 'top2': 0.5724235003227502, 'top3': 0.5725081953043248}",0.7468245490384356,0.5417238671053858,0.010860615428633374,0.6138777505811737,0.42072085731863107,student,0.38,
+4,112,5.3,116.2,"[0.37113155177689233, 0.4304098004089352, 0.20188035699616788, 0.32804902477829634, 0.5613490894740727, 0.16195989381006395, 0.22064107723935433, 0.2945484326528721]","[0.37930059726350224, 0.46679191904083284, 0.22973305931652954, 0.37275095341645953, 0.646127639407633, 0.19164165882853965, 0.2799476672005563, 0.3042154481257559]",1174561,112,False,,,,"[0.00154438979007035, 0.030645980759803813, 0.021255485098501642, 0.005227669234419145, 0.9375132175239673, 0.0002475702667783434, 0.00072944689996532, 0.0028362404264940613]",mixture,"{'mixture': 0.5607148781393188, 'uniform': 0.3658178463120181, 'top1': 0.5613490894740727, 'top2': 0.5624084345460589, 'top3': 0.5624875840270268}",0.7542581674695896,0.5300439451809495,0.012610876109984098,0.6451945548649818,0.4694781440108841,student,0.484,
+5,140,5.4,119.8,"[0.3686013985558847, 0.4316764686766728, 0.2017286174831427, 0.32834735201508225, 0.5580438444411788, 0.16184451751433904, 0.22400369081247648, 0.292771023767665]","[0.3818583838279288, 0.4695261790989113, 0.23319827221739736, 0.3783978491677929, 0.6413947113666308, 0.18985071416968632, 0.2782163770065046, 0.30448063459351216]",1440921,140,False,,,,"[0.0015526486686203665, 0.03200215204034361, 0.02188349146114073, 0.005406401675151992, 0.9352495152292364, 0.00025497758356847564, 0.0007778608562707616, 0.0028729524856675776]",mixture,"{'mixture': 0.5573747664340176, 'uniform': 0.36380727550724445, 'top1': 0.5580438444411788, 'top2': 0.5590948972738314, 'top3': 0.5591800198932264}",0.7539582315006403,0.536392270389954,0.01499025266145266,0.6441366353069468,0.4658413022561706,student,0.48,
+6,168,5.2,123.1,"[0.3734322017511206, 0.4356543154702399, 0.20464087198641312, 0.3322555657020442, 0.563837343525697, 0.16618033768548526, 0.22835194680726428, 0.2938739092673822]","[0.39086690892472775, 0.4766700450989547, 0.23643556022313506, 0.3796429239120413, 0.6442590563252943, 0.1929323791472062, 0.28010353436049357, 0.3105360609364328]",1701154,168,False,,,,"[0.0015400210244406964, 0.03147228651218212, 0.021293052134217047, 0.005313185861618959, 0.9366166665504763, 0.00025165512057501784, 0.0007678204876118409, 0.0027453123088780123]",mixture,"{'mixture': 0.563181371815908, 'uniform': 0.3701849952532835, 'top1': 0.563837343525697, 'top2': 0.5648706041858641, 'top3': 0.5649549898869545}",0.7571927107648795,0.5271002562939713,0.019880075923030854,0.6663378365932968,0.526872228608967,student,0.554,
+7,168,0.0,35.1,,,,,,,,,,,,0.7543813767791763,0.5235559405723194,0.02210728108908942,,,,,True

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-3/log.txt

@@ -0,0 +1,105 @@
+[10:18:06][      0s][START] target=snemi3d variant=macs-v11-norestart-acq-mcdropout budget=5.0% seed=3 | {"task": "membrane", "target": "snemi3d", "variant": "macs-v11-norestart-acq-mcdropout", "budget": 5.0, "seed": 3, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "mcdropout", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[10:18:06][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[10:18:06][      0s][DATA] snemi3d: train=3430 val=735 test=735 patches
+[10:18:08][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, mouse-brain=4.08
+[10:18:08][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[10:18:14][      7s][WEIGHTS] mixture: c-elegans=0.005, cerebellum=0.058, flywire=0.394, berghia=0.027, microns=0.477, oct-glia=0.007, oct-sfl=0.011, mouse-brain=0.021 | eff.sources=3.24 | MMD2 mix=0.0601 uniform=0.1529
+[10:18:14][      7s][WEIGHTS] per-source: c-elegans=0.003, cerebellum=0.308, flywire=0.031, berghia=0.005, microns=0.624, oct-glia=0.021, oct-sfl=0.005, mouse-brain=0.002 | min-MMD source (NeuroADDA rule) = microns
+[10:18:32][     25s][TEACHERS] target-val Dice: c-elegans=0.412, cerebellum=0.473, flywire=0.196, berghia=0.310, microns=0.589, oct-glia=0.137, oct-sfl=0.305, mouse-brain=0.350
+[10:18:32][     25s][TEACHERS] ensemble target-val Dice: mixture=0.421, mixture@raw=0.463, mixture@logit=0.367, per_source=0.589, per_source@raw=0.591, per_source@logit=0.578, uniform=0.411, uniform@raw=0.467, uniform@logit=0.228
+[10:18:32][     25s][WEIGHTS] using mixture: c-elegans=0.005, cerebellum=0.058, flywire=0.394, berghia=0.027, microns=0.477, oct-glia=0.007, oct-sfl=0.011, mouse-brain=0.021
+[10:18:32][     25s][INIT] student initialised from teacher microns (weight 0.477)
+[10:18:33][     27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[10:18:33][     27s][KD] distilling the mixture ensemble (prob space) into the student
+[10:19:12][     66s][KD] soft labels for 3430 patches from 8 teachers in 39s
+[10:19:43][     96s][KD] epoch 1/5 | patches=3430 | loss=0.4446 | 30s
+[10:20:14][    127s][KD] epoch 2/5 | patches=3430 | loss=0.3724 | 31s
+[10:20:45][    158s][KD] epoch 3/5 | patches=3430 | loss=0.3566 | 31s
+[10:21:15][    188s][KD] epoch 4/5 | patches=3430 | loss=0.3518 | 30s
+[10:21:45][    219s][KD] epoch 5/5 | patches=3430 | loss=0.3476 | 31s
+[10:21:47][    220s][ROUND0] before any target label: val_dice=0.3504 val_vi=0.6311
+[10:21:47][    220s][AL] patches=3430 (pool 3430 after dropping blank; free blank negatives: True) | budget=171 (5.0%) | rounds=6 | per_round=28 | acquisition=mcdropout | first_round=stratified
+[10:21:53][    227s][AL] round-1 uncertainty quality on 3430 patches: {'spearman_uncertainty_error': 0.4795522004497552, 'auroc_top20_error': 0.8258676231842175}
+[10:22:16][    250s][AL] round 1: stratified cold start: 14 patches from the membrane-rich half (ensemble prevalence >= 0.156) + 14 from the membrane-poor half of 2000 candidates
+[10:22:18][    252s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.560, uniform=0.405, top1=0.561, top2=0.563, top3=0.563 -> using mixture
+[10:22:18][    252s][WEIGHTS] round 1: student kept (labeled Dice 0.392 vs best teacher microns 0.561); distilling the corrected ensemble into it
+[10:22:18][    252s][WEIGHTS] round 1: teacher Dice on 28 labeled patches + 0 blank (288982 fg px): c-elegans=0.361, cerebellum=0.451, flywire=0.218, berghia=0.284, microns=0.561, oct-glia=0.186, oct-sfl=0.372, mouse-brain=0.290 | weights: c-elegans=0.001, cerebellum=0.037, flywire=0.025, berghia=0.003, microns=0.927, oct-glia=0.000, oct-sfl=0.003, mouse-brain=0.003
+[10:22:57][    291s][KD] soft labels for 3402 patches from 8 teachers in 39s
+[10:23:28][    322s][KD] epoch 1/2 | patches=3402 | loss=0.2479 | 31s
+[10:23:57][    351s][KD] epoch 2/2 | patches=3402 | loss=0.2446 | 29s
+[10:24:20][    374s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[10:24:21][    375s][FT] round 1/6 epoch 1/5 | patches=28 | loss=1.0334
+[10:24:22][    376s][FT] round 1/6 epoch 2/5 | patches=28 | loss=0.9354
+[10:24:23][    377s][FT] round 1/6 epoch 3/5 | patches=28 | loss=0.8876
+[10:24:24][    377s][FT] round 1/6 epoch 4/5 | patches=28 | loss=0.8557
+[10:24:25][    378s][FT] round 1/6 epoch 5/5 | patches=28 | loss=0.8477
+[10:24:26][    380s][AL] round 1/6 | labeled=28 | selection_seconds=29.4000 round_seconds=157.9000 restart_student_dice=0.3917 restart_teacher_dice=0.5609 val_dice=0.7324 val_vi=0.5694 val_ece=0.0110
+[10:24:32][    386s][AL] round 2: 1/3402 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:24:35][    389s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.444, uniform=0.325, top1=0.448, top2=0.454, top3=0.454 -> using top3
+[10:24:35][    389s][WEIGHTS] round 2: teacher Dice on 56 labeled patches + 0 blank (636921 fg px): c-elegans=0.359, cerebellum=0.446, flywire=0.212, berghia=0.318, microns=0.448, oct-glia=0.171, oct-sfl=0.264, mouse-brain=0.305 | weights: c-elegans=0.004, cerebellum=0.106, flywire=0.000, berghia=0.000, microns=0.890, oct-glia=0.000, oct-sfl=0.000, mouse-brain=0.000
+[10:24:40][    394s][CONS] round 2: on the 28 new patches student Dice=0.608 vs ensemble 0.318 -> pseudo-labels from the student | consistency x0.36
+[10:24:41][    395s][FT] round 2/6 epoch 1/5 | patches=56 | loss=0.8341
+[10:24:43][    396s][FT] round 2/6 epoch 2/5 | patches=56 | loss=0.7748
+[10:24:44][    398s][FT] round 2/6 epoch 3/5 | patches=56 | loss=0.7368
+[10:24:45][    399s][FT] round 2/6 epoch 4/5 | patches=56 | loss=0.7240
+[10:24:47][    400s][FT] round 2/6 epoch 5/5 | patches=56 | loss=0.7130
+[10:24:48][    402s][AL] round 2/6 | labeled=56 | selection_seconds=6.5000 round_seconds=20.8000 pseudo_student_dice=0.6085 pseudo_ensemble_dice=0.3182 consistency_scale=0.3620 val_dice=0.7421 val_vi=0.5560 val_ece=0.0160
+[10:24:54][    408s][AL] round 3: 1/3374 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:25:43][    456s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.397, cerebellum=0.488, flywire=0.243, berghia=0.387, microns=0.680, oct-glia=0.197, oct-sfl=0.278, mouse-brain=0.314 | blended with labeled Dice at 0.50
+[10:26:28][    502s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.561, uniform=0.376, top1=0.562, top2=0.563, top3=0.563 -> using mixture
+[10:26:28][    502s][WEIGHTS] round 3: teacher Dice on 84 labeled patches + 0 blank (943818 fg px): c-elegans=0.377, cerebellum=0.452, flywire=0.221, berghia=0.344, microns=0.562, oct-glia=0.178, oct-sfl=0.251, mouse-brain=0.306 | weights: c-elegans=0.002, cerebellum=0.037, flywire=0.025, berghia=0.006, microns=0.925, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[10:26:33][    507s][CONS] round 3: on the 28 new patches student Dice=0.645 vs ensemble 0.433 -> pseudo-labels from the student | consistency x0.48
+[10:26:35][    509s][FT] round 3/6 epoch 1/5 | patches=84 | loss=0.7785
+[10:26:37][    510s][FT] round 3/6 epoch 2/5 | patches=84 | loss=0.7679
+[10:26:39][    512s][FT] round 3/6 epoch 3/5 | patches=84 | loss=0.7383
+[10:26:40][    514s][FT] round 3/6 epoch 4/5 | patches=84 | loss=0.7332
+[10:26:42][    516s][FT] round 3/6 epoch 5/5 | patches=84 | loss=0.7084
+[10:26:44][    517s][AL] round 3/6 | labeled=84 | selection_seconds=6.4000 round_seconds=114.0000 pseudo_student_dice=0.6451 pseudo_ensemble_dice=0.4329 consistency_scale=0.4840 val_dice=0.7503 val_vi=0.5510 val_ece=0.0170
+[10:27:39][    573s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.432, cerebellum=0.506, flywire=0.236, berghia=0.386, microns=0.658, oct-glia=0.193, oct-sfl=0.267, mouse-brain=0.343 | blended with labeled Dice at 0.50
+[10:28:25][    618s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.558, uniform=0.362, top1=0.559, top2=0.560, top3=0.560 -> using mixture
+[10:28:25][    618s][WEIGHTS] round 4: teacher Dice on 112 labeled patches + 0 blank (1182224 fg px): c-elegans=0.390, cerebellum=0.449, flywire=0.215, berghia=0.337, microns=0.559, oct-glia=0.170, oct-sfl=0.239, mouse-brain=0.314 | weights: c-elegans=0.002, cerebellum=0.037, flywire=0.024, berghia=0.006, microns=0.926, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[10:28:30][    623s][CONS] round 4: on the 28 new patches student Dice=0.644 vs ensemble 0.513 -> pseudo-labels from the student | consistency x0.48
+[10:28:32][    626s][FT] round 4/6 epoch 1/5 | patches=112 | loss=0.7208
+[10:28:34][    628s][FT] round 4/6 epoch 2/5 | patches=112 | loss=0.7066
+[10:28:36][    630s][FT] round 4/6 epoch 3/5 | patches=112 | loss=0.6940
+[10:28:39][    632s][FT] round 4/6 epoch 4/5 | patches=112 | loss=0.6903
+[10:28:41][    634s][FT] round 4/6 epoch 5/5 | patches=112 | loss=0.6861
+[10:28:42][    636s][AL] round 4/6 | labeled=112 | selection_seconds=6.3000 round_seconds=117.2000 pseudo_student_dice=0.6443 pseudo_ensemble_dice=0.5135 consistency_scale=0.4810 val_dice=0.7523 val_vi=0.5246 val_ece=0.0165
+[10:29:38][    692s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.373, cerebellum=0.461, flywire=0.241, berghia=0.372, microns=0.648, oct-glia=0.189, oct-sfl=0.276, mouse-brain=0.293 | blended with labeled Dice at 0.50
+[10:30:25][    738s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.557, uniform=0.367, top1=0.558, top2=0.559, top3=0.559 -> using mixture
+[10:30:25][    738s][WEIGHTS] round 5: teacher Dice on 140 labeled patches + 0 blank (1488314 fg px): c-elegans=0.360, cerebellum=0.427, flywire=0.212, berghia=0.333, microns=0.558, oct-glia=0.165, oct-sfl=0.250, mouse-brain=0.290 | weights: c-elegans=0.001, cerebellum=0.030, flywire=0.024, berghia=0.006, microns=0.934, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[10:30:29][    743s][CONS] round 5: on the 28 new patches student Dice=0.673 vs ensemble 0.496 -> pseudo-labels from the student | consistency x0.58
+[10:30:32][    746s][FT] round 5/6 epoch 1/5 | patches=140 | loss=0.7061
+[10:30:35][    749s][FT] round 5/6 epoch 2/5 | patches=140 | loss=0.6878
+[10:30:38][    751s][FT] round 5/6 epoch 3/5 | patches=140 | loss=0.6845
+[10:30:41][    754s][FT] round 5/6 epoch 4/5 | patches=140 | loss=0.6812
+[10:30:43][    757s][FT] round 5/6 epoch 5/5 | patches=140 | loss=0.6732
+[10:30:45][    758s][AL] round 5/6 | labeled=140 | selection_seconds=6.4000 round_seconds=121.0000 pseudo_student_dice=0.6735 pseudo_ensemble_dice=0.4959 consistency_scale=0.5780 val_dice=0.7568 val_vi=0.5302 val_ece=0.0188
+[10:31:41][    815s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.395, cerebellum=0.484, flywire=0.241, berghia=0.384, microns=0.645, oct-glia=0.199, oct-sfl=0.281, mouse-brain=0.315 | blended with labeled Dice at 0.50
+[10:32:28][    862s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.560, uniform=0.374, top1=0.561, top2=0.562, top3=0.562 -> using mixture
+[10:32:28][    862s][WEIGHTS] round 6: teacher Dice on 168 labeled patches + 0 blank (1780338 fg px): c-elegans=0.374, cerebellum=0.437, flywire=0.213, berghia=0.339, microns=0.561, oct-glia=0.170, oct-sfl=0.254, mouse-brain=0.301 | weights: c-elegans=0.002, cerebellum=0.033, flywire=0.024, berghia=0.006, microns=0.932, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[10:32:33][    867s][CONS] round 6: on the 28 new patches student Dice=0.664 vs ensemble 0.513 -> pseudo-labels from the student | consistency x0.55
+[10:32:36][    870s][FT] round 6/6 epoch 1/5 | patches=168 | loss=0.6981
+[10:32:39][    873s][FT] round 6/6 epoch 2/5 | patches=168 | loss=0.6750
+[10:32:43][    876s][FT] round 6/6 epoch 3/5 | patches=168 | loss=0.6793
+[10:32:46][    879s][FT] round 6/6 epoch 4/5 | patches=168 | loss=0.6753
+[10:32:49][    883s][FT] round 6/6 epoch 5/5 | patches=168 | loss=0.6607
+[10:32:50][    884s][AL] round 6/6 | labeled=168 | selection_seconds=6.3000 round_seconds=124.2000 pseudo_student_dice=0.6640 pseudo_ensemble_dice=0.5131 consistency_scale=0.5470 val_dice=0.7577 val_vi=0.5182 val_ece=0.0166
+[10:32:54][    887s][CONS] final consolidation: pseudo-labels refreshed from the student
+[10:32:57][    890s][FT] final 1/10 epoch 1/1 | patches=168 | loss=0.6569
+[10:33:00][    894s][FT] final 2/10 epoch 1/1 | patches=168 | loss=0.6460
+[10:33:03][    897s][FT] final 3/10 epoch 1/1 | patches=168 | loss=0.6440
+[10:33:06][    900s][FT] final 4/10 epoch 1/1 | patches=168 | loss=0.6361
+[10:33:10][    903s][FT] final 5/10 epoch 1/1 | patches=168 | loss=0.6337
+[10:33:13][    906s][FT] final 6/10 epoch 1/1 | patches=168 | loss=0.6272
+[10:33:16][    910s][FT] final 7/10 epoch 1/1 | patches=168 | loss=0.6199
+[10:33:19][    913s][FT] final 8/10 epoch 1/1 | patches=168 | loss=0.6179
+[10:33:22][    916s][FT] final 9/10 epoch 1/1 | patches=168 | loss=0.6094
+[10:33:26][    919s][FT] final 10/10 epoch 1/1 | patches=168 | loss=0.6205
+[10:33:27][    921s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=35.2000 val_dice=0.7596 val_vi=0.5157 val_ece=0.0215
+[10:33:27][    921s][WEIGHTS] final: c-elegans=0.002, cerebellum=0.033, flywire=0.024, berghia=0.006, microns=0.932, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003
+[10:33:37][    931s][TEST] raw student (no EMA, no TTA): patch_dice=0.7530 patch_vi=0.5232
+[10:33:39][    933s][TEST] final model without TTA: patch_dice=0.7530 patch_vi=0.5232
+[10:33:39][    933s][RESOURCES] {"phases": {"setup": {"seconds": 1.16, "peak_cpu_rss_mb": 1027.9, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 25.23, "peak_cpu_rss_mb": 1809.6, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 192.42, "peak_cpu_rss_mb": 2896.9, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 700.54, "peak_cpu_rss_mb": 2896.9, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 11.74, "peak_cpu_rss_mb": 2896.9, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 735, "images": 15, "patches_per_second": 62.6, "seconds_per_image": 0.783}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2896.9, "total_seconds": 931.1}
+[10:33:39][    933s][TEST] patch_dice=0.7613 patch_iou=0.6242 patch_precision=0.7383 patch_recall=0.7886 patch_vi=0.5082 image_dice=0.8089 image_iou=0.6796 image_precision=0.7885 image_recall=0.8309 image_vi=0.5161 cal_ece=0.0111 cal_nll=0.1784 cal_brier=0.0565 cal_mean_confidence=0.9294 cal_accuracy=0.9183 cal_predicted_prevalence=0.2180 cal_true_prevalence=0.2059 cal_mean_probability=0.2176
+[10:33:39][    933s][DONE] snemi3d macs-v11-norestart-acq-mcdropout b5 s3 | test patch Dice=0.7613 VI=0.5082 | image Dice=0.8089 VI=0.5161 | 933s

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-3/result.json

@@ -0,0 +1,785 @@
+{
+  "target": "snemi3d",
+  "variant": "macs-v11-norestart-acq-mcdropout",
+  "budget": 5.0,
+  "seed": 3,
+  "config": {
+    "task": "membrane",
+    "target": "snemi3d",
+    "variant": "macs-v11-norestart-acq-mcdropout",
+    "budget": 5.0,
+    "seed": 3,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "mcdropout",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "snemi3d",
+      "--variants",
+      "macs-v11-norestart-acq-mcdropout",
+      "--budgets",
+      "5",
+      "--seeds",
+      "3",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.00526521078604608,
+    0.05775508166244069,
+    0.39371182687582884,
+    0.027420285915093805,
+    0.47701492758994885,
+    0.006835626005792793,
+    0.011200163178997314,
+    0.020796877985851614
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.00526521078604608,
+      0.05775508166244069,
+      0.39371182687582884,
+      0.027420285915093805,
+      0.47701492758994885,
+      0.006835626005792793,
+      0.011200163178997314,
+      0.020796877985851614
+    ],
+    "per_source": [
+      0.0030099345500300633,
+      0.30826329572645517,
+      0.030874948568488214,
+      0.005310476630139808,
+      0.6243402297379833,
+      0.021312995363178842,
+      0.004803663858606227,
+      0.0020844555651184054
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.04836304661062167,
+      "mmd2_mixture": 0.06012254011242163,
+      "mmd2_uniform": 0.15285652884935502,
+      "mmd2_single_source": [
+        0.4522663178230211,
+        0.14701205466558376,
+        0.12241636968711567,
+        0.27265250499014093,
+        0.08995147555294303,
+        0.3744252998076626,
+        0.3395783286723916,
+        0.23687871798038196
+      ],
+      "median_squared_distance": 0.18663079059841653,
+      "gammas": [
+        1.3395431654037109,
+        2.6790863308074218,
+        5.3581726616148435,
+        10.716345323229687,
+        21.432690646459374
+      ],
+      "max_weight": 0.47701492758994885,
+      "entropy": 1.1759493501799958,
+      "effective_sources": 3.241218535100111,
+      "min_weight": 0.00526521078604608
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.46244015980269426,
+        0.12211161125756897,
+        0.2912827113901817,
+        0.4206979810988831,
+        0.07022524085432302,
+        0.3185314378412357,
+        0.4280722723409719,
+        0.48945233449571823
+      ],
+      "standardized": [
+        0.9323157340081956,
+        -1.3822022358410426,
+        -0.2316976633920459,
+        0.6484341051420859,
+        -1.7350728286900672,
+        -0.0463835976280083,
+        0.6985854358714068,
+        1.1160210505294779
+      ],
+      "tau": 0.5,
+      "max_weight": 0.6243402297379833,
+      "entropy": 0.9300685916208571,
+      "effective_sources": 2.534683029672754,
+      "min_weight": 0.0020844555651184054
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "microns_basil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4124719184272143,
+      0.4732143695257148,
+      0.1963993253870582,
+      0.31013775029960944,
+      0.5894614335833763,
+      0.13744118666208724,
+      0.30544280939432356,
+      0.35012843982905756
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.4205090273501111,
+      "mixture@raw": 0.46291647296981747,
+      "mixture@logit": 0.3665771544587855,
+      "per_source": 0.5892233046139178,
+      "per_source@raw": 0.5905333874987907,
+      "per_source@logit": 0.5781025615679164,
+      "uniform": 0.41121937389276464,
+      "uniform@raw": 0.46695544697394986,
+      "uniform@logit": 0.22779528784246278
+    },
+    "val_patches": 735
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    4.075904369354248
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.47701492758994885,
+    "entropy": 1.1759493501799958,
+    "effective_sources": 3.241218535100111,
+    "min_weight": 0.00526521078604608
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.3571428571428572,
+    "spearman_weight_taxonomy": 0.50709255283711,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 192.4,
+  "round0_val": {
+    "val_dice": 0.3503957018118952,
+    "val_vi": 0.631135568604762,
+    "val_ece": 0.0757153987157513
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 28,
+      "selection_seconds": 29.4,
+      "round_seconds": 157.9,
+      "teacher_dice_labeled": [
+        0.36056026860693446,
+        0.4508413766529139,
+        0.21750134794876225,
+        0.28356610561779927,
+        0.560875393102026,
+        0.18601963732494034,
+        0.3724686404052985,
+        0.28968887034603186
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 288982,
+      "evidence_nonempty_patches": 28,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.391711079681163,
+      "restart_teacher_dice": 0.560875393102026,
+      "weights": [
+        0.0013803357494966194,
+        0.03734608199477827,
+        0.024685983611284195,
+        0.0033285842306526635,
+        0.9269546251229578,
+        0.000312843356437377,
+        0.003307582804027515,
+        0.00268396313036561
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5602980725744923,
+        "uniform": 0.40469001951724776,
+        "top1": 0.560875393102026,
+        "top2": 0.5626323977653401,
+        "top3": 0.5627847905743437
+      },
+      "val_dice": 0.7323560437845593,
+      "val_vi": 0.5694021627238198,
+      "val_ece": 0.011040160826576361
+    },
+    {
+      "round": 2,
+      "n_labeled": 56,
+      "selection_seconds": 6.5,
+      "round_seconds": 20.8,
+      "teacher_dice_labeled": [
+        0.3585644400140594,
+        0.44569513378054804,
+        0.21164106279410505,
+        0.31840836229424285,
+        0.44781682892290103,
+        0.17125172597809413,
+        0.2640817814474379,
+        0.30486247967348346
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 636921,
+      "evidence_nonempty_patches": 56,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.004025912650355965,
+        0.10554619536395982,
+        0.0,
+        0.0,
+        0.8904278919856843,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.4444176386674499,
+        "uniform": 0.32484691123850307,
+        "top1": 0.44781682892290103,
+        "top2": 0.45371291795123353,
+        "top3": 0.45396144205152433
+      },
+      "pseudo_student_dice": 0.6084848613600118,
+      "pseudo_ensemble_dice": 0.31824112897528706,
+      "pseudo_source": "student",
+      "consistency_scale": 0.362,
+      "val_dice": 0.7421486412029648,
+      "val_vi": 0.556016678738807,
+      "val_ece": 0.015960624579132065
+    },
+    {
+      "round": 3,
+      "n_labeled": 84,
+      "selection_seconds": 6.4,
+      "round_seconds": 114.0,
+      "teacher_dice_labeled": [
+        0.3772306767830903,
+        0.45163488231981797,
+        0.2206921454618092,
+        0.34409742181789726,
+        0.5617200882449684,
+        0.17804109167411933,
+        0.2511611799133086,
+        0.3060860498014537
+      ],
+      "teacher_dice_pool": [
+        0.39700488397740485,
+        0.4883830214540865,
+        0.24298449902950017,
+        0.38680546140851324,
+        0.6801822253345543,
+        0.19701607358506548,
+        0.2778733913497543,
+        0.31410917782810194
+      ],
+      "evidence_fg_pixels": 943818,
+      "evidence_nonempty_patches": 84,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0016144130645445057,
+        0.03726683086288565,
+        0.02523127275765228,
+        0.006036357907624515,
+        0.9254611767492045,
+        0.00028596166549658167,
+        0.000973442678300187,
+        0.003130544314291847
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5608050244269053,
+        "uniform": 0.3759099719289151,
+        "top1": 0.5617200882449684,
+        "top2": 0.563139067283982,
+        "top3": 0.5632320431152088
+      },
+      "pseudo_student_dice": 0.6450631392491858,
+      "pseudo_ensemble_dice": 0.4328560747160571,
+      "pseudo_source": "student",
+      "consistency_scale": 0.484,
+      "val_dice": 0.7503492358738133,
+      "val_vi": 0.5510445083549702,
+      "val_ece": 0.017044766730611508
+    },
+    {
+      "round": 4,
+      "n_labeled": 112,
+      "selection_seconds": 6.3,
+      "round_seconds": 117.2,
+      "teacher_dice_labeled": [
+        0.39048865645881586,
+        0.4491344212222539,
+        0.21475358314891385,
+        0.3369891850657348,
+        0.5589963369618478,
+        0.1701515660711788,
+        0.2394433322730992,
+        0.31370029514801345
+      ],
+      "teacher_dice_pool": [
+        0.4315052627319388,
+        0.506103214858209,
+        0.236099460618685,
+        0.3860519184665265,
+        0.6576926480768255,
+        0.19304244666447792,
+        0.26686329163719663,
+        0.3427654718654806
+      ],
+      "evidence_fg_pixels": 1182224,
+      "evidence_nonempty_patches": 112,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0018950403875757182,
+        0.037366999859284565,
+        0.024444065719803742,
+        0.005780021245019906,
+        0.9258790041630323,
+        0.0002716871988332553,
+        0.0008901138320830429,
+        0.0034730675943676205
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.557933292560871,
+        "uniform": 0.3623408682867729,
+        "top1": 0.5589963369618478,
+        "top2": 0.5602149432711729,
+        "top3": 0.5603361268165685
+      },
+      "pseudo_student_dice": 0.6443473735400261,
+      "pseudo_ensemble_dice": 0.5134618975328987,
+      "pseudo_source": "student",
+      "consistency_scale": 0.481,
+      "val_dice": 0.7522883045571649,
+      "val_vi": 0.524641625228006,
+      "val_ece": 0.016537528590195295
+    },
+    {
+      "round": 5,
+      "n_labeled": 140,
+      "selection_seconds": 6.4,
+      "round_seconds": 121.0,
+      "teacher_dice_labeled": [
+        0.36001979030446934,
+        0.4266018157765912,
+        0.21166665891913206,
+        0.3326090449700532,
+        0.5582016752054481,
+        0.16480849083774873,
+        0.25026909107295403,
+        0.28984831642964815
+      ],
+      "teacher_dice_pool": [
+        0.37284183823210754,
+        0.4612999481718947,
+        0.24096947941714422,
+        0.37192028122390125,
+        0.6479905837255117,
+        0.1889133895654166,
+        0.27553571279748673,
+        0.2933843567028957
+      ],
+      "evidence_fg_pixels": 1488314,
+      "evidence_nonempty_patches": 140,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0014214857178205984,
+        0.03034443133386405,
+        0.024111028469058503,
+        0.005628016509243339,
+        0.9344405975254152,
+        0.0002620071420808458,
+        0.0010090434270227865,
+        0.0027833898754947807
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5570908624941996,
+        "uniform": 0.3666543830200925,
+        "top1": 0.5582016752054481,
+        "top2": 0.5591297232748379,
+        "top3": 0.5591888606233547
+      },
+      "pseudo_student_dice": 0.6734744719434842,
+      "pseudo_ensemble_dice": 0.4958697336902917,
+      "pseudo_source": "student",
+      "consistency_scale": 0.578,
+      "val_dice": 0.7568385587791026,
+      "val_vi": 0.5301980889611158,
+      "val_ece": 0.018759114485960196
+    },
+    {
+      "round": 6,
+      "n_labeled": 168,
+      "selection_seconds": 6.3,
+      "round_seconds": 124.2,
+      "teacher_dice_labeled": [
+        0.37408712100817315,
+        0.4371792712153743,
+        0.21264084317704324,
+        0.33868259877237294,
+        0.5607169060423229,
+        0.16997856242071668,
+        0.2538120015052483,
+        0.3009675653092333
+      ],
+      "teacher_dice_pool": [
+        0.39522086924433447,
+        0.48409444015218395,
+        0.24088776217932312,
+        0.3843698480549815,
+        0.6450175213456133,
+        0.19881484436170063,
+        0.28143892579597607,
+        0.3146631055005121
+      ],
+      "evidence_fg_pixels": 1780338,
+      "evidence_nonempty_patches": 168,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0015910641009763598,
+        0.03279954685573042,
+        0.023675442220052585,
+        0.005815459119122665,
+        0.931808660430251,
+        0.00026829834944663096,
+        0.0010165952065912859,
+        0.003024933717829064
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.55973701472009,
+        "uniform": 0.37394979224484814,
+        "top1": 0.5607169060423229,
+        "top2": 0.5617850814822273,
+        "top3": 0.5618683450323583
+      },
+      "pseudo_student_dice": 0.6640267626039819,
+      "pseudo_ensemble_dice": 0.5130756838160614,
+      "pseudo_source": "student",
+      "consistency_scale": 0.547,
+      "val_dice": 0.7577150031400584,
+      "val_vi": 0.5181974799383757,
+      "val_ece": 0.01662966655423441
+    },
+    {
+      "round": 7,
+      "n_labeled": 168,
+      "selection_seconds": 0.0,
+      "round_seconds": 35.2,
+      "final_consolidation": true,
+      "val_dice": 0.759555873492989,
+      "val_vi": 0.515686209770237,
+      "val_ece": 0.02146679764354789
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.4795522004497552,
+    "auroc_top20_error": 0.8258676231842175
+  },
+  "al_seconds": 700.5,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.0015910641009763598,
+    0.03279954685573042,
+    0.023675442220052585,
+    0.005815459119122665,
+    0.931808660430251,
+    0.00026829834944663096,
+    0.0010165952065912859,
+    0.003024933717829064
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.931808660430251,
+    "entropy": 0.333468035026498,
+    "effective_sources": 1.395800429104696,
+    "min_weight": 0.00026829834944663096
+  },
+  "top_source_final": "microns_basil",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7530073262298932,
+    "patch_iou": 0.6136276658297826,
+    "patch_precision": 0.7270609913951437,
+    "patch_recall": 0.7839162274714548,
+    "patch_vi": 0.5231999082442164,
+    "n_patches": 735,
+    "cal_ece": 0.019476051786440497,
+    "cal_nll": 0.19016047656597235,
+    "cal_brier": 0.05974153790932348,
+    "cal_mean_confidence": 0.9343915492354201,
+    "cal_accuracy": 0.9149154974489796,
+    "cal_predicted_prevalence": 0.22031914328231292,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21871572358961278
+  },
+  "test_notta": {
+    "patch_dice": 0.7530073262298932,
+    "patch_iou": 0.6136276658297826,
+    "patch_precision": 0.7270609913951437,
+    "patch_recall": 0.7839162274714548,
+    "patch_vi": 0.5231999082442164,
+    "n_patches": 735,
+    "cal_ece": 0.019476051786440497,
+    "cal_nll": 0.19016047656597235,
+    "cal_brier": 0.05974153790932348,
+    "cal_mean_confidence": 0.9343915492354201,
+    "cal_accuracy": 0.9149154974489796,
+    "cal_predicted_prevalence": 0.22031914328231292,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21871572358961278
+  },
+  "test": {
+    "patch_dice": 0.7612862326469151,
+    "patch_iou": 0.6241756457763458,
+    "patch_precision": 0.7382643458817646,
+    "patch_recall": 0.7886357133088496,
+    "patch_vi": 0.5082244541709398,
+    "n_patches": 735,
+    "image_dice": 0.8089188850134597,
+    "image_iou": 0.6795995339893497,
+    "image_precision": 0.7884908624044854,
+    "image_recall": 0.8308934273184174,
+    "image_vi": 0.5161480607947462,
+    "n_images": 15,
+    "cal_ece": 0.011077641126509934,
+    "cal_nll": 0.17836683789955543,
+    "cal_brier": 0.056488108008727,
+    "cal_mean_confidence": 0.9293559680822921,
+    "cal_accuracy": 0.9182783269557823,
+    "cal_predicted_prevalence": 0.2179793792517007,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21756974344637195
+  },
+  "eval_seconds": 11.7,
+  "total_seconds": 932.7,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.16,
+        "peak_cpu_rss_mb": 1027.9,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 25.23,
+        "peak_cpu_rss_mb": 1809.6,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 192.42,
+        "peak_cpu_rss_mb": 2896.9,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4638.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 700.54,
+        "peak_cpu_rss_mb": 2896.9,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4646.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 11.74,
+        "peak_cpu_rss_mb": 2896.9,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4646.0,
+        "patches": 735,
+        "images": 15,
+        "patches_per_second": 62.6,
+        "seconds_per_image": 0.783
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2896.9,
+    "total_seconds": 931.1
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/snemi3d/macs-v11-norestart-acq-mcdropout_b5_s3.pth"
+}

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-3/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,28,29.4,157.9,"[0.36056026860693446, 0.4508413766529139, 0.21750134794876225, 0.28356610561779927, 0.560875393102026, 0.18601963732494034, 0.3724686404052985, 0.28968887034603186]",,288982,28,True,kept,0.391711079681163,0.560875393102026,"[0.0013803357494966194, 0.03734608199477827, 0.024685983611284195, 0.0033285842306526635, 0.9269546251229578, 0.000312843356437377, 0.003307582804027515, 0.00268396313036561]",mixture,"{'mixture': 0.5602980725744923, 'uniform': 0.40469001951724776, 'top1': 0.560875393102026, 'top2': 0.5626323977653401, 'top3': 0.5627847905743437}",0.7323560437845593,0.5694021627238198,0.011040160826576361,,,,,
+2,56,6.5,20.8,"[0.3585644400140594, 0.44569513378054804, 0.21164106279410505, 0.31840836229424285, 0.44781682892290103, 0.17125172597809413, 0.2640817814474379, 0.30486247967348346]",,636921,56,False,,,,"[0.004025912650355965, 0.10554619536395982, 0.0, 0.0, 0.8904278919856843, 0.0, 0.0, 0.0]",top3,"{'mixture': 0.4444176386674499, 'uniform': 0.32484691123850307, 'top1': 0.44781682892290103, 'top2': 0.45371291795123353, 'top3': 0.45396144205152433}",0.7421486412029648,0.556016678738807,0.015960624579132065,0.6084848613600118,0.31824112897528706,student,0.362,
+3,84,6.4,114.0,"[0.3772306767830903, 0.45163488231981797, 0.2206921454618092, 0.34409742181789726, 0.5617200882449684, 0.17804109167411933, 0.2511611799133086, 0.3060860498014537]","[0.39700488397740485, 0.4883830214540865, 0.24298449902950017, 0.38680546140851324, 0.6801822253345543, 0.19701607358506548, 0.2778733913497543, 0.31410917782810194]",943818,84,False,,,,"[0.0016144130645445057, 0.03726683086288565, 0.02523127275765228, 0.006036357907624515, 0.9254611767492045, 0.00028596166549658167, 0.000973442678300187, 0.003130544314291847]",mixture,"{'mixture': 0.5608050244269053, 'uniform': 0.3759099719289151, 'top1': 0.5617200882449684, 'top2': 0.563139067283982, 'top3': 0.5632320431152088}",0.7503492358738133,0.5510445083549702,0.017044766730611508,0.6450631392491858,0.4328560747160571,student,0.484,
+4,112,6.3,117.2,"[0.39048865645881586, 0.4491344212222539, 0.21475358314891385, 0.3369891850657348, 0.5589963369618478, 0.1701515660711788, 0.2394433322730992, 0.31370029514801345]","[0.4315052627319388, 0.506103214858209, 0.236099460618685, 0.3860519184665265, 0.6576926480768255, 0.19304244666447792, 0.26686329163719663, 0.3427654718654806]",1182224,112,False,,,,"[0.0018950403875757182, 0.037366999859284565, 0.024444065719803742, 0.005780021245019906, 0.9258790041630323, 0.0002716871988332553, 0.0008901138320830429, 0.0034730675943676205]",mixture,"{'mixture': 0.557933292560871, 'uniform': 0.3623408682867729, 'top1': 0.5589963369618478, 'top2': 0.5602149432711729, 'top3': 0.5603361268165685}",0.7522883045571649,0.524641625228006,0.016537528590195295,0.6443473735400261,0.5134618975328987,student,0.481,
+5,140,6.4,121.0,"[0.36001979030446934, 0.4266018157765912, 0.21166665891913206, 0.3326090449700532, 0.5582016752054481, 0.16480849083774873, 0.25026909107295403, 0.28984831642964815]","[0.37284183823210754, 0.4612999481718947, 0.24096947941714422, 0.37192028122390125, 0.6479905837255117, 0.1889133895654166, 0.27553571279748673, 0.2933843567028957]",1488314,140,False,,,,"[0.0014214857178205984, 0.03034443133386405, 0.024111028469058503, 0.005628016509243339, 0.9344405975254152, 0.0002620071420808458, 0.0010090434270227865, 0.0027833898754947807]",mixture,"{'mixture': 0.5570908624941996, 'uniform': 0.3666543830200925, 'top1': 0.5582016752054481, 'top2': 0.5591297232748379, 'top3': 0.5591888606233547}",0.7568385587791026,0.5301980889611158,0.018759114485960196,0.6734744719434842,0.4958697336902917,student,0.578,
+6,168,6.3,124.2,"[0.37408712100817315, 0.4371792712153743, 0.21264084317704324, 0.33868259877237294, 0.5607169060423229, 0.16997856242071668, 0.2538120015052483, 0.3009675653092333]","[0.39522086924433447, 0.48409444015218395, 0.24088776217932312, 0.3843698480549815, 0.6450175213456133, 0.19881484436170063, 0.28143892579597607, 0.3146631055005121]",1780338,168,False,,,,"[0.0015910641009763598, 0.03279954685573042, 0.023675442220052585, 0.005815459119122665, 0.931808660430251, 0.00026829834944663096, 0.0010165952065912859, 0.003024933717829064]",mixture,"{'mixture': 0.55973701472009, 'uniform': 0.37394979224484814, 'top1': 0.5607169060423229, 'top2': 0.5617850814822273, 'top3': 0.5618683450323583}",0.7577150031400584,0.5181974799383757,0.01662966655423441,0.6640267626039819,0.5130756838160614,student,0.547,
+7,168,0.0,35.2,,,,,,,,,,,,0.759555873492989,0.515686209770237,0.02146679764354789,,,,,True

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-random/budget-5/seed-3/log.txt

@@ -0,0 +1,103 @@
+[10:33:41][      0s][START] target=snemi3d variant=macs-v11-norestart-acq-random budget=5.0% seed=3 | {"task": "membrane", "target": "snemi3d", "variant": "macs-v11-norestart-acq-random", "budget": 5.0, "seed": 3, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "random", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 0, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[10:33:41][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[10:33:42][      0s][DATA] snemi3d: train=3430 val=735 test=735 patches
+[10:33:43][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, mouse-brain=4.08
+[10:33:43][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[10:33:49][      7s][WEIGHTS] mixture: c-elegans=0.005, cerebellum=0.058, flywire=0.394, berghia=0.027, microns=0.477, oct-glia=0.007, oct-sfl=0.011, mouse-brain=0.021 | eff.sources=3.24 | MMD2 mix=0.0601 uniform=0.1529
+[10:33:49][      7s][WEIGHTS] per-source: c-elegans=0.003, cerebellum=0.308, flywire=0.031, berghia=0.005, microns=0.624, oct-glia=0.021, oct-sfl=0.005, mouse-brain=0.002 | min-MMD source (NeuroADDA rule) = microns
+[10:34:07][     25s][TEACHERS] target-val Dice: c-elegans=0.412, cerebellum=0.473, flywire=0.196, berghia=0.310, microns=0.589, oct-glia=0.137, oct-sfl=0.305, mouse-brain=0.350
+[10:34:07][     25s][TEACHERS] ensemble target-val Dice: mixture=0.421, mixture@raw=0.463, mixture@logit=0.367, per_source=0.589, per_source@raw=0.591, per_source@logit=0.578, uniform=0.411, uniform@raw=0.467, uniform@logit=0.228
+[10:34:07][     25s][WEIGHTS] using mixture: c-elegans=0.005, cerebellum=0.058, flywire=0.394, berghia=0.027, microns=0.477, oct-glia=0.007, oct-sfl=0.011, mouse-brain=0.021
+[10:34:07][     25s][INIT] student initialised from teacher microns (weight 0.477)
+[10:34:08][     26s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[10:34:08][     26s][KD] distilling the mixture ensemble (prob space) into the student
+[10:34:47][     65s][KD] soft labels for 3430 patches from 8 teachers in 39s
+[10:35:18][     97s][KD] epoch 1/5 | patches=3430 | loss=0.4446 | 31s
+[10:35:49][    127s][KD] epoch 2/5 | patches=3430 | loss=0.3724 | 31s
+[10:36:20][    158s][KD] epoch 3/5 | patches=3430 | loss=0.3568 | 31s
+[10:36:50][    189s][KD] epoch 4/5 | patches=3430 | loss=0.3541 | 31s
+[10:37:20][    219s][KD] epoch 5/5 | patches=3430 | loss=0.3478 | 30s
+[10:37:22][    220s][ROUND0] before any target label: val_dice=0.3580 val_vi=0.6345
+[10:37:22][    220s][AL] patches=3430 (pool 3430 after dropping blank; free blank negatives: True) | budget=171 (5.0%) | rounds=6 | per_round=28 | acquisition=random | first_round=stratified
+[10:37:27][    226s][AL] round-1 uncertainty quality on 3430 patches: {'spearman_uncertainty_error': 0.005164451042246618, 'auroc_top20_error': 0.5020564348188254}
+[10:37:50][    249s][AL] round 1: stratified cold start: 14 patches from the membrane-rich half (ensemble prevalence >= 0.156) + 14 from the membrane-poor half of 2000 candidates
+[10:37:52][    251s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.560, uniform=0.405, top1=0.561, top2=0.563, top3=0.563 -> using mixture
+[10:37:53][    251s][WEIGHTS] round 1: student kept (labeled Dice 0.402 vs best teacher microns 0.561); distilling the corrected ensemble into it
+[10:37:53][    251s][WEIGHTS] round 1: teacher Dice on 28 labeled patches + 0 blank (288982 fg px): c-elegans=0.361, cerebellum=0.451, flywire=0.218, berghia=0.284, microns=0.561, oct-glia=0.186, oct-sfl=0.372, mouse-brain=0.290 | weights: c-elegans=0.001, cerebellum=0.037, flywire=0.025, berghia=0.003, microns=0.927, oct-glia=0.000, oct-sfl=0.003, mouse-brain=0.003
+[10:38:31][    290s][KD] soft labels for 3402 patches from 8 teachers in 39s
+[10:39:01][    320s][KD] epoch 1/2 | patches=3402 | loss=0.2481 | 30s
+[10:39:32][    350s][KD] epoch 2/2 | patches=3402 | loss=0.2479 | 30s
+[10:39:54][    373s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 23s
+[10:39:55][    374s][FT] round 1/6 epoch 1/5 | patches=28 | loss=1.0342
+[10:39:56][    375s][FT] round 1/6 epoch 2/5 | patches=28 | loss=0.9446
+[10:39:57][    375s][FT] round 1/6 epoch 3/5 | patches=28 | loss=0.8956
+[10:39:58][    376s][FT] round 1/6 epoch 4/5 | patches=28 | loss=0.8636
+[10:39:59][    377s][FT] round 1/6 epoch 5/5 | patches=28 | loss=0.8517
+[10:40:00][    379s][AL] round 1/6 | labeled=28 | selection_seconds=28.4000 round_seconds=156.9000 restart_student_dice=0.4024 restart_teacher_dice=0.5609 val_dice=0.7275 val_vi=0.5750 val_ece=0.0118
+[10:40:03][    382s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.557, uniform=0.374, top1=0.559, top2=0.559, top3=0.559 -> using mixture
+[10:40:03][    382s][WEIGHTS] round 2: teacher Dice on 56 labeled patches + 0 blank (565126 fg px): c-elegans=0.348, cerebellum=0.419, flywire=0.206, berghia=0.301, microns=0.559, oct-glia=0.176, oct-sfl=0.325, mouse-brain=0.283 | weights: c-elegans=0.001, cerebellum=0.028, flywire=0.023, berghia=0.004, microns=0.939, oct-glia=0.000, oct-sfl=0.002, mouse-brain=0.003
+[10:40:08][    387s][CONS] round 2: on the 28 new patches student Dice=0.685 vs ensemble 0.554 -> pseudo-labels from the student | consistency x0.62
+[10:40:10][    388s][FT] round 2/6 epoch 1/5 | patches=56 | loss=0.7638
+[10:40:11][    389s][FT] round 2/6 epoch 2/5 | patches=56 | loss=0.7486
+[10:40:12][    391s][FT] round 2/6 epoch 3/5 | patches=56 | loss=0.7512
+[10:40:14][    392s][FT] round 2/6 epoch 4/5 | patches=56 | loss=0.7207
+[10:40:15][    393s][FT] round 2/6 epoch 5/5 | patches=56 | loss=0.7258
+[10:40:16][    395s][AL] round 2/6 | labeled=56 | selection_seconds=0.2000 round_seconds=14.6000 pseudo_student_dice=0.6854 pseudo_ensemble_dice=0.5543 consistency_scale=0.6180 val_dice=0.7399 val_vi=0.5367 val_ece=0.0128
+[10:41:05][    443s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.362, cerebellum=0.462, flywire=0.235, berghia=0.363, microns=0.664, oct-glia=0.194, oct-sfl=0.281, mouse-brain=0.290 | blended with labeled Dice at 0.50
+[10:41:50][    488s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.615, uniform=0.410, top1=0.616, top2=0.616, top3=0.616 -> using mixture
+[10:41:50][    488s][WEIGHTS] round 3: teacher Dice on 84 labeled patches + 0 blank (846975 fg px): c-elegans=0.356, cerebellum=0.441, flywire=0.220, berghia=0.336, microns=0.616, oct-glia=0.181, oct-sfl=0.291, mouse-brain=0.284 | weights: c-elegans=0.001, cerebellum=0.020, flywire=0.015, berghia=0.003, microns=0.958, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002
+[10:41:55][    493s][CONS] round 3: on the 28 new patches student Dice=0.717 vs ensemble 0.582 -> pseudo-labels from the student | consistency x0.72
+[10:41:57][    495s][FT] round 3/6 epoch 1/5 | patches=84 | loss=0.7123
+[10:41:58][    497s][FT] round 3/6 epoch 2/5 | patches=84 | loss=0.6938
+[10:42:00][    499s][FT] round 3/6 epoch 3/5 | patches=84 | loss=0.6929
+[10:42:02][    501s][FT] round 3/6 epoch 4/5 | patches=84 | loss=0.6823
+[10:42:04][    503s][FT] round 3/6 epoch 5/5 | patches=84 | loss=0.6725
+[10:42:06][    504s][AL] round 3/6 | labeled=84 | selection_seconds=0.2000 round_seconds=107.8000 pseudo_student_dice=0.7165 pseudo_ensemble_dice=0.5822 consistency_scale=0.7220 val_dice=0.7494 val_vi=0.5370 val_ece=0.0221
+[10:42:54][    553s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.388, cerebellum=0.473, flywire=0.236, berghia=0.386, microns=0.654, oct-glia=0.192, oct-sfl=0.269, mouse-brain=0.306 | blended with labeled Dice at 0.50
+[10:43:40][    599s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.613, uniform=0.409, top1=0.613, top2=0.614, top3=0.614 -> using mixture
+[10:43:40][    599s][WEIGHTS] round 4: teacher Dice on 112 labeled patches + 0 blank (1109045 fg px): c-elegans=0.368, cerebellum=0.443, flywire=0.222, berghia=0.351, microns=0.613, oct-glia=0.183, oct-sfl=0.282, mouse-brain=0.291 | weights: c-elegans=0.001, cerebellum=0.021, flywire=0.016, berghia=0.004, microns=0.955, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002
+[10:43:45][    604s][CONS] round 4: on the 28 new patches student Dice=0.723 vs ensemble 0.588 -> pseudo-labels from the student | consistency x0.74
+[10:43:48][    606s][FT] round 4/6 epoch 1/5 | patches=112 | loss=0.6614
+[10:43:50][    608s][FT] round 4/6 epoch 2/5 | patches=112 | loss=0.6542
+[10:43:52][    611s][FT] round 4/6 epoch 3/5 | patches=112 | loss=0.6548
+[10:43:54][    613s][FT] round 4/6 epoch 4/5 | patches=112 | loss=0.6452
+[10:43:57][    615s][FT] round 4/6 epoch 5/5 | patches=112 | loss=0.6399
+[10:43:58][    617s][AL] round 4/6 | labeled=112 | selection_seconds=0.2000 round_seconds=111.1000 pseudo_student_dice=0.7229 pseudo_ensemble_dice=0.5877 consistency_scale=0.7430 val_dice=0.7508 val_vi=0.5354 val_ece=0.0242
+[10:44:48][    666s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.384, cerebellum=0.470, flywire=0.242, berghia=0.380, microns=0.652, oct-glia=0.192, oct-sfl=0.281, mouse-brain=0.304 | blended with labeled Dice at 0.50
+[10:45:34][    713s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.610, uniform=0.408, top1=0.611, top2=0.611, top3=0.611 -> using mixture
+[10:45:34][    713s][WEIGHTS] round 5: teacher Dice on 140 labeled patches + 0 blank (1372470 fg px): c-elegans=0.365, cerebellum=0.440, flywire=0.225, berghia=0.348, microns=0.611, oct-glia=0.181, oct-sfl=0.277, mouse-brain=0.288 | weights: c-elegans=0.001, cerebellum=0.021, flywire=0.017, berghia=0.004, microns=0.955, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002
+[10:45:39][    718s][CONS] round 5: on the 28 new patches student Dice=0.708 vs ensemble 0.558 -> pseudo-labels from the student | consistency x0.69
+[10:45:42][    721s][FT] round 5/6 epoch 1/5 | patches=140 | loss=0.6426
+[10:45:45][    723s][FT] round 5/6 epoch 2/5 | patches=140 | loss=0.6381
+[10:45:48][    726s][FT] round 5/6 epoch 3/5 | patches=140 | loss=0.6368
+[10:45:50][    729s][FT] round 5/6 epoch 4/5 | patches=140 | loss=0.6257
+[10:45:53][    732s][FT] round 5/6 epoch 5/5 | patches=140 | loss=0.6263
+[10:45:55][    733s][AL] round 5/6 | labeled=140 | selection_seconds=0.2000 round_seconds=115.0000 pseudo_student_dice=0.7077 pseudo_ensemble_dice=0.5583 consistency_scale=0.6920 val_dice=0.7525 val_vi=0.5317 val_ece=0.0236
+[10:46:45][    784s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.377, cerebellum=0.466, flywire=0.243, berghia=0.380, microns=0.646, oct-glia=0.198, oct-sfl=0.283, mouse-brain=0.299 | blended with labeled Dice at 0.50
+[10:47:32][    831s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.607, uniform=0.407, top1=0.608, top2=0.608, top3=0.608 -> using mixture
+[10:47:32][    831s][WEIGHTS] round 6: teacher Dice on 168 labeled patches + 0 blank (1676005 fg px): c-elegans=0.363, cerebellum=0.440, flywire=0.224, berghia=0.346, microns=0.608, oct-glia=0.184, oct-sfl=0.278, mouse-brain=0.287 | weights: c-elegans=0.001, cerebellum=0.022, flywire=0.017, berghia=0.004, microns=0.954, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002
+[10:47:37][    835s][CONS] round 6: on the 28 new patches student Dice=0.757 vs ensemble 0.564 -> pseudo-labels from the student | consistency x0.86
+[10:47:40][    839s][FT] round 6/6 epoch 1/5 | patches=168 | loss=0.6402
+[10:47:43][    842s][FT] round 6/6 epoch 2/5 | patches=168 | loss=0.6324
+[10:47:47][    845s][FT] round 6/6 epoch 3/5 | patches=168 | loss=0.6399
+[10:47:50][    848s][FT] round 6/6 epoch 4/5 | patches=168 | loss=0.6291
+[10:47:53][    851s][FT] round 6/6 epoch 5/5 | patches=168 | loss=0.6404
+[10:47:54][    853s][AL] round 6/6 | labeled=168 | selection_seconds=0.2000 round_seconds=118.4000 pseudo_student_dice=0.7570 pseudo_ensemble_dice=0.5641 consistency_scale=0.8570 val_dice=0.7530 val_vi=0.5240 val_ece=0.0258
+[10:47:58][    856s][CONS] final consolidation: pseudo-labels refreshed from the student
+[10:48:01][    859s][FT] final 1/10 epoch 1/1 | patches=168 | loss=0.6136
+[10:48:04][    863s][FT] final 2/10 epoch 1/1 | patches=168 | loss=0.6131
+[10:48:07][    866s][FT] final 3/10 epoch 1/1 | patches=168 | loss=0.6124
+[10:48:10][    869s][FT] final 4/10 epoch 1/1 | patches=168 | loss=0.6099
+[10:48:14][    872s][FT] final 5/10 epoch 1/1 | patches=168 | loss=0.5997
+[10:48:17][    875s][FT] final 6/10 epoch 1/1 | patches=168 | loss=0.5997
+[10:48:20][    879s][FT] final 7/10 epoch 1/1 | patches=168 | loss=0.5953
+[10:48:23][    882s][FT] final 8/10 epoch 1/1 | patches=168 | loss=0.5949
+[10:48:26][    885s][FT] final 9/10 epoch 1/1 | patches=168 | loss=0.5923
+[10:48:30][    888s][FT] final 10/10 epoch 1/1 | patches=168 | loss=0.5925
+[10:48:31][    890s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=35.2000 val_dice=0.7561 val_vi=0.5234 val_ece=0.0284
+[10:48:31][    890s][WEIGHTS] final: c-elegans=0.001, cerebellum=0.022, flywire=0.017, berghia=0.004, microns=0.954, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002
+[10:48:41][    900s][TEST] raw student (no EMA, no TTA): patch_dice=0.7512 patch_vi=0.5222
+[10:48:43][    901s][TEST] final model without TTA: patch_dice=0.7512 patch_vi=0.5222
+[10:48:43][    901s][RESOURCES] {"phases": {"setup": {"seconds": 1.14, "peak_cpu_rss_mb": 1027.6, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 24.99, "peak_cpu_rss_mb": 1810.2, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 192.59, "peak_cpu_rss_mb": 2896.7, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 669.28, "peak_cpu_rss_mb": 2896.7, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 11.6, "peak_cpu_rss_mb": 2896.7, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 735, "images": 15, "patches_per_second": 63.4, "seconds_per_image": 0.773}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 2896.7, "total_seconds": 899.6}
+[10:48:43][    901s][TEST] patch_dice=0.7595 patch_iou=0.6221 patch_precision=0.7419 patch_recall=0.7813 patch_vi=0.5086 image_dice=0.8065 image_iou=0.6762 image_precision=0.7927 image_recall=0.8212 image_vi=0.5177 cal_ece=0.0171 cal_nll=0.1809 cal_brier=0.0571 cal_mean_confidence=0.9353 cal_accuracy=0.9183 cal_predicted_prevalence=0.2149 cal_true_prevalence=0.2059 cal_mean_probability=0.2135
+[10:48:43][    901s][DONE] snemi3d macs-v11-norestart-acq-random b5 s3 | test patch Dice=0.7595 VI=0.5086 | image Dice=0.8065 VI=0.5177 | 901s

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-random/budget-5/seed-3/result.json

@@ -0,0 +1,785 @@
+{
+  "target": "snemi3d",
+  "variant": "macs-v11-norestart-acq-random",
+  "budget": 5.0,
+  "seed": 3,
+  "config": {
+    "task": "membrane",
+    "target": "snemi3d",
+    "variant": "macs-v11-norestart-acq-random",
+    "budget": 5.0,
+    "seed": 3,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "random",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 0,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "snemi3d",
+      "--variants",
+      "macs-v11-norestart-acq-random",
+      "--budgets",
+      "5",
+      "--seeds",
+      "3",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "whole-mouse-brain"
+  ],
+  "alphas": [
+    0.00526521078604608,
+    0.05775508166244069,
+    0.39371182687582884,
+    0.027420285915093805,
+    0.47701492758994885,
+    0.006835626005792793,
+    0.011200163178997314,
+    0.020796877985851614
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "whole-mouse-brain"
+    ],
+    "mixture": [
+      0.00526521078604608,
+      0.05775508166244069,
+      0.39371182687582884,
+      0.027420285915093805,
+      0.47701492758994885,
+      0.006835626005792793,
+      0.011200163178997314,
+      0.020796877985851614
+    ],
+    "per_source": [
+      0.0030099345500300633,
+      0.30826329572645517,
+      0.030874948568488214,
+      0.005310476630139808,
+      0.6243402297379833,
+      0.021312995363178842,
+      0.004803663858606227,
+      0.0020844555651184054
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.04836304661062167,
+      "mmd2_mixture": 0.06012254011242163,
+      "mmd2_uniform": 0.15285652884935502,
+      "mmd2_single_source": [
+        0.4522663178230211,
+        0.14701205466558376,
+        0.12241636968711567,
+        0.27265250499014093,
+        0.08995147555294303,
+        0.3744252998076626,
+        0.3395783286723916,
+        0.23687871798038196
+      ],
+      "median_squared_distance": 0.18663079059841653,
+      "gammas": [
+        1.3395431654037109,
+        2.6790863308074218,
+        5.3581726616148435,
+        10.716345323229687,
+        21.432690646459374
+      ],
+      "max_weight": 0.47701492758994885,
+      "entropy": 1.1759493501799958,
+      "effective_sources": 3.241218535100111,
+      "min_weight": 0.00526521078604608
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.46244015980269426,
+        0.12211161125756897,
+        0.2912827113901817,
+        0.4206979810988831,
+        0.07022524085432302,
+        0.3185314378412357,
+        0.4280722723409719,
+        0.48945233449571823
+      ],
+      "standardized": [
+        0.9323157340081956,
+        -1.3822022358410426,
+        -0.2316976633920459,
+        0.6484341051420859,
+        -1.7350728286900672,
+        -0.0463835976280083,
+        0.6985854358714068,
+        1.1160210505294779
+      ],
+      "tau": 0.5,
+      "max_weight": 0.6243402297379833,
+      "entropy": 0.9300685916208571,
+      "effective_sources": 2.534683029672754,
+      "min_weight": 0.0020844555651184054
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "microns_basil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.4124719184272143,
+      0.4732143695257148,
+      0.1963993253870582,
+      0.31013775029960944,
+      0.5894614335833763,
+      0.13744118666208724,
+      0.30544280939432356,
+      0.35012843982905756
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.4205090273501111,
+      "mixture@raw": 0.46291647296981747,
+      "mixture@logit": 0.3665771544587855,
+      "per_source": 0.5892233046139178,
+      "per_source@raw": 0.5905333874987907,
+      "per_source@logit": 0.5781025615679164,
+      "uniform": 0.41121937389276464,
+      "uniform@raw": 0.46695544697394986,
+      "uniform@logit": 0.22779528784246278
+    },
+    "val_patches": 735
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    4.075904369354248
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.47701492758994885,
+    "entropy": 1.1759493501799958,
+    "effective_sources": 3.241218535100111,
+    "min_weight": 0.00526521078604608
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.3571428571428572,
+    "spearman_weight_taxonomy": 0.50709255283711,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 192.6,
+  "round0_val": {
+    "val_dice": 0.35801025103979234,
+    "val_vi": 0.6344848740811435,
+    "val_ece": 0.07478441083095859
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 28,
+      "selection_seconds": 28.4,
+      "round_seconds": 156.9,
+      "teacher_dice_labeled": [
+        0.36056026860693446,
+        0.4508413766529139,
+        0.21750134794876225,
+        0.28356610561779927,
+        0.560875393102026,
+        0.18601963732494034,
+        0.3724686404052985,
+        0.28968887034603186
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 288982,
+      "evidence_nonempty_patches": 28,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.4024189407361833,
+      "restart_teacher_dice": 0.560875393102026,
+      "weights": [
+        0.0013803357494966194,
+        0.03734608199477827,
+        0.024685983611284195,
+        0.0033285842306526635,
+        0.9269546251229578,
+        0.000312843356437377,
+        0.003307582804027515,
+        0.00268396313036561
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5602980725744923,
+        "uniform": 0.40469001951724776,
+        "top1": 0.560875393102026,
+        "top2": 0.5626323977653401,
+        "top3": 0.5627847905743437
+      },
+      "val_dice": 0.727546285788197,
+      "val_vi": 0.5749785964732628,
+      "val_ece": 0.011788481275023454
+    },
+    {
+      "round": 2,
+      "n_labeled": 56,
+      "selection_seconds": 0.2,
+      "round_seconds": 14.6,
+      "teacher_dice_labeled": [
+        0.3479823236552919,
+        0.4186981967226685,
+        0.20557587495698246,
+        0.3007128919242538,
+        0.5589578670606897,
+        0.17607239437484073,
+        0.3247645305942548,
+        0.28265004696019047
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 565126,
+      "evidence_nonempty_patches": 56,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0012569812895489763,
+        0.02796522949451945,
+        0.022627061811221816,
+        0.004080341346961101,
+        0.9390747514634746,
+        0.00029247988784289957,
+        0.0021198435345643842,
+        0.0025833111718668077
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5571876015358332,
+        "uniform": 0.374202964280938,
+        "top1": 0.5589578670606897,
+        "top2": 0.5592919600765673,
+        "top3": 0.5593125851518601
+      },
+      "pseudo_student_dice": 0.6853953687609664,
+      "pseudo_ensemble_dice": 0.5542848049226852,
+      "pseudo_source": "student",
+      "consistency_scale": 0.618,
+      "val_dice": 0.7399397755468825,
+      "val_vi": 0.5367473626096051,
+      "val_ece": 0.012806996872993453
+    },
+    {
+      "round": 3,
+      "n_labeled": 84,
+      "selection_seconds": 0.2,
+      "round_seconds": 107.8,
+      "teacher_dice_labeled": [
+        0.3557500469410345,
+        0.4409271340874671,
+        0.21965401520887193,
+        0.3356542715086659,
+        0.6156078106682936,
+        0.1806652407278319,
+        0.29110385747882406,
+        0.2844647798935356
+      ],
+      "teacher_dice_pool": [
+        0.36221583553452585,
+        0.46154148860749067,
+        0.2351214263672385,
+        0.36298119639097676,
+        0.6642216725599263,
+        0.19426701924887152,
+        0.2809982631984566,
+        0.2902214086069471
+      ],
+      "evidence_fg_pixels": 846975,
+      "evidence_nonempty_patches": 84,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0007864934568305689,
+        0.020220362781744897,
+        0.015079962329175009,
+        0.0033502443752500594,
+        0.9579861938327272,
+        0.0001772858169266955,
+        0.0008764938397664921,
+        0.0015229635675790894
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6149834430342458,
+        "uniform": 0.41044819487495626,
+        "top1": 0.6156078106682936,
+        "top2": 0.6161620454596852,
+        "top3": 0.6161824348043965
+      },
+      "pseudo_student_dice": 0.7165017188224702,
+      "pseudo_ensemble_dice": 0.5822070619568362,
+      "pseudo_source": "student",
+      "consistency_scale": 0.722,
+      "val_dice": 0.7493534762151666,
+      "val_vi": 0.537037416881874,
+      "val_ece": 0.022052875547815737
+    },
+    {
+      "round": 4,
+      "n_labeled": 112,
+      "selection_seconds": 0.2,
+      "round_seconds": 111.1,
+      "teacher_dice_labeled": [
+        0.36799533483144575,
+        0.4434738198437197,
+        0.22237860981915997,
+        0.35119162408855786,
+        0.6132689554994883,
+        0.18307297983849213,
+        0.281595677466981,
+        0.2906295936185792
+      ],
+      "teacher_dice_pool": [
+        0.38838436897907,
+        0.47321562487468605,
+        0.23612988589273762,
+        0.3855286998335696,
+        0.6543285906911006,
+        0.19182863371607134,
+        0.26864523780903776,
+        0.30642074866250274
+      ],
+      "evidence_fg_pixels": 1109045,
+      "evidence_nonempty_patches": 112,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0009075705663560531,
+        0.021176478328176308,
+        0.01582113731605874,
+        0.003995388643561608,
+        0.9554465806535456,
+        0.00018541093767134145,
+        0.0008136920001201178,
+        0.0016537415545100844
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6125056280699729,
+        "uniform": 0.4087578286493045,
+        "top1": 0.6132689554994883,
+        "top2": 0.6137450526434824,
+        "top3": 0.6137686488987392
+      },
+      "pseudo_student_dice": 0.7228817582417583,
+      "pseudo_ensemble_dice": 0.5876644214959471,
+      "pseudo_source": "student",
+      "consistency_scale": 0.743,
+      "val_dice": 0.7508233511614852,
+      "val_vi": 0.5353514012696726,
+      "val_ece": 0.024242308398439487
+    },
+    {
+      "round": 5,
+      "n_labeled": 140,
+      "selection_seconds": 0.2,
+      "round_seconds": 115.0,
+      "teacher_dice_labeled": [
+        0.36545827887620963,
+        0.44001314710549466,
+        0.2250317553851608,
+        0.3482704748138672,
+        0.6108450152408786,
+        0.18098147746558904,
+        0.2772557088193389,
+        0.28833909699736204
+      ],
+      "teacher_dice_pool": [
+        0.38441743915813703,
+        0.4699472512144608,
+        0.242043795701895,
+        0.38036188631835627,
+        0.6520519630893523,
+        0.19162936980644238,
+        0.28072756803528426,
+        0.30415916811101773
+      ],
+      "evidence_fg_pixels": 1372470,
+      "evidence_nonempty_patches": 140,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0009060261815732048,
+        0.020946085232241562,
+        0.016635612810557043,
+        0.0039732992909113215,
+        0.9549002581410955,
+        0.00018592196889683152,
+        0.0007977934031757694,
+        0.0016550029715487065
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.61001214706305,
+        "uniform": 0.40779566499551556,
+        "top1": 0.6108450152408786,
+        "top2": 0.6112827862046616,
+        "top3": 0.6113057282316456
+      },
+      "pseudo_student_dice": 0.7077233122972116,
+      "pseudo_ensemble_dice": 0.5583120517817071,
+      "pseudo_source": "student",
+      "consistency_scale": 0.692,
+      "val_dice": 0.7525420432275504,
+      "val_vi": 0.5316956049707146,
+      "val_ece": 0.02357947966251423
+    },
+    {
+      "round": 6,
+      "n_labeled": 168,
+      "selection_seconds": 0.2,
+      "round_seconds": 118.4,
+      "teacher_dice_labeled": [
+        0.3634730961229432,
+        0.4400121039280854,
+        0.22399140525566236,
+        0.3461002739895306,
+        0.6077325689334545,
+        0.18446004079543676,
+        0.27816643604219415,
+        0.2869076314362251
+      ],
+      "teacher_dice_pool": [
+        0.37718160768208964,
+        0.4658483575908974,
+        0.24283830333599785,
+        0.37970663291846574,
+        0.6464730400378235,
+        0.19800368751348005,
+        0.2827052621021349,
+        0.29906355224330206
+      ],
+      "evidence_fg_pixels": 1676005,
+      "evidence_nonempty_patches": 168,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.0009152615352997819,
+        0.02158362259693189,
+        0.01696471714785591,
+        0.0040063807443571735,
+        0.9538208919175786,
+        0.00019836448588131552,
+        0.0008296056124843809,
+        0.0016811559596108942
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.6068742421539629,
+        "uniform": 0.4065183273644009,
+        "top1": 0.6077325689334545,
+        "top2": 0.6082575257609559,
+        "top3": 0.6082816859281865
+      },
+      "pseudo_student_dice": 0.7569865654765929,
+      "pseudo_ensemble_dice": 0.5640826712486684,
+      "pseudo_source": "student",
+      "consistency_scale": 0.857,
+      "val_dice": 0.7529501278915827,
+      "val_vi": 0.5239912101194711,
+      "val_ece": 0.02575743938369193
+    },
+    {
+      "round": 7,
+      "n_labeled": 168,
+      "selection_seconds": 0.0,
+      "round_seconds": 35.2,
+      "final_consolidation": true,
+      "val_dice": 0.7560981292366642,
+      "val_vi": 0.5234298513862702,
+      "val_ece": 0.028411418873243124
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.005164451042246618,
+    "auroc_top20_error": 0.5020564348188254
+  },
+  "al_seconds": 669.3,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.0009152615352997819,
+    0.02158362259693189,
+    0.01696471714785591,
+    0.0040063807443571735,
+    0.9538208919175786,
+    0.00019836448588131552,
+    0.0008296056124843809,
+    0.0016811559596108942
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9538208919175786,
+    "entropy": 0.24388029175697593,
+    "effective_sources": 1.2761915506968406,
+    "min_weight": 0.00019836448588131552
+  },
+  "top_source_final": "microns_basil",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7511768441625362,
+    "patch_iou": 0.6116278656599864,
+    "patch_precision": 0.7322982005032396,
+    "patch_recall": 0.7746409359216964,
+    "patch_vi": 0.5222147527748564,
+    "n_patches": 735,
+    "cal_ece": 0.02511359686460786,
+    "cal_nll": 0.19427339443574926,
+    "cal_brier": 0.060363387699489846,
+    "cal_mean_confidence": 0.9400941984802542,
+    "cal_accuracy": 0.9149806016156462,
+    "cal_predicted_prevalence": 0.21563031462585033,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21331799387816586
+  },
+  "test_notta": {
+    "patch_dice": 0.7511768441625362,
+    "patch_iou": 0.6116278656599864,
+    "patch_precision": 0.7322982005032396,
+    "patch_recall": 0.7746409359216964,
+    "patch_vi": 0.5222147527748564,
+    "n_patches": 735,
+    "cal_ece": 0.02511359686460786,
+    "cal_nll": 0.19427339443574926,
+    "cal_brier": 0.060363387699489846,
+    "cal_mean_confidence": 0.9400941984802542,
+    "cal_accuracy": 0.9149806016156462,
+    "cal_predicted_prevalence": 0.21563031462585033,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21331799387816586
+  },
+  "test": {
+    "patch_dice": 0.7595023671503549,
+    "patch_iou": 0.6220891369072583,
+    "patch_precision": 0.7419020438291963,
+    "patch_recall": 0.7812715250362608,
+    "patch_vi": 0.5085818716439209,
+    "n_patches": 735,
+    "image_dice": 0.8065017348049895,
+    "image_iou": 0.6762415070696743,
+    "image_precision": 0.7927211103610438,
+    "image_recall": 0.8211628232225955,
+    "image_vi": 0.5176808942475893,
+    "n_images": 15,
+    "cal_ece": 0.01706907210860156,
+    "cal_nll": 0.1809449537218227,
+    "cal_brier": 0.05709824919899649,
+    "cal_mean_confidence": 0.935303017479304,
+    "cal_accuracy": 0.918304900085034,
+    "cal_predicted_prevalence": 0.21492081207482994,
+    "cal_true_prevalence": 0.20588727678571428,
+    "cal_mean_probability": 0.21349476381479376
+  },
+  "eval_seconds": 11.6,
+  "total_seconds": 901.3,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.14,
+        "peak_cpu_rss_mb": 1027.6,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 24.99,
+        "peak_cpu_rss_mb": 1810.2,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 192.59,
+        "peak_cpu_rss_mb": 2896.7,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4638.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 669.28,
+        "peak_cpu_rss_mb": 2896.7,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4646.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 11.6,
+        "peak_cpu_rss_mb": 2896.7,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4646.0,
+        "patches": 735,
+        "images": 15,
+        "patches_per_second": 63.4,
+        "seconds_per_image": 0.773
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 2896.7,
+    "total_seconds": 899.6
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/snemi3d/macs-v11-norestart-acq-random_b5_s3.pth"
+}

macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-random/budget-5/seed-3/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,28,28.4,156.9,"[0.36056026860693446, 0.4508413766529139, 0.21750134794876225, 0.28356610561779927, 0.560875393102026, 0.18601963732494034, 0.3724686404052985, 0.28968887034603186]",,288982,28,True,kept,0.4024189407361833,0.560875393102026,"[0.0013803357494966194, 0.03734608199477827, 0.024685983611284195, 0.0033285842306526635, 0.9269546251229578, 0.000312843356437377, 0.003307582804027515, 0.00268396313036561]",mixture,"{'mixture': 0.5602980725744923, 'uniform': 0.40469001951724776, 'top1': 0.560875393102026, 'top2': 0.5626323977653401, 'top3': 0.5627847905743437}",0.727546285788197,0.5749785964732628,0.011788481275023454,,,,,
+2,56,0.2,14.6,"[0.3479823236552919, 0.4186981967226685, 0.20557587495698246, 0.3007128919242538, 0.5589578670606897, 0.17607239437484073, 0.3247645305942548, 0.28265004696019047]",,565126,56,False,,,,"[0.0012569812895489763, 0.02796522949451945, 0.022627061811221816, 0.004080341346961101, 0.9390747514634746, 0.00029247988784289957, 0.0021198435345643842, 0.0025833111718668077]",mixture,"{'mixture': 0.5571876015358332, 'uniform': 0.374202964280938, 'top1': 0.5589578670606897, 'top2': 0.5592919600765673, 'top3': 0.5593125851518601}",0.7399397755468825,0.5367473626096051,0.012806996872993453,0.6853953687609664,0.5542848049226852,student,0.618,
+3,84,0.2,107.8,"[0.3557500469410345, 0.4409271340874671, 0.21965401520887193, 0.3356542715086659, 0.6156078106682936, 0.1806652407278319, 0.29110385747882406, 0.2844647798935356]","[0.36221583553452585, 0.46154148860749067, 0.2351214263672385, 0.36298119639097676, 0.6642216725599263, 0.19426701924887152, 0.2809982631984566, 0.2902214086069471]",846975,84,False,,,,"[0.0007864934568305689, 0.020220362781744897, 0.015079962329175009, 0.0033502443752500594, 0.9579861938327272, 0.0001772858169266955, 0.0008764938397664921, 0.0015229635675790894]",mixture,"{'mixture': 0.6149834430342458, 'uniform': 0.41044819487495626, 'top1': 0.6156078106682936, 'top2': 0.6161620454596852, 'top3': 0.6161824348043965}",0.7493534762151666,0.537037416881874,0.022052875547815737,0.7165017188224702,0.5822070619568362,student,0.722,
+4,112,0.2,111.1,"[0.36799533483144575, 0.4434738198437197, 0.22237860981915997, 0.35119162408855786, 0.6132689554994883, 0.18307297983849213, 0.281595677466981, 0.2906295936185792]","[0.38838436897907, 0.47321562487468605, 0.23612988589273762, 0.3855286998335696, 0.6543285906911006, 0.19182863371607134, 0.26864523780903776, 0.30642074866250274]",1109045,112,False,,,,"[0.0009075705663560531, 0.021176478328176308, 0.01582113731605874, 0.003995388643561608, 0.9554465806535456, 0.00018541093767134145, 0.0008136920001201178, 0.0016537415545100844]",mixture,"{'mixture': 0.6125056280699729, 'uniform': 0.4087578286493045, 'top1': 0.6132689554994883, 'top2': 0.6137450526434824, 'top3': 0.6137686488987392}",0.7508233511614852,0.5353514012696726,0.024242308398439487,0.7228817582417583,0.5876644214959471,student,0.743,
+5,140,0.2,115.0,"[0.36545827887620963, 0.44001314710549466, 0.2250317553851608, 0.3482704748138672, 0.6108450152408786, 0.18098147746558904, 0.2772557088193389, 0.28833909699736204]","[0.38441743915813703, 0.4699472512144608, 0.242043795701895, 0.38036188631835627, 0.6520519630893523, 0.19162936980644238, 0.28072756803528426, 0.30415916811101773]",1372470,140,False,,,,"[0.0009060261815732048, 0.020946085232241562, 0.016635612810557043, 0.0039732992909113215, 0.9549002581410955, 0.00018592196889683152, 0.0007977934031757694, 0.0016550029715487065]",mixture,"{'mixture': 0.61001214706305, 'uniform': 0.40779566499551556, 'top1': 0.6108450152408786, 'top2': 0.6112827862046616, 'top3': 0.6113057282316456}",0.7525420432275504,0.5316956049707146,0.02357947966251423,0.7077233122972116,0.5583120517817071,student,0.692,
+6,168,0.2,118.4,"[0.3634730961229432, 0.4400121039280854, 0.22399140525566236, 0.3461002739895306, 0.6077325689334545, 0.18446004079543676, 0.27816643604219415, 0.2869076314362251]","[0.37718160768208964, 0.4658483575908974, 0.24283830333599785, 0.37970663291846574, 0.6464730400378235, 0.19800368751348005, 0.2827052621021349, 0.29906355224330206]",1676005,168,False,,,,"[0.0009152615352997819, 0.02158362259693189, 0.01696471714785591, 0.0040063807443571735, 0.9538208919175786, 0.00019836448588131552, 0.0008296056124843809, 0.0016811559596108942]",mixture,"{'mixture': 0.6068742421539629, 'uniform': 0.4065183273644009, 'top1': 0.6077325689334545, 'top2': 0.6082575257609559, 'top3': 0.6082816859281865}",0.7529501278915827,0.5239912101194711,0.02575743938369193,0.7569865654765929,0.5640826712486684,student,0.857,
+7,168,0.0,35.2,,,,,,,,,,,,0.7560981292366642,0.5234298513862702,0.028411418873243124,,,,,True

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-3/log.txt

@@ -0,0 +1,106 @@
+[10:48:45][      0s][START] target=whole-mouse-brain variant=macs-v11-norestart-acq-entropy budget=5.0% seed=3 | {"task": "membrane", "target": "whole-mouse-brain", "variant": "macs-v11-norestart-acq-entropy", "budget": 5.0, "seed": 3, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "entropy", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[10:48:45][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[10:48:45][      0s][DATA] whole-mouse-brain: train=5925 val=1250 test=1300 patches
+[10:48:46][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, snemi=3.55
+[10:48:47][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[10:48:53][      7s][WEIGHTS] mixture: c-elegans=0.008, cerebellum=0.256, flywire=0.039, berghia=0.075, microns=0.376, oct-glia=0.134, oct-sfl=0.096, snemi=0.016 | eff.sources=5.13 | MMD2 mix=0.1295 uniform=0.1553
+[10:48:53][      7s][WEIGHTS] per-source: c-elegans=0.145, cerebellum=0.042, flywire=0.001, berghia=0.142, microns=0.020, oct-glia=0.402, oct-sfl=0.243, snemi=0.004 | min-MMD source (NeuroADDA rule) = oct-glia
+[10:49:16][     31s][TEACHERS] target-val Dice: c-elegans=0.386, cerebellum=0.316, flywire=0.110, berghia=0.259, microns=0.329, oct-glia=0.407, oct-sfl=0.440, snemi=0.305
+[10:49:16][     31s][TEACHERS] ensemble target-val Dice: mixture=0.385, mixture@raw=0.400, mixture@logit=0.380, per_source=0.420, per_source@raw=0.422, per_source@logit=0.419, uniform=0.413, uniform@raw=0.388, uniform@logit=0.405
+[10:49:16][     31s][WEIGHTS] using mixture: c-elegans=0.008, cerebellum=0.256, flywire=0.039, berghia=0.075, microns=0.376, oct-glia=0.134, oct-sfl=0.096, snemi=0.016
+[10:49:16][     31s][INIT] student initialised from teacher microns (weight 0.376)
+[10:49:18][     33s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[10:49:18][     33s][KD] distilling the mixture ensemble (prob space) into the student
+[10:50:25][    100s][KD] soft labels for 5925 patches from 8 teachers in 67s
+[10:51:18][    153s][KD] epoch 1/5 | patches=5925 | loss=0.6690 | 53s
+[10:52:11][    205s][KD] epoch 2/5 | patches=5925 | loss=0.5006 | 52s
+[10:53:03][    258s][KD] epoch 3/5 | patches=5925 | loss=0.4703 | 52s
+[10:53:55][    310s][KD] epoch 4/5 | patches=5925 | loss=0.4599 | 52s
+[10:54:46][    361s][KD] epoch 5/5 | patches=5925 | loss=0.4555 | 51s
+[10:54:48][    363s][ROUND0] before any target label: val_dice=0.4218 val_vi=0.5128
+[10:54:48][    363s][AL] patches=5925 (pool 5925 after dropping blank; free blank negatives: True) | budget=296 (5.0%) | rounds=6 | per_round=49 | acquisition=entropy | first_round=stratified
+[10:54:57][    372s][AL] round-1 uncertainty quality on 5925 patches: {'spearman_uncertainty_error': -0.20489397455499517, 'auroc_top20_error': 0.39909434029446844}
+[10:55:20][    395s][AL] round 1: stratified cold start: 25 patches from the membrane-rich half (ensemble prevalence >= 0.215) + 24 from the membrane-poor half of 2000 candidates
+[10:55:23][    398s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.433, uniform=0.435, top1=0.471, top2=0.445, top3=0.446 -> using top1
+[10:55:23][    398s][WEIGHTS] round 1: student kept (labeled Dice 0.396 vs best teacher oct-sfl 0.471); distilling the corrected ensemble into it
+[10:55:23][    398s][WEIGHTS] round 1: teacher Dice on 49 labeled patches + 0 blank (640544 fg px): c-elegans=0.367, cerebellum=0.350, flywire=0.141, berghia=0.235, microns=0.310, oct-glia=0.385, oct-sfl=0.471, snemi=0.254 | weights: c-elegans=0.000, cerebellum=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=1.000, snemi=0.000
+[10:55:32][    407s][KD] soft labels for 5876 patches from 8 teachers in 9s
+[10:56:23][    458s][KD] epoch 1/2 | patches=5876 | loss=0.1745 | 51s
+[10:57:14][    508s][KD] epoch 2/2 | patches=5876 | loss=0.1632 | 51s
+[10:57:17][    512s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 3s
+[10:57:18][    513s][FT] round 1/6 epoch 1/5 | patches=49 | loss=1.3008
+[10:57:19][    514s][FT] round 1/6 epoch 2/5 | patches=49 | loss=1.0758
+[10:57:21][    516s][FT] round 1/6 epoch 3/5 | patches=49 | loss=1.3884
+[10:57:22][    517s][FT] round 1/6 epoch 4/5 | patches=49 | loss=1.0120
+[10:57:23][    518s][FT] round 1/6 epoch 5/5 | patches=49 | loss=1.0581
+[10:57:25][    520s][AL] round 1/6 | labeled=49 | selection_seconds=32.2000 round_seconds=155.3000 restart_student_dice=0.3963 restart_teacher_dice=0.4707 val_dice=0.6465 val_vi=0.4665 val_ece=0.0076
+[10:57:34][    529s][AL] round 2: 530/5876 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:57:39][    534s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.491, uniform=0.279, top1=0.461, top2=0.461, top3=0.460 -> using mixture
+[10:57:39][    534s][WEIGHTS] round 2: teacher Dice on 98 labeled patches + 0 blank (1735724 fg px): c-elegans=0.396, cerebellum=0.461, flywire=0.101, berghia=0.203, microns=0.255, oct-glia=0.258, oct-sfl=0.332, snemi=0.393 | weights: c-elegans=0.011, cerebellum=0.699, flywire=0.003, berghia=0.015, microns=0.130, oct-glia=0.048, oct-sfl=0.071, snemi=0.022
+[10:57:41][    536s][CONS] round 2: on the 49 new patches student Dice=0.548 vs ensemble 0.557 -> pseudo-labels from the ensemble | consistency x1.00
+[10:57:43][    538s][FT] round 2/6 epoch 1/5 | patches=98 | loss=1.2651
+[10:57:45][    540s][FT] round 2/6 epoch 2/5 | patches=98 | loss=1.3382
+[10:57:47][    542s][FT] round 2/6 epoch 3/5 | patches=98 | loss=1.1291
+[10:57:49][    544s][FT] round 2/6 epoch 4/5 | patches=98 | loss=1.0845
+[10:57:51][    546s][FT] round 2/6 epoch 5/5 | patches=98 | loss=1.0443
+[10:57:53][    548s][AL] round 2/6 | labeled=98 | selection_seconds=9.0000 round_seconds=26.0000 pseudo_student_dice=0.5477 pseudo_ensemble_dice=0.5568 consistency_scale=1.0000 val_dice=0.6380 val_vi=0.5154 val_ece=0.0264
+[10:58:02][    557s][AL] round 3: 503/5827 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:58:09][    564s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.460, uniform=0.204, top1=0.456, top2=0.456, top3=0.456 -> using mixture
+[10:58:09][    564s][WEIGHTS] round 3: teacher Dice on 147 labeled patches + 0 blank (2564465 fg px): c-elegans=0.356, cerebellum=0.456, flywire=0.073, berghia=0.161, microns=0.188, oct-glia=0.189, oct-sfl=0.246, snemi=0.399 | weights: c-elegans=0.009, cerebellum=0.802, flywire=0.003, berghia=0.012, microns=0.080, oct-glia=0.029, oct-sfl=0.037, snemi=0.028
+[10:58:11][    566s][CONS] round 3: on the 49 new patches student Dice=0.221 vs ensemble 0.440 -> pseudo-labels from the ensemble | consistency x1.00
+[10:58:14][    569s][FT] round 3/6 epoch 1/5 | patches=147 | loss=1.2710
+[10:58:17][    572s][FT] round 3/6 epoch 2/5 | patches=147 | loss=1.3728
+[10:58:20][    575s][FT] round 3/6 epoch 3/5 | patches=147 | loss=1.2233
+[10:58:23][    578s][FT] round 3/6 epoch 4/5 | patches=147 | loss=1.1298
+[10:58:26][    581s][FT] round 3/6 epoch 5/5 | patches=147 | loss=1.0388
+[10:58:28][    583s][AL] round 3/6 | labeled=147 | selection_seconds=9.0000 round_seconds=32.4000 pseudo_student_dice=0.2214 pseudo_ensemble_dice=0.4398 consistency_scale=1.0000 val_dice=0.6743 val_vi=0.4691 val_ece=0.0284
+[10:58:37][    591s][AL] round 4: 697/5778 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[10:58:46][    601s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.389, uniform=0.206, top1=0.388, top2=0.388, top3=0.388 -> using mixture
+[10:58:46][    601s][WEIGHTS] round 4: teacher Dice on 196 labeled patches + 0 blank (3036531 fg px): c-elegans=0.330, cerebellum=0.388, flywire=0.072, berghia=0.168, microns=0.184, oct-glia=0.198, oct-sfl=0.243, snemi=0.357 | weights: c-elegans=0.011, cerebellum=0.685, flywire=0.004, berghia=0.022, microns=0.132, oct-glia=0.054, oct-sfl=0.060, snemi=0.031
+[10:58:51][    606s][CONS] round 4: on the 49 new patches student Dice=0.459 vs ensemble 0.213 -> pseudo-labels from the student | consistency x0.00
+[10:58:53][    608s][FT] round 4/6 epoch 1/5 | patches=196 | loss=1.0920
+[10:58:55][    610s][FT] round 4/6 epoch 2/5 | patches=196 | loss=0.9350
+[10:58:57][    612s][FT] round 4/6 epoch 3/5 | patches=196 | loss=0.9393
+[10:58:59][    614s][FT] round 4/6 epoch 4/5 | patches=196 | loss=0.9013
+[10:59:01][    616s][FT] round 4/6 epoch 5/5 | patches=196 | loss=0.9538
+[10:59:03][    618s][AL] round 4/6 | labeled=196 | selection_seconds=9.0000 round_seconds=32.9000 pseudo_student_dice=0.4592 pseudo_ensemble_dice=0.2127 consistency_scale=0.0000 val_dice=0.7087 val_vi=0.4190 val_ece=0.0029
+[10:59:11][    626s][AL] round 5: 567/5729 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:00:03][    678s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.413, cerebellum=0.421, flywire=0.151, berghia=0.286, microns=0.320, oct-glia=0.441, oct-sfl=0.539, snemi=0.334 | blended with labeled Dice at 0.50
+[11:00:52][    727s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.450, uniform=0.317, top1=0.412, top2=0.429, top3=0.429 -> using mixture
+[11:00:52][    727s][WEIGHTS] round 5: teacher Dice on 245 labeled patches + 0 blank (4134678 fg px): c-elegans=0.381, cerebellum=0.412, flywire=0.107, berghia=0.232, microns=0.233, oct-glia=0.319, oct-sfl=0.385, snemi=0.362 | weights: c-elegans=0.012, cerebellum=0.540, flywire=0.004, berghia=0.026, microns=0.132, oct-glia=0.112, oct-sfl=0.154, snemi=0.020
+[11:00:57][    732s][CONS] round 5: on the 49 new patches student Dice=0.700 vs ensemble 0.319 -> pseudo-labels from the student | consistency x0.67
+[11:01:01][    736s][FT] round 5/6 epoch 1/5 | patches=245 | loss=1.0706
+[11:01:06][    741s][FT] round 5/6 epoch 2/5 | patches=245 | loss=1.2142
+[11:01:11][    745s][FT] round 5/6 epoch 3/5 | patches=245 | loss=1.0638
+[11:01:15][    750s][FT] round 5/6 epoch 4/5 | patches=245 | loss=1.0411
+[11:01:20][    754s][FT] round 5/6 epoch 5/5 | patches=245 | loss=1.0409
+[11:01:22][    757s][AL] round 5/6 | labeled=245 | selection_seconds=8.8000 round_seconds=136.9000 pseudo_student_dice=0.6995 pseudo_ensemble_dice=0.3187 consistency_scale=0.6650 val_dice=0.7157 val_vi=0.3772 val_ece=0.0090
+[11:01:30][    765s][AL] round 6: 707/5680 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:02:23][    818s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.392, cerebellum=0.395, flywire=0.157, berghia=0.265, microns=0.334, oct-glia=0.443, oct-sfl=0.542, snemi=0.297 | blended with labeled Dice at 0.50
+[11:03:13][    868s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.419, uniform=0.323, top1=0.399, top2=0.419, top3=0.419 -> using mixture
+[11:03:13][    868s][WEIGHTS] round 6: teacher Dice on 294 labeled patches + 0 blank (5686541 fg px): c-elegans=0.385, cerebellum=0.399, flywire=0.111, berghia=0.251, microns=0.231, oct-glia=0.326, oct-sfl=0.397, snemi=0.368 | weights: c-elegans=0.013, cerebellum=0.490, flywire=0.004, berghia=0.033, microns=0.134, oct-glia=0.124, oct-sfl=0.179, snemi=0.022
+[11:03:18][    873s][CONS] round 6: on the 49 new patches student Dice=0.482 vs ensemble 0.215 -> pseudo-labels from the student | consistency x0.00
+[11:03:21][    876s][FT] round 6/6 epoch 1/5 | patches=294 | loss=1.0749
+[11:03:24][    879s][FT] round 6/6 epoch 2/5 | patches=294 | loss=0.9716
+[11:03:27][    882s][FT] round 6/6 epoch 3/5 | patches=294 | loss=0.9355
+[11:03:30][    885s][FT] round 6/6 epoch 4/5 | patches=294 | loss=0.9278
+[11:03:33][    887s][FT] round 6/6 epoch 5/5 | patches=294 | loss=0.8386
+[11:03:34][    889s][AL] round 6/6 | labeled=294 | selection_seconds=8.8000 round_seconds=130.9000 pseudo_student_dice=0.4819 pseudo_ensemble_dice=0.2152 consistency_scale=0.0000 val_dice=0.7126 val_vi=0.3923 val_ece=0.0091
+[11:03:37][    892s][CONS] final consolidation: pseudo-labels refreshed from the student
+[11:03:40][    895s][FT] final 1/10 epoch 1/1 | patches=294 | loss=0.8899
+[11:03:43][    898s][FT] final 2/10 epoch 1/1 | patches=294 | loss=0.8436
+[11:03:46][    901s][FT] final 3/10 epoch 1/1 | patches=294 | loss=0.7589
+[11:03:48][    903s][FT] final 4/10 epoch 1/1 | patches=294 | loss=0.8406
+[11:03:51][    906s][FT] final 5/10 epoch 1/1 | patches=294 | loss=0.8507
+[11:03:54][    909s][FT] final 6/10 epoch 1/1 | patches=294 | loss=0.7447
+[11:03:57][    912s][FT] final 7/10 epoch 1/1 | patches=294 | loss=0.7217
+[11:04:00][    914s][FT] final 8/10 epoch 1/1 | patches=294 | loss=0.6748
+[11:04:02][    917s][FT] final 9/10 epoch 1/1 | patches=294 | loss=0.7146
+[11:04:05][    920s][FT] final 10/10 epoch 1/1 | patches=294 | loss=0.6861
+[11:04:07][    922s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=30.5000 val_dice=0.7200 val_vi=0.3878 val_ece=0.0197
+[11:04:07][    922s][WEIGHTS] final: c-elegans=0.013, cerebellum=0.490, flywire=0.004, berghia=0.033, microns=0.134, oct-glia=0.124, oct-sfl=0.179, snemi=0.022
+[11:04:25][    939s][TEST] raw student (no EMA, no TTA): patch_dice=0.7044 patch_vi=0.3511
+[11:04:27][    942s][TEST] final model without TTA: patch_dice=0.7044 patch_vi=0.3511
+[11:04:27][    942s][RESOURCES] {"phases": {"setup": {"seconds": 1.12, "peak_cpu_rss_mb": 1027.4, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 31.35, "peak_cpu_rss_mb": 1809.2, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 328.23, "peak_cpu_rss_mb": 3676.9, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 558.98, "peak_cpu_rss_mb": 3676.9, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 20.15, "peak_cpu_rss_mb": 3676.9, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 1300, "images": 52, "patches_per_second": 64.5, "seconds_per_image": 0.387}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 3676.9, "total_seconds": 939.8}
+[11:04:27][    942s][TEST] patch_dice=0.7253 patch_iou=0.6459 patch_precision=0.7409 patch_recall=0.8422 patch_vi=0.3269 image_dice=0.8342 image_iou=0.7308 image_precision=0.8019 image_recall=0.8938 image_vi=0.3823 cal_ece=0.0061 cal_nll=0.1796 cal_brier=0.0510 cal_mean_confidence=0.9385 cal_accuracy=0.9328 cal_predicted_prevalence=0.2326 cal_true_prevalence=0.2108 cal_mean_probability=0.2424
+[11:04:27][    942s][DONE] whole-mouse-brain macs-v11-norestart-acq-entropy b5 s3 | test patch Dice=0.7253 VI=0.3269 | image Dice=0.8342 VI=0.3823 | 942s

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-3/result.json

@@ -0,0 +1,767 @@
+{
+  "target": "whole-mouse-brain",
+  "variant": "macs-v11-norestart-acq-entropy",
+  "budget": 5.0,
+  "seed": 3,
+  "config": {
+    "task": "membrane",
+    "target": "whole-mouse-brain",
+    "variant": "macs-v11-norestart-acq-entropy",
+    "budget": 5.0,
+    "seed": 3,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "entropy",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "whole-mouse-brain",
+      "--variants",
+      "macs-v11-norestart-acq-entropy",
+      "--budgets",
+      "5",
+      "--seeds",
+      "3",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "snemi3d"
+  ],
+  "alphas": [
+    0.007617762850063339,
+    0.2559312601209969,
+    0.03865993944852264,
+    0.07533345106002441,
+    0.3764169834180784,
+    0.13449516240019677,
+    0.09581340232315401,
+    0.0157320383789636
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "snemi3d"
+    ],
+    "mixture": [
+      0.007617762850063339,
+      0.2559312601209969,
+      0.03865993944852264,
+      0.07533345106002441,
+      0.3764169834180784,
+      0.13449516240019677,
+      0.09581340232315401,
+      0.0157320383789636
+    ],
+    "per_source": [
+      0.1450746039368088,
+      0.041950766450651496,
+      0.0008814721099573598,
+      0.1418325124471684,
+      0.02035312744213396,
+      0.4023730240336797,
+      0.24338555075976573,
+      0.004148942819834595
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.11311670518772092,
+      "mmd2_mixture": 0.12945823384265215,
+      "mmd2_uniform": 0.155288743679944,
+      "mmd2_single_source": [
+        0.38071639717042494,
+        0.17697894982207135,
+        0.3194061281637849,
+        0.24037642211575033,
+        0.16266554797152122,
+        0.2734526560532348,
+        0.2563986706812869,
+        0.2660414632502267
+      ],
+      "median_squared_distance": 0.18592532667757627,
+      "gammas": [
+        1.3446258477390725,
+        2.689251695478145,
+        5.37850339095629,
+        10.75700678191258,
+        21.51401356382516
+      ],
+      "max_weight": 0.3764169834180784,
+      "entropy": 1.6341528654931232,
+      "effective_sources": 5.125114496180783,
+      "min_weight": 0.007617762850063339
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.33220085501273156,
+        0.46442573060595516,
+        0.8760630244687474,
+        0.33460943356720163,
+        0.5415026072699861,
+        0.2234870921926152,
+        0.2770625113402161,
+        0.7109869456345189
+      ],
+      "standardized": [
+        -0.6467275403932012,
+        -0.02635183204683673,
+        1.9049774780863755,
+        -0.6354269162093369,
+        0.33527922824378936,
+        -1.1567932698101913,
+        -0.9054268876938927,
+        1.130469739823295
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4023730240336797,
+      "entropy": 1.508581788687784,
+      "effective_sources": 4.520315483702711,
+      "min_weight": 0.0008814721099573598
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.38573376502042933,
+      0.31643357804427197,
+      0.10997258971952562,
+      0.2588485480329741,
+      0.3288178379711084,
+      0.4072116167356654,
+      0.4397876396557443,
+      0.30535999081848075
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.3845088921696238,
+      "mixture@raw": 0.3998907914691259,
+      "mixture@logit": 0.3802050170706561,
+      "per_source": 0.4199022574247113,
+      "per_source@raw": 0.42186088103254527,
+      "per_source@logit": 0.41929713845028616,
+      "uniform": 0.4128319943503538,
+      "uniform@raw": 0.3877140229453584,
+      "uniform@logit": 0.4054972948683435
+    },
+    "val_patches": 1024
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    3.5461363792419434
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.3764169834180784,
+    "entropy": 1.6341528654931232,
+    "effective_sources": 5.125114496180783,
+    "min_weight": 0.007617762850063339
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.261904761904762,
+    "spearman_weight_taxonomy": 0.39440531887330776,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 328.2,
+  "round0_val": {
+    "val_dice": 0.42175774709778946,
+    "val_vi": 0.5128062998414359,
+    "val_ece": 0.14443836415563802
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 49,
+      "selection_seconds": 32.2,
+      "round_seconds": 155.3,
+      "teacher_dice_labeled": [
+        0.36738741011177045,
+        0.34973487377165097,
+        0.14100291682142935,
+        0.2353012898060949,
+        0.30996524321190877,
+        0.38452188964078277,
+        0.47067622607709797,
+        0.25385926187706875
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 640544,
+      "evidence_nonempty_patches": 39,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.3962524167755847,
+      "restart_teacher_dice": 0.47067622607709797,
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.43280781937383384,
+        "uniform": 0.43493152236966137,
+        "top1": 0.47067622607709797,
+        "top2": 0.44454578875106776,
+        "top3": 0.4462152205328611
+      },
+      "val_dice": 0.646540074905994,
+      "val_vi": 0.4665198719972955,
+      "val_ece": 0.0075807375838696745
+    },
+    {
+      "round": 2,
+      "n_labeled": 98,
+      "selection_seconds": 9.0,
+      "round_seconds": 26.0,
+      "teacher_dice_labeled": [
+        0.3964949872852674,
+        0.46138364251508823,
+        0.1010757881063798,
+        0.20256512904065874,
+        0.25460547393201843,
+        0.2577079727691892,
+        0.33156287339796325,
+        0.39328567712650325
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1735724,
+      "evidence_nonempty_patches": 84,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.010880776159647547,
+        0.699461172663094,
+        0.0028780896424400485,
+        0.015473680225314972,
+        0.1301018016412744,
+        0.04795067726707895,
+        0.07149276637716834,
+        0.02176103602398173
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.4910491059378369,
+        "uniform": 0.2787527762668842,
+        "top1": 0.46138364251508823,
+        "top2": 0.4609649416796084,
+        "top3": 0.45997326697860197
+      },
+      "pseudo_student_dice": 0.5476616446618572,
+      "pseudo_ensemble_dice": 0.556802486489929,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6379528251954637,
+      "val_vi": 0.5153946538049008,
+      "val_ece": 0.02644548907751808
+    },
+    {
+      "round": 3,
+      "n_labeled": 147,
+      "selection_seconds": 9.0,
+      "round_seconds": 32.4,
+      "teacher_dice_labeled": [
+        0.35633572198115654,
+        0.4561676677016383,
+        0.07341647254380124,
+        0.16051329920740956,
+        0.18766835678359275,
+        0.18921733847825212,
+        0.24562380051934984,
+        0.39910204344838873
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 2564465,
+      "evidence_nonempty_patches": 113,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.008798245853375434,
+        0.8021518907392878,
+        0.0026371032015051134,
+        0.012277516649360495,
+        0.08048682486773641,
+        0.029207164112188905,
+        0.036574562641144275,
+        0.027866691935401616
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.460341006527611,
+        "uniform": 0.2037175342073936,
+        "top1": 0.4561676677016383,
+        "top2": 0.4559419693890312,
+        "top3": 0.45570403413587507
+      },
+      "pseudo_student_dice": 0.22137009624307094,
+      "pseudo_ensemble_dice": 0.4397540958983941,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6743464690510582,
+      "val_vi": 0.4690869017455335,
+      "val_ece": 0.02842435821653777
+    },
+    {
+      "round": 4,
+      "n_labeled": 196,
+      "selection_seconds": 9.0,
+      "round_seconds": 32.9,
+      "teacher_dice_labeled": [
+        0.3299771096263301,
+        0.38804163172125516,
+        0.07160612127896214,
+        0.16794057559069106,
+        0.18443843255050638,
+        0.19807899270143609,
+        0.24278902382700712,
+        0.3569940772081713
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 3036531,
+      "evidence_nonempty_patches": 153,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.01141576602419868,
+        0.6854439711596396,
+        0.004373688359092133,
+        0.022333137905060703,
+        0.13160742045978957,
+        0.05389618569239227,
+        0.060041402346467036,
+        0.030888428053359913
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.3890792156145218,
+        "uniform": 0.20577203391014331,
+        "top1": 0.38804163172125516,
+        "top2": 0.3879763730386121,
+        "top3": 0.38780666396278846
+      },
+      "pseudo_student_dice": 0.45920203220412686,
+      "pseudo_ensemble_dice": 0.21272979415981438,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.7087167370495862,
+      "val_vi": 0.4190034422858896,
+      "val_ece": 0.002885292617739875
+    },
+    {
+      "round": 5,
+      "n_labeled": 245,
+      "selection_seconds": 8.8,
+      "round_seconds": 136.9,
+      "teacher_dice_labeled": [
+        0.38112177883028797,
+        0.4121840427291803,
+        0.10655492598420999,
+        0.23207856359270612,
+        0.2329279883222823,
+        0.3189693538938384,
+        0.3850992144352713,
+        0.362244453467251
+      ],
+      "teacher_dice_pool": [
+        0.4131676741656513,
+        0.42103200046233735,
+        0.1509888135870009,
+        0.2862589597725059,
+        0.3200670355049752,
+        0.4407886865128643,
+        0.5394127796666625,
+        0.3344276364677657
+      ],
+      "evidence_fg_pixels": 4134678,
+      "evidence_nonempty_patches": 193,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.011779082999612873,
+        0.5398947030835544,
+        0.003838101998014584,
+        0.0262413054958334,
+        0.13223785280483505,
+        0.11170333649150943,
+        0.15416439364109832,
+        0.020141223485542006
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.45044626372960384,
+        "uniform": 0.3174082551184142,
+        "top1": 0.4121840427291803,
+        "top2": 0.4287747359626445,
+        "top3": 0.42858896234699534
+      },
+      "pseudo_student_dice": 0.6995360803402509,
+      "pseudo_ensemble_dice": 0.31871799632824066,
+      "pseudo_source": "student",
+      "consistency_scale": 0.665,
+      "val_dice": 0.7156779814900249,
+      "val_vi": 0.3771571907269592,
+      "val_ece": 0.009027793309739329
+    },
+    {
+      "round": 6,
+      "n_labeled": 294,
+      "selection_seconds": 8.8,
+      "round_seconds": 130.9,
+      "teacher_dice_labeled": [
+        0.3854258140408431,
+        0.399042101167703,
+        0.11107803619300319,
+        0.250959391571319,
+        0.23089472687251988,
+        0.32607846011094965,
+        0.3967225724361204,
+        0.36848059917839887
+      ],
+      "teacher_dice_pool": [
+        0.39226917381905135,
+        0.3954502206402,
+        0.1565966359168242,
+        0.265017680659922,
+        0.3343952632446626,
+        0.4432951310911836,
+        0.541844091574079,
+        0.2973078419837905
+      ],
+      "evidence_fg_pixels": 5686541,
+      "evidence_nonempty_patches": 237,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.012736176365196153,
+        0.4903088525913747,
+        0.004159063249292493,
+        0.03282609531881109,
+        0.13420257271254782,
+        0.12421555848074521,
+        0.17934904829414433,
+        0.02220263298788817
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.4187754258232841,
+        "uniform": 0.3225095176404339,
+        "top1": 0.399042101167703,
+        "top2": 0.4194592007402292,
+        "top3": 0.41927779087623196
+      },
+      "pseudo_student_dice": 0.48187702682941513,
+      "pseudo_ensemble_dice": 0.21516510877757866,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.7125637098332068,
+      "val_vi": 0.3922931931632452,
+      "val_ece": 0.009128083181487133
+    },
+    {
+      "round": 7,
+      "n_labeled": 294,
+      "selection_seconds": 0.0,
+      "round_seconds": 30.5,
+      "final_consolidation": true,
+      "val_dice": 0.7199579951007331,
+      "val_vi": 0.3878433725098941,
+      "val_ece": 0.01970009244399582
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.20489397455499517,
+    "auroc_top20_error": 0.39909434029446844
+  },
+  "al_seconds": 559.0,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.012736176365196153,
+    0.4903088525913747,
+    0.004159063249292493,
+    0.03282609531881109,
+    0.13420257271254782,
+    0.12421555848074521,
+    0.17934904829414433,
+    0.02220263298788817
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.4903088525913747,
+    "entropy": 1.461326814289988,
+    "effective_sources": 4.3116765288398,
+    "min_weight": 0.004159063249292493
+  },
+  "top_source_final": "dhanyasi-P14-mouse-cerebellum",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7044237441791064,
+    "patch_iou": 0.6208068695363332,
+    "patch_precision": 0.7148284002150658,
+    "patch_recall": 0.8299271833672399,
+    "patch_vi": 0.351098248100767,
+    "n_patches": 1300,
+    "cal_ece": 0.019415714926569648,
+    "cal_nll": 0.19742326733203194,
+    "cal_brier": 0.05522876654738458,
+    "cal_mean_confidence": 0.9472860574746466,
+    "cal_accuracy": 0.9278703425480769,
+    "cal_predicted_prevalence": 0.23288611778846155,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.23892273644979425
+  },
+  "test_notta": {
+    "patch_dice": 0.7044237441791064,
+    "patch_iou": 0.6208068695363332,
+    "patch_precision": 0.7148284002150658,
+    "patch_recall": 0.8299271833672399,
+    "patch_vi": 0.351098248100767,
+    "n_patches": 1300,
+    "cal_ece": 0.019415714926569648,
+    "cal_nll": 0.19742326733203194,
+    "cal_brier": 0.05522876654738458,
+    "cal_mean_confidence": 0.9472860574746466,
+    "cal_accuracy": 0.9278703425480769,
+    "cal_predicted_prevalence": 0.23288611778846155,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.23892273644979425
+  },
+  "test": {
+    "patch_dice": 0.7253355817647213,
+    "patch_iou": 0.6458858156903308,
+    "patch_precision": 0.7408743616027792,
+    "patch_recall": 0.842185801710432,
+    "patch_vi": 0.3268501847439328,
+    "n_patches": 1300,
+    "image_dice": 0.8342029779726485,
+    "image_iou": 0.7307789106547949,
+    "image_precision": 0.8018712059262382,
+    "image_recall": 0.8937876571487801,
+    "image_vi": 0.38229628640130375,
+    "n_images": 52,
+    "cal_ece": 0.006070841455767178,
+    "cal_nll": 0.17962633679523854,
+    "cal_brier": 0.05098272736595117,
+    "cal_mean_confidence": 0.9384511861769915,
+    "cal_accuracy": 0.9328049879807693,
+    "cal_predicted_prevalence": 0.23264197716346155,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.24242487424387058
+  },
+  "eval_seconds": 20.1,
+  "total_seconds": 942.0,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.12,
+        "peak_cpu_rss_mb": 1027.4,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 31.35,
+        "peak_cpu_rss_mb": 1809.2,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 328.23,
+        "peak_cpu_rss_mb": 3676.9,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4638.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 558.98,
+        "peak_cpu_rss_mb": 3676.9,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4646.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 20.15,
+        "peak_cpu_rss_mb": 3676.9,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4646.0,
+        "patches": 1300,
+        "images": 52,
+        "patches_per_second": 64.5,
+        "seconds_per_image": 0.387
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 3676.9,
+    "total_seconds": 939.8
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/whole-mouse-brain/macs-v11-norestart-acq-entropy_b5_s3.pth"
+}

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-3/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,49,32.2,155.3,"[0.36738741011177045, 0.34973487377165097, 0.14100291682142935, 0.2353012898060949, 0.30996524321190877, 0.38452188964078277, 0.47067622607709797, 0.25385926187706875]",,640544,39,True,kept,0.3962524167755847,0.47067622607709797,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.43280781937383384, 'uniform': 0.43493152236966137, 'top1': 0.47067622607709797, 'top2': 0.44454578875106776, 'top3': 0.4462152205328611}",0.646540074905994,0.4665198719972955,0.0075807375838696745,,,,,
+2,98,9.0,26.0,"[0.3964949872852674, 0.46138364251508823, 0.1010757881063798, 0.20256512904065874, 0.25460547393201843, 0.2577079727691892, 0.33156287339796325, 0.39328567712650325]",,1735724,84,False,,,,"[0.010880776159647547, 0.699461172663094, 0.0028780896424400485, 0.015473680225314972, 0.1301018016412744, 0.04795067726707895, 0.07149276637716834, 0.02176103602398173]",mixture,"{'mixture': 0.4910491059378369, 'uniform': 0.2787527762668842, 'top1': 0.46138364251508823, 'top2': 0.4609649416796084, 'top3': 0.45997326697860197}",0.6379528251954637,0.5153946538049008,0.02644548907751808,0.5476616446618572,0.556802486489929,ensemble,1.0,
+3,147,9.0,32.4,"[0.35633572198115654, 0.4561676677016383, 0.07341647254380124, 0.16051329920740956, 0.18766835678359275, 0.18921733847825212, 0.24562380051934984, 0.39910204344838873]",,2564465,113,False,,,,"[0.008798245853375434, 0.8021518907392878, 0.0026371032015051134, 0.012277516649360495, 0.08048682486773641, 0.029207164112188905, 0.036574562641144275, 0.027866691935401616]",mixture,"{'mixture': 0.460341006527611, 'uniform': 0.2037175342073936, 'top1': 0.4561676677016383, 'top2': 0.4559419693890312, 'top3': 0.45570403413587507}",0.6743464690510582,0.4690869017455335,0.02842435821653777,0.22137009624307094,0.4397540958983941,ensemble,1.0,
+4,196,9.0,32.9,"[0.3299771096263301, 0.38804163172125516, 0.07160612127896214, 0.16794057559069106, 0.18443843255050638, 0.19807899270143609, 0.24278902382700712, 0.3569940772081713]",,3036531,153,False,,,,"[0.01141576602419868, 0.6854439711596396, 0.004373688359092133, 0.022333137905060703, 0.13160742045978957, 0.05389618569239227, 0.060041402346467036, 0.030888428053359913]",mixture,"{'mixture': 0.3890792156145218, 'uniform': 0.20577203391014331, 'top1': 0.38804163172125516, 'top2': 0.3879763730386121, 'top3': 0.38780666396278846}",0.7087167370495862,0.4190034422858896,0.002885292617739875,0.45920203220412686,0.21272979415981438,student,0.0,
+5,245,8.8,136.9,"[0.38112177883028797, 0.4121840427291803, 0.10655492598420999, 0.23207856359270612, 0.2329279883222823, 0.3189693538938384, 0.3850992144352713, 0.362244453467251]","[0.4131676741656513, 0.42103200046233735, 0.1509888135870009, 0.2862589597725059, 0.3200670355049752, 0.4407886865128643, 0.5394127796666625, 0.3344276364677657]",4134678,193,False,,,,"[0.011779082999612873, 0.5398947030835544, 0.003838101998014584, 0.0262413054958334, 0.13223785280483505, 0.11170333649150943, 0.15416439364109832, 0.020141223485542006]",mixture,"{'mixture': 0.45044626372960384, 'uniform': 0.3174082551184142, 'top1': 0.4121840427291803, 'top2': 0.4287747359626445, 'top3': 0.42858896234699534}",0.7156779814900249,0.3771571907269592,0.009027793309739329,0.6995360803402509,0.31871799632824066,student,0.665,
+6,294,8.8,130.9,"[0.3854258140408431, 0.399042101167703, 0.11107803619300319, 0.250959391571319, 0.23089472687251988, 0.32607846011094965, 0.3967225724361204, 0.36848059917839887]","[0.39226917381905135, 0.3954502206402, 0.1565966359168242, 0.265017680659922, 0.3343952632446626, 0.4432951310911836, 0.541844091574079, 0.2973078419837905]",5686541,237,False,,,,"[0.012736176365196153, 0.4903088525913747, 0.004159063249292493, 0.03282609531881109, 0.13420257271254782, 0.12421555848074521, 0.17934904829414433, 0.02220263298788817]",mixture,"{'mixture': 0.4187754258232841, 'uniform': 0.3225095176404339, 'top1': 0.399042101167703, 'top2': 0.4194592007402292, 'top3': 0.41927779087623196}",0.7125637098332068,0.3922931931632452,0.009128083181487133,0.48187702682941513,0.21516510877757866,student,0.0,
+7,294,0.0,30.5,,,,,,,,,,,,0.7199579951007331,0.3878433725098941,0.01970009244399582,,,,,True

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-margin/budget-5/seed-3/log.txt

@@ -0,0 +1,106 @@
+[11:04:30][      0s][START] target=whole-mouse-brain variant=macs-v11-norestart-acq-margin budget=5.0% seed=3 | {"task": "membrane", "target": "whole-mouse-brain", "variant": "macs-v11-norestart-acq-margin", "budget": 5.0, "seed": 3, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "margin", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[11:04:30][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[11:04:30][      0s][DATA] whole-mouse-brain: train=5925 val=1250 test=1300 patches
+[11:04:31][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, snemi=3.55
+[11:04:31][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s
+[11:04:37][      7s][WEIGHTS] mixture: c-elegans=0.008, cerebellum=0.256, flywire=0.039, berghia=0.075, microns=0.376, oct-glia=0.134, oct-sfl=0.096, snemi=0.016 | eff.sources=5.13 | MMD2 mix=0.1295 uniform=0.1553
+[11:04:37][      7s][WEIGHTS] per-source: c-elegans=0.145, cerebellum=0.042, flywire=0.001, berghia=0.142, microns=0.020, oct-glia=0.402, oct-sfl=0.243, snemi=0.004 | min-MMD source (NeuroADDA rule) = oct-glia
+[11:05:01][     31s][TEACHERS] target-val Dice: c-elegans=0.386, cerebellum=0.316, flywire=0.110, berghia=0.259, microns=0.329, oct-glia=0.407, oct-sfl=0.440, snemi=0.305
+[11:05:01][     31s][TEACHERS] ensemble target-val Dice: mixture=0.385, mixture@raw=0.400, mixture@logit=0.380, per_source=0.420, per_source@raw=0.422, per_source@logit=0.419, uniform=0.413, uniform@raw=0.388, uniform@logit=0.405
+[11:05:01][     31s][WEIGHTS] using mixture: c-elegans=0.008, cerebellum=0.256, flywire=0.039, berghia=0.075, microns=0.376, oct-glia=0.134, oct-sfl=0.096, snemi=0.016
+[11:05:01][     31s][INIT] student initialised from teacher microns (weight 0.376)
+[11:05:03][     33s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[11:05:03][     33s][KD] distilling the mixture ensemble (prob space) into the student
+[11:06:10][    100s][KD] soft labels for 5925 patches from 8 teachers in 67s
+[11:07:03][    153s][KD] epoch 1/5 | patches=5925 | loss=0.6683 | 53s
+[11:07:55][    205s][KD] epoch 2/5 | patches=5925 | loss=0.4975 | 52s
+[11:08:47][    258s][KD] epoch 3/5 | patches=5925 | loss=0.4701 | 52s
+[11:09:39][    309s][KD] epoch 4/5 | patches=5925 | loss=0.4595 | 52s
+[11:10:31][    362s][KD] epoch 5/5 | patches=5925 | loss=0.4557 | 52s
+[11:10:33][    364s][ROUND0] before any target label: val_dice=0.4161 val_vi=0.5124
+[11:10:33][    364s][AL] patches=5925 (pool 5925 after dropping blank; free blank negatives: True) | budget=296 (5.0%) | rounds=6 | per_round=49 | acquisition=margin | first_round=stratified
+[11:10:43][    373s][AL] round-1 uncertainty quality on 5925 patches: {'spearman_uncertainty_error': -0.24881207404064967, 'auroc_top20_error': 0.3594799622567608}
+[11:11:06][    396s][AL] round 1: stratified cold start: 25 patches from the membrane-rich half (ensemble prevalence >= 0.215) + 24 from the membrane-poor half of 2000 candidates
+[11:11:08][    399s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.433, uniform=0.435, top1=0.471, top2=0.445, top3=0.446 -> using top1
+[11:11:09][    399s][WEIGHTS] round 1: student kept (labeled Dice 0.384 vs best teacher oct-sfl 0.471); distilling the corrected ensemble into it
+[11:11:09][    399s][WEIGHTS] round 1: teacher Dice on 49 labeled patches + 0 blank (640544 fg px): c-elegans=0.367, cerebellum=0.350, flywire=0.141, berghia=0.235, microns=0.310, oct-glia=0.385, oct-sfl=0.471, snemi=0.254 | weights: c-elegans=0.000, cerebellum=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=1.000, snemi=0.000
+[11:11:18][    408s][KD] soft labels for 5876 patches from 8 teachers in 9s
+[11:12:10][    460s][KD] epoch 1/2 | patches=5876 | loss=0.1740 | 52s
+[11:13:01][    512s][KD] epoch 2/2 | patches=5876 | loss=0.1640 | 52s
+[11:13:05][    515s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 3s
+[11:13:06][    516s][FT] round 1/6 epoch 1/5 | patches=49 | loss=1.3153
+[11:13:07][    517s][FT] round 1/6 epoch 2/5 | patches=49 | loss=1.0823
+[11:13:08][    519s][FT] round 1/6 epoch 3/5 | patches=49 | loss=1.2696
+[11:13:10][    520s][FT] round 1/6 epoch 4/5 | patches=49 | loss=1.0443
+[11:13:11][    521s][FT] round 1/6 epoch 5/5 | patches=49 | loss=1.0469
+[11:13:13][    523s][AL] round 1/6 | labeled=49 | selection_seconds=32.3000 round_seconds=157.5000 restart_student_dice=0.3843 restart_teacher_dice=0.4707 val_dice=0.6424 val_vi=0.4584 val_ece=0.0183
+[11:13:22][    532s][AL] round 2: 525/5876 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:13:27][    537s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.448, uniform=0.275, top1=0.479, top2=0.449, top3=0.449 -> using top1
+[11:13:27][    537s][WEIGHTS] round 2: teacher Dice on 98 labeled patches + 0 blank (1973085 fg px): c-elegans=0.479, cerebellum=0.447, flywire=0.164, berghia=0.213, microns=0.222, oct-glia=0.255, oct-sfl=0.330, snemi=0.432 | weights: c-elegans=1.000, cerebellum=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.000, snemi=0.000
+[11:13:29][    539s][CONS] round 2: on the 49 new patches student Dice=0.571 vs ensemble 0.563 -> pseudo-labels from the ensemble | consistency x1.00
+[11:13:31][    541s][FT] round 2/6 epoch 1/5 | patches=98 | loss=1.2983
+[11:13:33][    543s][FT] round 2/6 epoch 2/5 | patches=98 | loss=1.2472
+[11:13:35][    545s][FT] round 2/6 epoch 3/5 | patches=98 | loss=1.1352
+[11:13:37][    547s][FT] round 2/6 epoch 4/5 | patches=98 | loss=1.1293
+[11:13:39][    549s][FT] round 2/6 epoch 5/5 | patches=98 | loss=1.0466
+[11:13:41][    551s][AL] round 2/6 | labeled=98 | selection_seconds=9.0000 round_seconds=26.2000 pseudo_student_dice=0.5708 pseudo_ensemble_dice=0.5629 consistency_scale=1.0000 val_dice=0.6499 val_vi=0.4739 val_ece=0.0387
+[11:13:50][    560s][AL] round 3: 556/5827 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:13:57][    567s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.518, uniform=0.193, top1=0.518, top2=0.519, top3=0.519 -> using mixture
+[11:13:57][    567s][WEIGHTS] round 3: teacher Dice on 147 labeled patches + 0 blank (3447972 fg px): c-elegans=0.468, cerebellum=0.518, flywire=0.109, berghia=0.178, microns=0.156, oct-glia=0.188, oct-sfl=0.243, snemi=0.502 | weights: c-elegans=0.015, cerebellum=0.858, flywire=0.002, berghia=0.008, microns=0.034, oct-glia=0.017, oct-sfl=0.020, snemi=0.045
+[11:13:59][    569s][CONS] round 3: on the 49 new patches student Dice=0.369 vs ensemble 0.598 -> pseudo-labels from the ensemble | consistency x1.00
+[11:14:02][    572s][FT] round 3/6 epoch 1/5 | patches=147 | loss=1.5031
+[11:14:05][    575s][FT] round 3/6 epoch 2/5 | patches=147 | loss=1.4121
+[11:14:07][    578s][FT] round 3/6 epoch 3/5 | patches=147 | loss=1.2386
+[11:14:10][    581s][FT] round 3/6 epoch 4/5 | patches=147 | loss=1.1837
+[11:14:13][    584s][FT] round 3/6 epoch 5/5 | patches=147 | loss=1.1168
+[11:14:15][    586s][AL] round 3/6 | labeled=147 | selection_seconds=8.9000 round_seconds=32.2000 pseudo_student_dice=0.3691 pseudo_ensemble_dice=0.5979 consistency_scale=1.0000 val_dice=0.6700 val_vi=0.4160 val_ece=0.0603
+[11:14:24][    594s][AL] round 4: 811/5778 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:14:33][    603s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.507, uniform=0.168, top1=0.507, top2=0.508, top3=0.508 -> using mixture
+[11:14:33][    603s][WEIGHTS] round 4: teacher Dice on 196 labeled patches + 0 blank (4475353 fg px): c-elegans=0.452, cerebellum=0.507, flywire=0.093, berghia=0.175, microns=0.140, oct-glia=0.174, oct-sfl=0.225, snemi=0.490 | weights: c-elegans=0.015, cerebellum=0.861, flywire=0.002, berghia=0.009, microns=0.032, oct-glia=0.016, oct-sfl=0.019, snemi=0.045
+[11:14:38][    608s][CONS] round 4: on the 49 new patches student Dice=0.571 vs ensemble 0.466 -> pseudo-labels from the student | consistency x0.24
+[11:14:42][    612s][FT] round 4/6 epoch 1/5 | patches=196 | loss=1.1983
+[11:14:45][    616s][FT] round 4/6 epoch 2/5 | patches=196 | loss=1.1145
+[11:14:49][    620s][FT] round 4/6 epoch 3/5 | patches=196 | loss=1.0540
+[11:14:53][    623s][FT] round 4/6 epoch 4/5 | patches=196 | loss=0.9975
+[11:14:57][    627s][FT] round 4/6 epoch 5/5 | patches=196 | loss=1.0069
+[11:14:59][    629s][AL] round 4/6 | labeled=196 | selection_seconds=8.9000 round_seconds=41.5000 pseudo_student_dice=0.5709 pseudo_ensemble_dice=0.4664 consistency_scale=0.2360 val_dice=0.6300 val_vi=0.5393 val_ece=0.0166
+[11:15:07][    638s][AL] round 5: 374/5729 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:15:59][    690s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.452, cerebellum=0.459, flywire=0.132, berghia=0.303, microns=0.285, oct-glia=0.406, oct-sfl=0.497, snemi=0.405 | blended with labeled Dice at 0.50
+[11:16:48][    738s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.451, uniform=0.289, top1=0.446, top2=0.446, top3=0.447 -> using mixture
+[11:16:48][    738s][WEIGHTS] round 5: teacher Dice on 245 labeled patches + 0 blank (4632452 fg px): c-elegans=0.422, cerebellum=0.446, flywire=0.112, berghia=0.237, microns=0.214, oct-glia=0.292, oct-sfl=0.363, snemi=0.414 | weights: c-elegans=0.015, cerebellum=0.653, flywire=0.003, berghia=0.024, microns=0.094, oct-glia=0.074, oct-sfl=0.107, snemi=0.029
+[11:16:53][    743s][CONS] round 5: on the 49 new patches student Dice=0.157 vs ensemble 0.069 -> pseudo-labels from the student | consistency x0.00
+[11:16:55][    746s][FT] round 5/6 epoch 1/5 | patches=245 | loss=1.0619
+[11:16:58][    748s][FT] round 5/6 epoch 2/5 | patches=245 | loss=1.0047
+[11:17:00][    751s][FT] round 5/6 epoch 3/5 | patches=245 | loss=0.9691
+[11:17:03][    753s][FT] round 5/6 epoch 4/5 | patches=245 | loss=0.9047
+[11:17:05][    755s][FT] round 5/6 epoch 5/5 | patches=245 | loss=0.8771
+[11:17:07][    757s][AL] round 5/6 | labeled=245 | selection_seconds=8.8000 round_seconds=126.3000 pseudo_student_dice=0.1569 pseudo_ensemble_dice=0.0693 consistency_scale=0.0000 val_dice=0.6217 val_vi=0.5293 val_ece=0.0085
+[11:17:15][    766s][AL] round 6: 298/5680 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:18:08][    819s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.459, cerebellum=0.467, flywire=0.130, berghia=0.303, microns=0.278, oct-glia=0.400, oct-sfl=0.490, snemi=0.427 | blended with labeled Dice at 0.50
+[11:18:58][    869s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.443, uniform=0.283, top1=0.439, top2=0.439, top3=0.440 -> using mixture
+[11:18:58][    869s][WEIGHTS] round 6: teacher Dice on 294 labeled patches + 0 blank (5289236 fg px): c-elegans=0.420, cerebellum=0.439, flywire=0.109, berghia=0.254, microns=0.206, oct-glia=0.296, oct-sfl=0.361, snemi=0.417 | weights: c-elegans=0.016, cerebellum=0.639, flywire=0.004, berghia=0.030, microns=0.091, oct-glia=0.081, oct-sfl=0.109, snemi=0.032
+[11:19:03][    874s][CONS] round 6: on the 49 new patches student Dice=0.439 vs ensemble 0.169 -> pseudo-labels from the student | consistency x0.00
+[11:19:06][    876s][FT] round 6/6 epoch 1/5 | patches=294 | loss=1.1014
+[11:19:09][    879s][FT] round 6/6 epoch 2/5 | patches=294 | loss=1.0341
+[11:19:12][    882s][FT] round 6/6 epoch 3/5 | patches=294 | loss=0.9358
+[11:19:15][    885s][FT] round 6/6 epoch 4/5 | patches=294 | loss=0.9625
+[11:19:17][    888s][FT] round 6/6 epoch 5/5 | patches=294 | loss=0.8936
+[11:19:19][    890s][AL] round 6/6 | labeled=294 | selection_seconds=8.8000 round_seconds=130.4000 pseudo_student_dice=0.4388 pseudo_ensemble_dice=0.1688 consistency_scale=0.0000 val_dice=0.7115 val_vi=0.3985 val_ece=0.0027
+[11:19:22][    893s][CONS] final consolidation: pseudo-labels refreshed from the student
+[11:19:25][    896s][FT] final 1/10 epoch 1/1 | patches=294 | loss=0.9068
+[11:19:28][    898s][FT] final 2/10 epoch 1/1 | patches=294 | loss=0.9436
+[11:19:31][    901s][FT] final 3/10 epoch 1/1 | patches=294 | loss=0.8761
+[11:19:34][    904s][FT] final 4/10 epoch 1/1 | patches=294 | loss=0.9009
+[11:19:36][    907s][FT] final 5/10 epoch 1/1 | patches=294 | loss=0.7947
+[11:19:39][    910s][FT] final 6/10 epoch 1/1 | patches=294 | loss=0.7844
+[11:19:42][    913s][FT] final 7/10 epoch 1/1 | patches=294 | loss=0.7615
+[11:19:45][    915s][FT] final 8/10 epoch 1/1 | patches=294 | loss=0.7173
+[11:19:48][    918s][FT] final 9/10 epoch 1/1 | patches=294 | loss=0.7306
+[11:19:50][    921s][FT] final 10/10 epoch 1/1 | patches=294 | loss=0.7398
+[11:19:52][    923s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=31.1000 val_dice=0.7158 val_vi=0.3846 val_ece=0.0160
+[11:19:52][    923s][WEIGHTS] final: c-elegans=0.016, cerebellum=0.639, flywire=0.004, berghia=0.030, microns=0.091, oct-glia=0.081, oct-sfl=0.109, snemi=0.032
+[11:20:10][    941s][TEST] raw student (no EMA, no TTA): patch_dice=0.7032 patch_vi=0.3580
+[11:20:13][    943s][TEST] final model without TTA: patch_dice=0.7032 patch_vi=0.3580
+[11:20:13][    943s][RESOURCES] {"phases": {"setup": {"seconds": 1.13, "peak_cpu_rss_mb": 1027.7, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 31.55, "peak_cpu_rss_mb": 1809.5, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 328.93, "peak_cpu_rss_mb": 3677.4, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 559.03, "peak_cpu_rss_mb": 3677.4, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 20.19, "peak_cpu_rss_mb": 3677.4, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 1300, "images": 52, "patches_per_second": 64.4, "seconds_per_image": 0.388}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 3677.4, "total_seconds": 940.8}
+[11:20:13][    943s][TEST] patch_dice=0.7254 patch_iou=0.6456 patch_precision=0.7365 patch_recall=0.8469 patch_vi=0.3307 image_dice=0.8376 image_iou=0.7339 image_precision=0.8064 image_recall=0.8949 image_vi=0.3832 cal_ece=0.0033 cal_nll=0.1739 cal_brier=0.0503 cal_mean_confidence=0.9339 cal_accuracy=0.9333 cal_predicted_prevalence=0.2312 cal_true_prevalence=0.2108 cal_mean_probability=0.2435
+[11:20:13][    943s][DONE] whole-mouse-brain macs-v11-norestart-acq-margin b5 s3 | test patch Dice=0.7254 VI=0.3307 | image Dice=0.8376 VI=0.3832 | 943s

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-margin/budget-5/seed-3/result.json

@@ -0,0 +1,767 @@
+{
+  "target": "whole-mouse-brain",
+  "variant": "macs-v11-norestart-acq-margin",
+  "budget": 5.0,
+  "seed": 3,
+  "config": {
+    "task": "membrane",
+    "target": "whole-mouse-brain",
+    "variant": "macs-v11-norestart-acq-margin",
+    "budget": 5.0,
+    "seed": 3,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "margin",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "whole-mouse-brain",
+      "--variants",
+      "macs-v11-norestart-acq-margin",
+      "--budgets",
+      "5",
+      "--seeds",
+      "3",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "snemi3d"
+  ],
+  "alphas": [
+    0.007617762850063339,
+    0.2559312601209969,
+    0.03865993944852264,
+    0.07533345106002441,
+    0.3764169834180784,
+    0.13449516240019677,
+    0.09581340232315401,
+    0.0157320383789636
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "snemi3d"
+    ],
+    "mixture": [
+      0.007617762850063339,
+      0.2559312601209969,
+      0.03865993944852264,
+      0.07533345106002441,
+      0.3764169834180784,
+      0.13449516240019677,
+      0.09581340232315401,
+      0.0157320383789636
+    ],
+    "per_source": [
+      0.1450746039368088,
+      0.041950766450651496,
+      0.0008814721099573598,
+      0.1418325124471684,
+      0.02035312744213396,
+      0.4023730240336797,
+      0.24338555075976573,
+      0.004148942819834595
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.11311670518772092,
+      "mmd2_mixture": 0.12945823384265215,
+      "mmd2_uniform": 0.155288743679944,
+      "mmd2_single_source": [
+        0.38071639717042494,
+        0.17697894982207135,
+        0.3194061281637849,
+        0.24037642211575033,
+        0.16266554797152122,
+        0.2734526560532348,
+        0.2563986706812869,
+        0.2660414632502267
+      ],
+      "median_squared_distance": 0.18592532667757627,
+      "gammas": [
+        1.3446258477390725,
+        2.689251695478145,
+        5.37850339095629,
+        10.75700678191258,
+        21.51401356382516
+      ],
+      "max_weight": 0.3764169834180784,
+      "entropy": 1.6341528654931232,
+      "effective_sources": 5.125114496180783,
+      "min_weight": 0.007617762850063339
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.33220085501273156,
+        0.46442573060595516,
+        0.8760630244687474,
+        0.33460943356720163,
+        0.5415026072699861,
+        0.2234870921926152,
+        0.2770625113402161,
+        0.7109869456345189
+      ],
+      "standardized": [
+        -0.6467275403932012,
+        -0.02635183204683673,
+        1.9049774780863755,
+        -0.6354269162093369,
+        0.33527922824378936,
+        -1.1567932698101913,
+        -0.9054268876938927,
+        1.130469739823295
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4023730240336797,
+      "entropy": 1.508581788687784,
+      "effective_sources": 4.520315483702711,
+      "min_weight": 0.0008814721099573598
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.38573376502042933,
+      0.31643357804427197,
+      0.10997258971952562,
+      0.2588485480329741,
+      0.3288178379711084,
+      0.4072116167356654,
+      0.4397876396557443,
+      0.30535999081848075
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.3845088921696238,
+      "mixture@raw": 0.3998907914691259,
+      "mixture@logit": 0.3802050170706561,
+      "per_source": 0.4199022574247113,
+      "per_source@raw": 0.42186088103254527,
+      "per_source@logit": 0.41929713845028616,
+      "uniform": 0.4128319943503538,
+      "uniform@raw": 0.3877140229453584,
+      "uniform@logit": 0.4054972948683435
+    },
+    "val_patches": 1024
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    3.5461363792419434
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.3764169834180784,
+    "entropy": 1.6341528654931232,
+    "effective_sources": 5.125114496180783,
+    "min_weight": 0.007617762850063339
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.261904761904762,
+    "spearman_weight_taxonomy": 0.39440531887330776,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 328.9,
+  "round0_val": {
+    "val_dice": 0.41606177435746006,
+    "val_vi": 0.5123552235778059,
+    "val_ece": 0.14459192201571608
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 49,
+      "selection_seconds": 32.3,
+      "round_seconds": 157.5,
+      "teacher_dice_labeled": [
+        0.36738741011177045,
+        0.34973487377165097,
+        0.14100291682142935,
+        0.2353012898060949,
+        0.30996524321190877,
+        0.38452188964078277,
+        0.47067622607709797,
+        0.25385926187706875
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 640544,
+      "evidence_nonempty_patches": 39,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.3843282859342692,
+      "restart_teacher_dice": 0.47067622607709797,
+      "weights": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.43280781937383384,
+        "uniform": 0.43493152236966137,
+        "top1": 0.47067622607709797,
+        "top2": 0.44454578875106776,
+        "top3": 0.4462152205328611
+      },
+      "val_dice": 0.6423608970665755,
+      "val_vi": 0.4583910108960369,
+      "val_ece": 0.018296731206981286
+    },
+    {
+      "round": 2,
+      "n_labeled": 98,
+      "selection_seconds": 9.0,
+      "round_seconds": 26.2,
+      "teacher_dice_labeled": [
+        0.4790010720265582,
+        0.4471774808650101,
+        0.1635805331835569,
+        0.2126323076912074,
+        0.22214865194975508,
+        0.2548709404689054,
+        0.3297295554084256,
+        0.43233750559567696
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1973085,
+      "evidence_nonempty_patches": 84,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.44770498278126886,
+        "uniform": 0.2748470197475036,
+        "top1": 0.4790010720265582,
+        "top2": 0.4485533237171078,
+        "top3": 0.4488417346971152
+      },
+      "pseudo_student_dice": 0.5707606242464341,
+      "pseudo_ensemble_dice": 0.5629199461807312,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6498913510594475,
+      "val_vi": 0.4739385598038719,
+      "val_ece": 0.03870070742571032
+    },
+    {
+      "round": 3,
+      "n_labeled": 147,
+      "selection_seconds": 8.9,
+      "round_seconds": 32.2,
+      "teacher_dice_labeled": [
+        0.4675352679158566,
+        0.51830134874921,
+        0.10851799996363382,
+        0.17803503365781362,
+        0.1555896524596807,
+        0.18820785764113224,
+        0.24280354165733387,
+        0.5021665894135083
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 3447972,
+      "evidence_nonempty_patches": 121,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.015380486145659464,
+        0.8584998767475599,
+        0.0021538308841062364,
+        0.008410986483131318,
+        0.033577567429096726,
+        0.016624391153563504,
+        0.020444259760168546,
+        0.04490860139671429
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5183846978967368,
+        "uniform": 0.1932272503274666,
+        "top1": 0.51830134874921,
+        "top2": 0.5188958884279018,
+        "top3": 0.5190050811393596
+      },
+      "pseudo_student_dice": 0.3690770983734586,
+      "pseudo_ensemble_dice": 0.5979417700316343,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6700320030612286,
+      "val_vi": 0.41596008318263494,
+      "val_ece": 0.0603106358419856
+    },
+    {
+      "round": 4,
+      "n_labeled": 196,
+      "selection_seconds": 8.9,
+      "round_seconds": 41.5,
+      "teacher_dice_labeled": [
+        0.4521817725179459,
+        0.5070976087675007,
+        0.09348266121186623,
+        0.17504878863175116,
+        0.14020116045283942,
+        0.17422245803812333,
+        0.22538984408502558,
+        0.490394045937209
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 4475353,
+      "evidence_nonempty_patches": 163,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.014804203531187304,
+        0.8613454825173498,
+        0.0020797365351979775,
+        0.009161618543295417,
+        0.03230816007313526,
+        0.01622192158502868,
+        0.01927698149206271,
+        0.04480189572274281
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.506573562715877,
+        "uniform": 0.16827833119491115,
+        "top1": 0.5070976087675007,
+        "top2": 0.5079794458246457,
+        "top3": 0.5080628048620779
+      },
+      "pseudo_student_dice": 0.5709141779489735,
+      "pseudo_ensemble_dice": 0.4664405385589357,
+      "pseudo_source": "student",
+      "consistency_scale": 0.236,
+      "val_dice": 0.6299671169810852,
+      "val_vi": 0.5393323978378546,
+      "val_ece": 0.01660194871051421
+    },
+    {
+      "round": 5,
+      "n_labeled": 245,
+      "selection_seconds": 8.8,
+      "round_seconds": 126.3,
+      "teacher_dice_labeled": [
+        0.4222786728452023,
+        0.44601629440755297,
+        0.11182095835834197,
+        0.23742002382343114,
+        0.2138810167448923,
+        0.29239287155867766,
+        0.36293542540080537,
+        0.41378682372365055
+      ],
+      "teacher_dice_pool": [
+        0.45211503388520596,
+        0.4586268562508522,
+        0.13206520212610942,
+        0.3034451235507459,
+        0.2853314220634787,
+        0.4058655940805688,
+        0.4968728262174501,
+        0.40482826965176766
+      ],
+      "evidence_fg_pixels": 4632452,
+      "evidence_nonempty_patches": 192,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.015338916240729503,
+        0.6534039312404931,
+        0.003490821758896233,
+        0.02388494131779684,
+        0.09431416337331966,
+        0.07389023085416159,
+        0.10657836184751224,
+        0.02909863336709079
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.450717108090974,
+        "uniform": 0.28857802117043396,
+        "top1": 0.44601629440755297,
+        "top2": 0.44629305482704884,
+        "top3": 0.44681775473389657
+      },
+      "pseudo_student_dice": 0.15691007580540497,
+      "pseudo_ensemble_dice": 0.06933827105338418,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.6217327713376608,
+      "val_vi": 0.5292694280761847,
+      "val_ece": 0.008471389509414293
+    },
+    {
+      "round": 6,
+      "n_labeled": 294,
+      "selection_seconds": 8.8,
+      "round_seconds": 130.4,
+      "teacher_dice_labeled": [
+        0.4204763259856441,
+        0.4391522114337465,
+        0.10897358309180741,
+        0.2540118077271043,
+        0.20582054260607813,
+        0.296415072255611,
+        0.36062790679412976,
+        0.4173540295143803
+      ],
+      "teacher_dice_pool": [
+        0.4590040784121988,
+        0.4673811142986949,
+        0.12976248045272956,
+        0.3033754435936046,
+        0.2781588715045815,
+        0.40047257679930554,
+        0.490289506131676,
+        0.42686846374501874
+      ],
+      "evidence_fg_pixels": 5289236,
+      "evidence_nonempty_patches": 229,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.01577566582337092,
+        0.6388397404019729,
+        0.003552893550954598,
+        0.02952581508890634,
+        0.09111528485345537,
+        0.08055188987539245,
+        0.10906066204483804,
+        0.03157804836110938
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.44275603813402087,
+        "uniform": 0.28250357953908506,
+        "top1": 0.4391522114337465,
+        "top2": 0.4394290706035419,
+        "top3": 0.4402034746697075
+      },
+      "pseudo_student_dice": 0.43881066214162834,
+      "pseudo_ensemble_dice": 0.16880114668182505,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.7114711579891998,
+      "val_vi": 0.3985485231582955,
+      "val_ece": 0.002656605495560469
+    },
+    {
+      "round": 7,
+      "n_labeled": 294,
+      "selection_seconds": 0.0,
+      "round_seconds": 31.1,
+      "final_consolidation": true,
+      "val_dice": 0.7158009490992879,
+      "val_vi": 0.38461948070584473,
+      "val_ece": 0.016036664127452485
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.24881207404064967,
+    "auroc_top20_error": 0.3594799622567608
+  },
+  "al_seconds": 559.0,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.01577566582337092,
+    0.6388397404019729,
+    0.003552893550954598,
+    0.02952581508890634,
+    0.09111528485345537,
+    0.08055188987539245,
+    0.10906066204483804,
+    0.03157804836110938
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.6388397404019729,
+    "entropy": 1.24771609642705,
+    "effective_sources": 3.4823804469747293,
+    "min_weight": 0.003552893550954598
+  },
+  "top_source_final": "dhanyasi-P14-mouse-cerebellum",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7031837305499606,
+    "patch_iou": 0.6186873385247618,
+    "patch_precision": 0.7015658527093758,
+    "patch_recall": 0.8501698170797606,
+    "patch_vi": 0.35798487868286893,
+    "n_patches": 1300,
+    "cal_ece": 0.014296868795295074,
+    "cal_nll": 0.19317754581198462,
+    "cal_brier": 0.055458809854427006,
+    "cal_mean_confidence": 0.9406963579779875,
+    "cal_accuracy": 0.9263994891826923,
+    "cal_predicted_prevalence": 0.2381670673076923,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.2453708863664394
+  },
+  "test_notta": {
+    "patch_dice": 0.7031837305499606,
+    "patch_iou": 0.6186873385247618,
+    "patch_precision": 0.7015658527093758,
+    "patch_recall": 0.8501698170797606,
+    "patch_vi": 0.35798487868286893,
+    "n_patches": 1300,
+    "cal_ece": 0.014296868795295074,
+    "cal_nll": 0.19317754581198462,
+    "cal_brier": 0.055458809854427006,
+    "cal_mean_confidence": 0.9406963579779875,
+    "cal_accuracy": 0.9263994891826923,
+    "cal_predicted_prevalence": 0.2381670673076923,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.2453708863664394
+  },
+  "test": {
+    "patch_dice": 0.7253981718195581,
+    "patch_iou": 0.6455533937479556,
+    "patch_precision": 0.7364661124292978,
+    "patch_recall": 0.8468879653434368,
+    "patch_vi": 0.33066736927078855,
+    "n_patches": 1300,
+    "image_dice": 0.8376132017999095,
+    "image_iou": 0.7338580259277356,
+    "image_precision": 0.8064019432082945,
+    "image_recall": 0.8948911544055365,
+    "image_vi": 0.3832235482166312,
+    "n_images": 52,
+    "cal_ece": 0.0032693595258011564,
+    "cal_nll": 0.17391863248092237,
+    "cal_brier": 0.05027658671011288,
+    "cal_mean_confidence": 0.9338601700799212,
+    "cal_accuracy": 0.9332977764423077,
+    "cal_predicted_prevalence": 0.23124924879807693,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.2435347047852322
+  },
+  "eval_seconds": 20.2,
+  "total_seconds": 943.0,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.13,
+        "peak_cpu_rss_mb": 1027.7,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 31.55,
+        "peak_cpu_rss_mb": 1809.5,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 328.93,
+        "peak_cpu_rss_mb": 3677.4,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4638.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 559.03,
+        "peak_cpu_rss_mb": 3677.4,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4646.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 20.19,
+        "peak_cpu_rss_mb": 3677.4,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4646.0,
+        "patches": 1300,
+        "images": 52,
+        "patches_per_second": 64.4,
+        "seconds_per_image": 0.388
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 3677.4,
+    "total_seconds": 940.8
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/whole-mouse-brain/macs-v11-norestart-acq-margin_b5_s3.pth"
+}

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-margin/budget-5/seed-3/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,49,32.3,157.5,"[0.36738741011177045, 0.34973487377165097, 0.14100291682142935, 0.2353012898060949, 0.30996524321190877, 0.38452188964078277, 0.47067622607709797, 0.25385926187706875]",,640544,39,True,kept,0.3843282859342692,0.47067622607709797,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.43280781937383384, 'uniform': 0.43493152236966137, 'top1': 0.47067622607709797, 'top2': 0.44454578875106776, 'top3': 0.4462152205328611}",0.6423608970665755,0.4583910108960369,0.018296731206981286,,,,,
+2,98,9.0,26.2,"[0.4790010720265582, 0.4471774808650101, 0.1635805331835569, 0.2126323076912074, 0.22214865194975508, 0.2548709404689054, 0.3297295554084256, 0.43233750559567696]",,1973085,84,False,,,,"[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.44770498278126886, 'uniform': 0.2748470197475036, 'top1': 0.4790010720265582, 'top2': 0.4485533237171078, 'top3': 0.4488417346971152}",0.6498913510594475,0.4739385598038719,0.03870070742571032,0.5707606242464341,0.5629199461807312,ensemble,1.0,
+3,147,8.9,32.2,"[0.4675352679158566, 0.51830134874921, 0.10851799996363382, 0.17803503365781362, 0.1555896524596807, 0.18820785764113224, 0.24280354165733387, 0.5021665894135083]",,3447972,121,False,,,,"[0.015380486145659464, 0.8584998767475599, 0.0021538308841062364, 0.008410986483131318, 0.033577567429096726, 0.016624391153563504, 0.020444259760168546, 0.04490860139671429]",mixture,"{'mixture': 0.5183846978967368, 'uniform': 0.1932272503274666, 'top1': 0.51830134874921, 'top2': 0.5188958884279018, 'top3': 0.5190050811393596}",0.6700320030612286,0.41596008318263494,0.0603106358419856,0.3690770983734586,0.5979417700316343,ensemble,1.0,
+4,196,8.9,41.5,"[0.4521817725179459, 0.5070976087675007, 0.09348266121186623, 0.17504878863175116, 0.14020116045283942, 0.17422245803812333, 0.22538984408502558, 0.490394045937209]",,4475353,163,False,,,,"[0.014804203531187304, 0.8613454825173498, 0.0020797365351979775, 0.009161618543295417, 0.03230816007313526, 0.01622192158502868, 0.01927698149206271, 0.04480189572274281]",mixture,"{'mixture': 0.506573562715877, 'uniform': 0.16827833119491115, 'top1': 0.5070976087675007, 'top2': 0.5079794458246457, 'top3': 0.5080628048620779}",0.6299671169810852,0.5393323978378546,0.01660194871051421,0.5709141779489735,0.4664405385589357,student,0.236,
+5,245,8.8,126.3,"[0.4222786728452023, 0.44601629440755297, 0.11182095835834197, 0.23742002382343114, 0.2138810167448923, 0.29239287155867766, 0.36293542540080537, 0.41378682372365055]","[0.45211503388520596, 0.4586268562508522, 0.13206520212610942, 0.3034451235507459, 0.2853314220634787, 0.4058655940805688, 0.4968728262174501, 0.40482826965176766]",4632452,192,False,,,,"[0.015338916240729503, 0.6534039312404931, 0.003490821758896233, 0.02388494131779684, 0.09431416337331966, 0.07389023085416159, 0.10657836184751224, 0.02909863336709079]",mixture,"{'mixture': 0.450717108090974, 'uniform': 0.28857802117043396, 'top1': 0.44601629440755297, 'top2': 0.44629305482704884, 'top3': 0.44681775473389657}",0.6217327713376608,0.5292694280761847,0.008471389509414293,0.15691007580540497,0.06933827105338418,student,0.0,
+6,294,8.8,130.4,"[0.4204763259856441, 0.4391522114337465, 0.10897358309180741, 0.2540118077271043, 0.20582054260607813, 0.296415072255611, 0.36062790679412976, 0.4173540295143803]","[0.4590040784121988, 0.4673811142986949, 0.12976248045272956, 0.3033754435936046, 0.2781588715045815, 0.40047257679930554, 0.490289506131676, 0.42686846374501874]",5289236,229,False,,,,"[0.01577566582337092, 0.6388397404019729, 0.003552893550954598, 0.02952581508890634, 0.09111528485345537, 0.08055188987539245, 0.10906066204483804, 0.03157804836110938]",mixture,"{'mixture': 0.44275603813402087, 'uniform': 0.28250357953908506, 'top1': 0.4391522114337465, 'top2': 0.4394290706035419, 'top3': 0.4402034746697075}",0.7114711579891998,0.3985485231582955,0.002656605495560469,0.43881066214162834,0.16880114668182505,student,0.0,
+7,294,0.0,31.1,,,,,,,,,,,,0.7158009490992879,0.38461948070584473,0.016036664127452485,,,,,True

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/log.txt

@@ -0,0 +1,108 @@
+[05:58:12][      0s][START] target=whole-mouse-brain variant=macs-v11-norestart-acq-mcdropout budget=5.0% seed=2 | {"task": "membrane", "target": "whole-mouse-brain", "variant": "macs-v11-norestart-acq-mcdropout", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "mcdropout", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[05:58:12][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[05:58:12][      0s][DATA] whole-mouse-brain: train=5925 val=1250 test=1300 patches
+[05:58:14][      1s][CALIB] source-val temperatures: c-elegans=2.58, cerebellum=4.56, flywire=1.99, berghia=8.00, microns=1.92, oct-glia=2.55, oct-sfl=2.64, snemi=3.55
+[05:58:14][      2s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 1s
+[05:58:20][      8s][WEIGHTS] mixture: c-elegans=0.008, cerebellum=0.256, flywire=0.039, berghia=0.075, microns=0.376, oct-glia=0.134, oct-sfl=0.096, snemi=0.016 | eff.sources=5.13 | MMD2 mix=0.1295 uniform=0.1553
+[05:58:20][      8s][WEIGHTS] per-source: c-elegans=0.145, cerebellum=0.042, flywire=0.001, berghia=0.142, microns=0.020, oct-glia=0.402, oct-sfl=0.243, snemi=0.004 | min-MMD source (NeuroADDA rule) = oct-glia
+[05:58:44][     32s][TEACHERS] target-val Dice: c-elegans=0.386, cerebellum=0.316, flywire=0.110, berghia=0.259, microns=0.329, oct-glia=0.407, oct-sfl=0.440, snemi=0.305
+[05:58:44][     32s][TEACHERS] ensemble target-val Dice: mixture=0.385, mixture@raw=0.400, mixture@logit=0.380, per_source=0.420, per_source@raw=0.422, per_source@logit=0.419, uniform=0.413, uniform@raw=0.388, uniform@logit=0.405
+[05:58:44][     32s][WEIGHTS] using mixture: c-elegans=0.008, cerebellum=0.256, flywire=0.039, berghia=0.075, microns=0.376, oct-glia=0.134, oct-sfl=0.096, snemi=0.016
+[05:58:44][     32s][INIT] student initialised from teacher microns (weight 0.376)
+[05:58:45][     33s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[05:58:45][     33s][KD] distilling the mixture ensemble (prob space) into the student
+[05:59:53][    100s][KD] soft labels for 5925 patches from 8 teachers in 67s
+[06:00:45][    153s][KD] epoch 1/5 | patches=5925 | loss=0.6663 | 52s
+[06:01:38][    205s][KD] epoch 2/5 | patches=5925 | loss=0.4936 | 52s
+[06:02:30][    258s][KD] epoch 3/5 | patches=5925 | loss=0.4654 | 52s
+[06:03:21][    309s][KD] epoch 4/5 | patches=5925 | loss=0.4610 | 51s
+[06:04:13][    360s][KD] epoch 5/5 | patches=5925 | loss=0.4566 | 52s
+[06:04:15][    362s][ROUND0] before any target label: val_dice=0.3948 val_vi=0.5092
+[06:04:15][    362s][AL] patches=5925 (pool 5925 after dropping blank; free blank negatives: True) | budget=296 (5.0%) | rounds=6 | per_round=49 | acquisition=mcdropout | first_round=stratified
+[06:04:26][    373s][AL] round-1 uncertainty quality on 5925 patches: {'spearman_uncertainty_error': -0.20304753183898672, 'auroc_top20_error': 0.41592640068365117}
+[06:04:49][    396s][AL] round 1: stratified cold start: 25 patches from the membrane-rich half (ensemble prevalence >= 0.213) + 24 from the membrane-poor half of 2000 candidates
+[06:04:51][    399s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.417, uniform=0.396, top1=0.453, top2=0.456, top3=0.424 -> using top2
+[06:04:52][    399s][WEIGHTS] round 1: student kept (labeled Dice 0.302 vs best teacher oct-sfl 0.453); distilling the corrected ensemble into it
+[06:04:52][    399s][WEIGHTS] round 1: teacher Dice on 49 labeled patches + 0 blank (745730 fg px): c-elegans=0.417, cerebellum=0.359, flywire=0.153, berghia=0.310, microns=0.248, oct-glia=0.404, oct-sfl=0.453, snemi=0.315 | weights: c-elegans=0.052, cerebellum=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.948, snemi=0.000
+[06:05:09][    417s][KD] soft labels for 5876 patches from 8 teachers in 17s
+[06:06:00][    468s][KD] epoch 1/2 | patches=5876 | loss=0.2661 | 51s
+[06:06:52][    520s][KD] epoch 2/2 | patches=5876 | loss=0.2587 | 52s
+[06:06:58][    525s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 6s
+[06:06:59][    527s][FT] round 1/6 epoch 1/5 | patches=49 | loss=1.4576
+[06:07:00][    528s][FT] round 1/6 epoch 2/5 | patches=49 | loss=1.2015
+[06:07:02][    529s][FT] round 1/6 epoch 3/5 | patches=49 | loss=1.1189
+[06:07:03][    531s][FT] round 1/6 epoch 4/5 | patches=49 | loss=0.9829
+[06:07:04][    532s][FT] round 1/6 epoch 5/5 | patches=49 | loss=1.0922
+[06:07:06][    534s][AL] round 1/6 | labeled=49 | selection_seconds=34.0000 round_seconds=169.4000 restart_student_dice=0.3018 restart_teacher_dice=0.4531 val_dice=0.6500 val_vi=0.4698 val_ece=0.0157
+[06:07:17][    545s][AL] round 2: 477/5876 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[06:07:22][    549s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.411, uniform=0.230, top1=0.414, top2=0.414, top3=0.414 -> using mixture
+[06:07:22][    549s][WEIGHTS] round 2: teacher Dice on 98 labeled patches + 0 blank (1552702 fg px): c-elegans=0.366, cerebellum=0.414, flywire=0.094, berghia=0.242, microns=0.154, oct-glia=0.249, oct-sfl=0.298, snemi=0.355 | weights: c-elegans=0.013, cerebellum=0.695, flywire=0.004, berghia=0.037, microns=0.076, oct-glia=0.070, oct-sfl=0.082, snemi=0.024
+[06:07:27][    555s][CONS] round 2: on the 49 new patches student Dice=0.533 vs ensemble 0.450 -> pseudo-labels from the student | consistency x0.11
+[06:07:29][    557s][FT] round 2/6 epoch 1/5 | patches=98 | loss=1.2128
+[06:07:31][    559s][FT] round 2/6 epoch 2/5 | patches=98 | loss=1.1617
+[06:07:33][    561s][FT] round 2/6 epoch 3/5 | patches=98 | loss=1.1708
+[06:07:35][    563s][FT] round 2/6 epoch 4/5 | patches=98 | loss=1.1620
+[06:07:37][    565s][FT] round 2/6 epoch 5/5 | patches=98 | loss=1.0921
+[06:07:39][    567s][AL] round 2/6 | labeled=98 | selection_seconds=10.9000 round_seconds=31.3000 pseudo_student_dice=0.5334 pseudo_ensemble_dice=0.4504 consistency_scale=0.1110 val_dice=0.6392 val_vi=0.5097 val_ece=0.0480
+[06:07:50][    578s][AL] round 3: 542/5827 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[06:08:40][    628s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.425, cerebellum=0.432, flywire=0.134, berghia=0.279, microns=0.286, oct-glia=0.402, oct-sfl=0.503, snemi=0.374 | blended with labeled Dice at 0.50
+[06:09:26][    674s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.406, uniform=0.302, top1=0.385, top2=0.401, top3=0.401 -> using mixture
+[06:09:26][    674s][WEIGHTS] round 3: teacher Dice on 147 labeled patches + 0 blank (2045013 fg px): c-elegans=0.368, cerebellum=0.385, flywire=0.104, berghia=0.247, microns=0.216, oct-glia=0.308, oct-sfl=0.377, snemi=0.350 | weights: c-elegans=0.012, cerebellum=0.497, flywire=0.005, berghia=0.037, microns=0.135, oct-glia=0.121, oct-sfl=0.171, snemi=0.021
+[06:09:31][    679s][CONS] round 3: on the 49 new patches student Dice=0.321 vs ensemble 0.316 -> pseudo-labels from the student | consistency x0.00
+[06:09:33][    681s][FT] round 3/6 epoch 1/5 | patches=147 | loss=1.1314
+[06:09:34][    682s][FT] round 3/6 epoch 2/5 | patches=147 | loss=1.1234
+[06:09:36][    684s][FT] round 3/6 epoch 3/5 | patches=147 | loss=1.0685
+[06:09:37][    685s][FT] round 3/6 epoch 4/5 | patches=147 | loss=1.0192
+[06:09:39][    687s][FT] round 3/6 epoch 5/5 | patches=147 | loss=1.0596
+[06:09:41][    688s][AL] round 3/6 | labeled=147 | selection_seconds=10.8000 round_seconds=119.4000 pseudo_student_dice=0.3206 pseudo_ensemble_dice=0.3164 consistency_scale=0.0000 val_dice=0.6527 val_vi=0.4833 val_ece=0.0120
+[06:09:51][    699s][AL] round 4: 486/5778 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[06:10:42][    750s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.429, cerebellum=0.420, flywire=0.147, berghia=0.276, microns=0.315, oct-glia=0.427, oct-sfl=0.522, snemi=0.351 | blended with labeled Dice at 0.50
+[06:11:30][    798s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.375, uniform=0.314, top1=0.387, top2=0.390, top3=0.394 -> using top3
+[06:11:30][    798s][WEIGHTS] round 4: teacher Dice on 196 labeled patches + 0 blank (2546814 fg px): c-elegans=0.367, cerebellum=0.364, flywire=0.108, berghia=0.253, microns=0.226, oct-glia=0.320, oct-sfl=0.387, snemi=0.331 | weights: c-elegans=0.020, cerebellum=0.665, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.315, snemi=0.000
+[06:11:35][    803s][CONS] round 4: on the 49 new patches student Dice=0.372 vs ensemble 0.243 -> pseudo-labels from the student | consistency x0.00
+[06:11:37][    805s][FT] round 4/6 epoch 1/5 | patches=196 | loss=1.1221
+[06:11:39][    806s][FT] round 4/6 epoch 2/5 | patches=196 | loss=1.1144
+[06:11:41][    808s][FT] round 4/6 epoch 3/5 | patches=196 | loss=1.0672
+[06:11:43][    810s][FT] round 4/6 epoch 4/5 | patches=196 | loss=1.0260
+[06:11:44][    812s][FT] round 4/6 epoch 5/5 | patches=196 | loss=1.0100
+[06:11:46][    814s][AL] round 4/6 | labeled=196 | selection_seconds=10.7000 round_seconds=123.8000 pseudo_student_dice=0.3715 pseudo_ensemble_dice=0.2432 consistency_scale=0.0000 val_dice=0.7141 val_vi=0.3944 val_ece=0.0050
+[06:11:57][    824s][AL] round 5: 847/5729 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[06:12:49][    877s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.375, cerebellum=0.383, flywire=0.160, berghia=0.253, microns=0.320, oct-glia=0.437, oct-sfl=0.518, snemi=0.281 | blended with labeled Dice at 0.50
+[06:13:38][    925s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.360, uniform=0.321, top1=0.384, top2=0.386, top3=0.384 -> using top3
+[06:13:38][    925s][WEIGHTS] round 5: teacher Dice on 245 labeled patches + 0 blank (3577893 fg px): c-elegans=0.350, cerebellum=0.348, flywire=0.116, berghia=0.262, microns=0.226, oct-glia=0.332, oct-sfl=0.384, snemi=0.313 | weights: c-elegans=0.019, cerebellum=0.639, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.341, snemi=0.000
+[06:13:43][    930s][CONS] round 5: on the 49 new patches student Dice=0.588 vs ensemble 0.292 -> pseudo-labels from the student | consistency x0.29
+[06:13:47][    935s][FT] round 5/6 epoch 1/5 | patches=245 | loss=1.0574
+[06:13:52][    939s][FT] round 5/6 epoch 2/5 | patches=245 | loss=1.0164
+[06:13:56][    944s][FT] round 5/6 epoch 3/5 | patches=245 | loss=0.9954
+[06:14:01][    948s][FT] round 5/6 epoch 4/5 | patches=245 | loss=0.9892
+[06:14:05][    953s][FT] round 5/6 epoch 5/5 | patches=245 | loss=0.9721
+[06:14:07][    955s][AL] round 5/6 | labeled=245 | selection_seconds=10.6000 round_seconds=138.8000 pseudo_student_dice=0.5885 pseudo_ensemble_dice=0.2919 consistency_scale=0.2950 val_dice=0.7095 val_vi=0.3976 val_ece=0.0023
+[06:14:17][    965s][AL] round 6: 650/5680 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[06:15:11][   1018s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.375, cerebellum=0.365, flywire=0.158, berghia=0.262, microns=0.330, oct-glia=0.464, oct-sfl=0.533, snemi=0.281 | blended with labeled Dice at 0.50
+[06:16:00][   1068s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.365, uniform=0.322, top1=0.385, top2=0.387, top3=0.392 -> using top3
+[06:16:00][   1068s][WEIGHTS] round 6: teacher Dice on 294 labeled patches + 0 blank (4756463 fg px): c-elegans=0.365, cerebellum=0.359, flywire=0.110, berghia=0.258, microns=0.223, oct-glia=0.340, oct-sfl=0.385, snemi=0.331 | weights: c-elegans=0.021, cerebellum=0.659, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.320, snemi=0.000
+[06:16:06][   1073s][CONS] round 6: on the 49 new patches student Dice=0.540 vs ensemble 0.578 -> pseudo-labels from the student | consistency x0.13
+[06:16:11][   1079s][FT] round 6/6 epoch 1/5 | patches=294 | loss=1.0482
+[06:16:16][   1084s][FT] round 6/6 epoch 2/5 | patches=294 | loss=1.0003
+[06:16:22][   1089s][FT] round 6/6 epoch 3/5 | patches=294 | loss=0.9577
+[06:16:27][   1095s][FT] round 6/6 epoch 4/5 | patches=294 | loss=0.9268
+[06:16:32][   1100s][FT] round 6/6 epoch 5/5 | patches=294 | loss=0.8717
+[06:16:34][   1102s][AL] round 6/6 | labeled=294 | selection_seconds=10.6000 round_seconds=145.1000 pseudo_student_dice=0.5404 pseudo_ensemble_dice=0.5780 consistency_scale=0.1350 val_dice=0.7143 val_vi=0.4083 val_ece=0.0072
+[06:16:37][   1105s][CONS] final consolidation: pseudo-labels refreshed from the student
+[06:16:43][   1110s][FT] final 1/10 epoch 1/1 | patches=294 | loss=0.8734
+[06:16:48][   1116s][FT] final 2/10 epoch 1/1 | patches=294 | loss=0.9310
+[06:16:53][   1121s][FT] final 3/10 epoch 1/1 | patches=294 | loss=0.8816
+[06:16:59][   1126s][FT] final 4/10 epoch 1/1 | patches=294 | loss=0.8120
+[06:17:04][   1132s][FT] final 5/10 epoch 1/1 | patches=294 | loss=0.7529
+[06:17:09][   1137s][FT] final 6/10 epoch 1/1 | patches=294 | loss=0.7757
+[06:17:15][   1142s][FT] final 7/10 epoch 1/1 | patches=294 | loss=0.7468
+[06:17:20][   1148s][FT] final 8/10 epoch 1/1 | patches=294 | loss=0.7143
+[06:17:25][   1153s][FT] final 9/10 epoch 1/1 | patches=294 | loss=0.6793
+[06:17:31][   1158s][FT] final 10/10 epoch 1/1 | patches=294 | loss=0.7078
+[06:17:33][   1161s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=56.6000 val_dice=0.7309 val_vi=0.3714 val_ece=0.0158
+[06:17:33][   1161s][WEIGHTS] final: c-elegans=0.021, cerebellum=0.659, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.320, snemi=0.000
+[06:17:51][   1178s][TEST] raw student (no EMA, no TTA): patch_dice=0.7117 patch_vi=0.3410
+[06:17:53][   1181s][TEST] final model without TTA: patch_dice=0.7117 patch_vi=0.3410
+[06:17:53][   1181s][RESOURCES] {"phases": {"setup": {"seconds": 1.15, "peak_cpu_rss_mb": 1027.9, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 31.61, "peak_cpu_rss_mb": 1801.2, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 327.55, "peak_cpu_rss_mb": 3676.1, "peak_gpu_alloc_mb": 4303.7, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 798.09, "peak_cpu_rss_mb": 3676.1, "peak_gpu_alloc_mb": 4424.0, "peak_gpu_reserved_mb": 4648.0, "rounds": 6}, "inference": {"seconds": 20.38, "peak_cpu_rss_mb": 3676.1, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4648.0, "patches": 1300, "images": 52, "patches_per_second": 63.8, "seconds_per_image": 0.392}}, "peak_gpu_alloc_mb": 4424.0, "peak_cpu_rss_mb": 3676.1, "total_seconds": 1178.8}
+[06:17:53][   1181s][TEST] patch_dice=0.7329 patch_iou=0.6539 patch_precision=0.7795 patch_recall=0.8102 patch_vi=0.3151 image_dice=0.8398 image_iou=0.7371 image_precision=0.8421 image_recall=0.8598 image_vi=0.3651 cal_ece=0.0033 cal_nll=0.1590 cal_brier=0.0466 cal_mean_confidence=0.9397 cal_accuracy=0.9366 cal_predicted_prevalence=0.2156 cal_true_prevalence=0.2108 cal_mean_probability=0.2234
+[06:17:53][   1181s][DONE] whole-mouse-brain macs-v11-norestart-acq-mcdropout b5 s2 | test patch Dice=0.7329 VI=0.3151 | image Dice=0.8398 VI=0.3651 | 1181s

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/result.json

@@ -0,0 +1,785 @@
+{
+  "target": "whole-mouse-brain",
+  "variant": "macs-v11-norestart-acq-mcdropout",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "membrane",
+    "target": "whole-mouse-brain",
+    "variant": "macs-v11-norestart-acq-mcdropout",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "mcdropout",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "whole-mouse-brain",
+      "--variants",
+      "macs-v11-norestart-acq-mcdropout",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "c-elegans-dauer-stage",
+    "dhanyasi-P14-mouse-cerebellum",
+    "flywire",
+    "katz-lab-berghia-connective",
+    "microns_basil",
+    "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "octopus-vulgaris-vertical-lobe-sfltract",
+    "snemi3d"
+  ],
+  "alphas": [
+    0.007617762850063339,
+    0.2559312601209969,
+    0.03865993944852264,
+    0.07533345106002441,
+    0.3764169834180784,
+    0.13449516240019677,
+    0.09581340232315401,
+    0.0157320383789636
+  ],
+  "weights_all": {
+    "sources": [
+      "c-elegans-dauer-stage",
+      "dhanyasi-P14-mouse-cerebellum",
+      "flywire",
+      "katz-lab-berghia-connective",
+      "microns_basil",
+      "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+      "octopus-vulgaris-vertical-lobe-sfltract",
+      "snemi3d"
+    ],
+    "mixture": [
+      0.007617762850063339,
+      0.2559312601209969,
+      0.03865993944852264,
+      0.07533345106002441,
+      0.3764169834180784,
+      0.13449516240019677,
+      0.09581340232315401,
+      0.0157320383789636
+    ],
+    "per_source": [
+      0.1450746039368088,
+      0.041950766450651496,
+      0.0008814721099573598,
+      0.1418325124471684,
+      0.02035312744213396,
+      0.4023730240336797,
+      0.24338555075976573,
+      0.004148942819834595
+    ],
+    "uniform": [
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125,
+      0.125
+    ],
+    "mixture_info": {
+      "objective": 0.11311670518772092,
+      "mmd2_mixture": 0.12945823384265215,
+      "mmd2_uniform": 0.155288743679944,
+      "mmd2_single_source": [
+        0.38071639717042494,
+        0.17697894982207135,
+        0.3194061281637849,
+        0.24037642211575033,
+        0.16266554797152122,
+        0.2734526560532348,
+        0.2563986706812869,
+        0.2660414632502267
+      ],
+      "median_squared_distance": 0.18592532667757627,
+      "gammas": [
+        1.3446258477390725,
+        2.689251695478145,
+        5.37850339095629,
+        10.75700678191258,
+        21.51401356382516
+      ],
+      "max_weight": 0.3764169834180784,
+      "entropy": 1.6341528654931232,
+      "effective_sources": 5.125114496180783,
+      "min_weight": 0.007617762850063339
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.33220085501273156,
+        0.46442573060595516,
+        0.8760630244687474,
+        0.33460943356720163,
+        0.5415026072699861,
+        0.2234870921926152,
+        0.2770625113402161,
+        0.7109869456345189
+      ],
+      "standardized": [
+        -0.6467275403932012,
+        -0.02635183204683673,
+        1.9049774780863755,
+        -0.6354269162093369,
+        0.33527922824378936,
+        -1.1567932698101913,
+        -0.9054268876938927,
+        1.130469739823295
+      ],
+      "tau": 0.5,
+      "max_weight": 0.4023730240336797,
+      "entropy": 1.508581788687784,
+      "effective_sources": 4.520315483702711,
+      "min_weight": 0.0008814721099573598
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil",
+    "seconds": 6.2
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.38573376502042933,
+      0.31643357804427197,
+      0.10997258971952562,
+      0.2588485480329741,
+      0.3288178379711084,
+      0.4072116167356654,
+      0.4397876396557443,
+      0.30535999081848075
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.3845088921696238,
+      "mixture@raw": 0.3998907914691259,
+      "mixture@logit": 0.3802050170706561,
+      "per_source": 0.4199022574247113,
+      "per_source@raw": 0.42186088103254527,
+      "per_source@logit": 0.41929713845028616,
+      "uniform": 0.4128319943503538,
+      "uniform@raw": 0.3877140229453584,
+      "uniform@logit": 0.4054972948683435
+    },
+    "val_patches": 1024
+  },
+  "temperatures": [
+    2.575807571411133,
+    4.557675361633301,
+    1.990083932876587,
+    8.0,
+    1.9150984287261963,
+    2.5513463020324707,
+    2.6421806812286377,
+    3.5461363792419434
+  ],
+  "top_source": "microns_basil",
+  "weight_stats": {
+    "max_weight": 0.3764169834180784,
+    "entropy": 1.6341528654931232,
+    "effective_sources": 5.125114496180783,
+    "min_weight": 0.007617762850063339
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.261904761904762,
+    "spearman_weight_taxonomy": 0.39440531887330776,
+    "top_source": "microns_basil",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "microns_basil",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 327.6,
+  "round0_val": {
+    "val_dice": 0.3947513405618871,
+    "val_vi": 0.5091646853725434,
+    "val_ece": 0.14075125837595415
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 49,
+      "selection_seconds": 34.0,
+      "round_seconds": 169.4,
+      "teacher_dice_labeled": [
+        0.4166773458661978,
+        0.35928987385280325,
+        0.153101914721845,
+        0.3102874105923009,
+        0.24827571594727177,
+        0.40443287728637833,
+        0.45309617093263127,
+        0.3151852254719067
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 745730,
+      "evidence_nonempty_patches": 43,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.30178618768780596,
+      "restart_teacher_dice": 0.45309617093263127,
+      "weights": [
+        0.0523462898796494,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.9476537101203506,
+        0.0
+      ],
+      "ensemble_choice": "top2",
+      "candidate_dice": {
+        "mixture": 0.41707851191221385,
+        "uniform": 0.3962786850936092,
+        "top1": 0.45309617093263127,
+        "top2": 0.4560007178672816,
+        "top3": 0.4242840493249194
+      },
+      "val_dice": 0.649954056799635,
+      "val_vi": 0.46975670281025395,
+      "val_ece": 0.015682493902244875
+    },
+    {
+      "round": 2,
+      "n_labeled": 98,
+      "selection_seconds": 10.9,
+      "round_seconds": 31.3,
+      "teacher_dice_labeled": [
+        0.36611084588621456,
+        0.41390494215223367,
+        0.09410596399488945,
+        0.24220693615965957,
+        0.15365108848199444,
+        0.24885494485948062,
+        0.29824941598594545,
+        0.355243465280686
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1552702,
+      "evidence_nonempty_patches": 76,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.012823011947318825,
+        0.6947891955189256,
+        0.004286685040611986,
+        0.036731834491645354,
+        0.07570608809152632,
+        0.07008633566128063,
+        0.08182206111278438,
+        0.023754788135906927
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.41136455031192193,
+        "uniform": 0.2299235728798188,
+        "top1": 0.41390494215223367,
+        "top2": 0.4141551875692704,
+        "top3": 0.4140180268546756
+      },
+      "pseudo_student_dice": 0.5334437898129569,
+      "pseudo_ensemble_dice": 0.4503609960060156,
+      "pseudo_source": "student",
+      "consistency_scale": 0.111,
+      "val_dice": 0.6391891344279756,
+      "val_vi": 0.5097387246188724,
+      "val_ece": 0.04795928938682097
+    },
+    {
+      "round": 3,
+      "n_labeled": 147,
+      "selection_seconds": 10.8,
+      "round_seconds": 119.4,
+      "teacher_dice_labeled": [
+        0.36768339021727037,
+        0.3852192186783384,
+        0.10416450316146425,
+        0.24673547320391392,
+        0.2164043919966956,
+        0.3083610023527579,
+        0.37673879727901105,
+        0.34984706011994526
+      ],
+      "teacher_dice_pool": [
+        0.4249715484975724,
+        0.43163808487226474,
+        0.13402683120843648,
+        0.27919123067816787,
+        0.28581213424874735,
+        0.4018351343096836,
+        0.5029478154698421,
+        0.3738482728359761
+      ],
+      "evidence_fg_pixels": 2045013,
+      "evidence_nonempty_patches": 111,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.012424062776149956,
+        0.49741245892995734,
+        0.004521154756190894,
+        0.03665674701240556,
+        0.13524121291846142,
+        0.12120177924292354,
+        0.17107619568913765,
+        0.02146638867477356
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.40562761102178846,
+        "uniform": 0.3019242904620139,
+        "top1": 0.3852192186783384,
+        "top2": 0.40122242697931965,
+        "top3": 0.40084130714444877
+      },
+      "pseudo_student_dice": 0.3206224561686304,
+      "pseudo_ensemble_dice": 0.31636739169836314,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.6526551922105499,
+      "val_vi": 0.48330637836874146,
+      "val_ece": 0.01201782947328647
+    },
+    {
+      "round": 4,
+      "n_labeled": 196,
+      "selection_seconds": 10.7,
+      "round_seconds": 123.8,
+      "teacher_dice_labeled": [
+        0.3669733595372897,
+        0.3639770152858429,
+        0.10806844150474738,
+        0.25307971683616926,
+        0.22584242770873106,
+        0.320093417456483,
+        0.38747785383420064,
+        0.3311470636749818
+      ],
+      "teacher_dice_pool": [
+        0.42873382874298777,
+        0.4201025611225556,
+        0.1465960471130387,
+        0.27571882770491135,
+        0.3152241379813523,
+        0.4274018046756167,
+        0.5223327910218584,
+        0.3509337191657171
+      ],
+      "evidence_fg_pixels": 2546814,
+      "evidence_nonempty_patches": 151,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.020389457321379638,
+        0.6647963031539222,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.31481423952469817,
+        0.0
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.3754659094523485,
+        "uniform": 0.31416604860326686,
+        "top1": 0.38747785383420064,
+        "top2": 0.3897138845510255,
+        "top3": 0.39374676780961954
+      },
+      "pseudo_student_dice": 0.3715079220082423,
+      "pseudo_ensemble_dice": 0.24320091987757675,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.71405290677071,
+      "val_vi": 0.3943821589053846,
+      "val_ece": 0.0050473695531796405
+    },
+    {
+      "round": 5,
+      "n_labeled": 245,
+      "selection_seconds": 10.6,
+      "round_seconds": 138.8,
+      "teacher_dice_labeled": [
+        0.3502083366531371,
+        0.34845351115348644,
+        0.11604729844495462,
+        0.2621757924425775,
+        0.22619772716440484,
+        0.33203357450989535,
+        0.3840164960993779,
+        0.313400181441156
+      ],
+      "teacher_dice_pool": [
+        0.37548772339205577,
+        0.3833013808355072,
+        0.1600921317997613,
+        0.25321508996470743,
+        0.31990754606797517,
+        0.4370099709404248,
+        0.5179453131615449,
+        0.280774759856714
+      ],
+      "evidence_fg_pixels": 3577893,
+      "evidence_nonempty_patches": 189,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.01936135902517909,
+        0.6391615107247445,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.3414771302500763,
+        0.0
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.3599807604482505,
+        "uniform": 0.3211029686375517,
+        "top1": 0.3840164960993779,
+        "top2": 0.38579360695658055,
+        "top3": 0.3835683473682102
+      },
+      "pseudo_student_dice": 0.5884572297615537,
+      "pseudo_ensemble_dice": 0.2919429375125578,
+      "pseudo_source": "student",
+      "consistency_scale": 0.295,
+      "val_dice": 0.7095045778301132,
+      "val_vi": 0.3975579165871835,
+      "val_ece": 0.0022580185247888613
+    },
+    {
+      "round": 6,
+      "n_labeled": 294,
+      "selection_seconds": 10.6,
+      "round_seconds": 145.1,
+      "teacher_dice_labeled": [
+        0.3652716200083902,
+        0.3585971398683318,
+        0.10992564142649246,
+        0.2578325740522286,
+        0.22277450818970798,
+        0.33973973894072296,
+        0.3846495074209926,
+        0.3305959501728769
+      ],
+      "teacher_dice_pool": [
+        0.37545897608442663,
+        0.36517019215223734,
+        0.15792099420152225,
+        0.2615905184780217,
+        0.330487145457808,
+        0.46403301342316666,
+        0.532613356475908,
+        0.28124866584265834
+      ],
+      "evidence_fg_pixels": 4756463,
+      "evidence_nonempty_patches": 232,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.020966771040059683,
+        0.6589319110869489,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.3201013178729914,
+        0.0
+      ],
+      "ensemble_choice": "top3",
+      "candidate_dice": {
+        "mixture": 0.3652920718151623,
+        "uniform": 0.3223976533149485,
+        "top1": 0.3846495074209926,
+        "top2": 0.38678028399483,
+        "top3": 0.39248325722874766
+      },
+      "pseudo_student_dice": 0.5404097731059258,
+      "pseudo_ensemble_dice": 0.5780078399479068,
+      "pseudo_source": "student",
+      "consistency_scale": 0.135,
+      "val_dice": 0.7143403163792835,
+      "val_vi": 0.40830035269899356,
+      "val_ece": 0.007214722442033303
+    },
+    {
+      "round": 7,
+      "n_labeled": 294,
+      "selection_seconds": 0.0,
+      "round_seconds": 56.6,
+      "final_consolidation": true,
+      "val_dice": 0.7309256869658374,
+      "val_vi": 0.3713934767695636,
+      "val_ece": 0.015823260301908625
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.20304753183898672,
+    "auroc_top20_error": 0.41592640068365117
+  },
+  "al_seconds": 798.1,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.020966771040059683,
+    0.6589319110869489,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.3201013178729914,
+    0.0
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.6589319110869489,
+    "entropy": 0.7205294126502648,
+    "effective_sources": 2.055521141531504,
+    "min_weight": 0.0
+  },
+  "top_source_final": "dhanyasi-P14-mouse-cerebellum",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7117326324003791,
+    "patch_iou": 0.6289278430482521,
+    "patch_precision": 0.7384001631619221,
+    "patch_recall": 0.8160601597829159,
+    "patch_vi": 0.34099656215267415,
+    "n_patches": 1300,
+    "cal_ece": 0.014679185918989899,
+    "cal_nll": 0.17894412255736725,
+    "cal_brier": 0.051316040682139796,
+    "cal_mean_confidence": 0.9461751954929392,
+    "cal_accuracy": 0.9315347055288462,
+    "cal_predicted_prevalence": 0.22203125,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.22613713504306693
+  },
+  "test_notta": {
+    "patch_dice": 0.7117326324003791,
+    "patch_iou": 0.6289278430482521,
+    "patch_precision": 0.7384001631619221,
+    "patch_recall": 0.8160601597829159,
+    "patch_vi": 0.34099656215267415,
+    "n_patches": 1300,
+    "cal_ece": 0.014679185918989899,
+    "cal_nll": 0.17894412255736725,
+    "cal_brier": 0.051316040682139796,
+    "cal_mean_confidence": 0.9461751954929392,
+    "cal_accuracy": 0.9315347055288462,
+    "cal_predicted_prevalence": 0.22203125,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.22613713504306693
+  },
+  "test": {
+    "patch_dice": 0.7328983606873765,
+    "patch_iou": 0.6539298086166608,
+    "patch_precision": 0.7794708341773859,
+    "patch_recall": 0.8101709827186498,
+    "patch_vi": 0.31509139568672867,
+    "n_patches": 1300,
+    "image_dice": 0.8398286469075072,
+    "image_iou": 0.737109970022708,
+    "image_precision": 0.8420662330155432,
+    "image_recall": 0.8598087525051639,
+    "image_vi": 0.3650706363380156,
+    "n_images": 52,
+    "cal_ece": 0.003320093970549238,
+    "cal_nll": 0.15897419372119356,
+    "cal_brier": 0.04664112409434639,
+    "cal_mean_confidence": 0.9397174220413317,
+    "cal_accuracy": 0.9365895432692307,
+    "cal_predicted_prevalence": 0.21556715745192306,
+    "cal_true_prevalence": 0.21078049879807692,
+    "cal_mean_probability": 0.2233936774178567
+  },
+  "eval_seconds": 20.4,
+  "total_seconds": 1180.9,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.15,
+        "peak_cpu_rss_mb": 1027.9,
+        "peak_gpu_alloc_mb": 1066.3,
+        "peak_gpu_reserved_mb": 1086.0
+      },
+      "features_and_weights": {
+        "seconds": 31.61,
+        "peak_cpu_rss_mb": 1801.2,
+        "peak_gpu_alloc_mb": 2820.1,
+        "peak_gpu_reserved_mb": 3212.0
+      },
+      "distillation": {
+        "seconds": 327.55,
+        "peak_cpu_rss_mb": 3676.1,
+        "peak_gpu_alloc_mb": 4303.7,
+        "peak_gpu_reserved_mb": 4638.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 798.09,
+        "peak_cpu_rss_mb": 3676.1,
+        "peak_gpu_alloc_mb": 4424.0,
+        "peak_gpu_reserved_mb": 4648.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 20.38,
+        "peak_cpu_rss_mb": 3676.1,
+        "peak_gpu_alloc_mb": 2623.4,
+        "peak_gpu_reserved_mb": 4648.0,
+        "patches": 1300,
+        "images": 52,
+        "patches_per_second": 63.8,
+        "seconds_per_image": 0.392
+      }
+    },
+    "peak_gpu_alloc_mb": 4424.0,
+    "peak_cpu_rss_mb": 3676.1,
+    "total_seconds": 1178.8
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/whole-mouse-brain/macs-v11-norestart-acq-mcdropout_b5_s2.pth"
+}

macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,49,34.0,169.4,"[0.4166773458661978, 0.35928987385280325, 0.153101914721845, 0.3102874105923009, 0.24827571594727177, 0.40443287728637833, 0.45309617093263127, 0.3151852254719067]",,745730,43,True,kept,0.30178618768780596,0.45309617093263127,"[0.0523462898796494, 0.0, 0.0, 0.0, 0.0, 0.0, 0.9476537101203506, 0.0]",top2,"{'mixture': 0.41707851191221385, 'uniform': 0.3962786850936092, 'top1': 0.45309617093263127, 'top2': 0.4560007178672816, 'top3': 0.4242840493249194}",0.649954056799635,0.46975670281025395,0.015682493902244875,,,,,
+2,98,10.9,31.3,"[0.36611084588621456, 0.41390494215223367, 0.09410596399488945, 0.24220693615965957, 0.15365108848199444, 0.24885494485948062, 0.29824941598594545, 0.355243465280686]",,1552702,76,False,,,,"[0.012823011947318825, 0.6947891955189256, 0.004286685040611986, 0.036731834491645354, 0.07570608809152632, 0.07008633566128063, 0.08182206111278438, 0.023754788135906927]",mixture,"{'mixture': 0.41136455031192193, 'uniform': 0.2299235728798188, 'top1': 0.41390494215223367, 'top2': 0.4141551875692704, 'top3': 0.4140180268546756}",0.6391891344279756,0.5097387246188724,0.04795928938682097,0.5334437898129569,0.4503609960060156,student,0.111,
+3,147,10.8,119.4,"[0.36768339021727037, 0.3852192186783384, 0.10416450316146425, 0.24673547320391392, 0.2164043919966956, 0.3083610023527579, 0.37673879727901105, 0.34984706011994526]","[0.4249715484975724, 0.43163808487226474, 0.13402683120843648, 0.27919123067816787, 0.28581213424874735, 0.4018351343096836, 0.5029478154698421, 0.3738482728359761]",2045013,111,False,,,,"[0.012424062776149956, 0.49741245892995734, 0.004521154756190894, 0.03665674701240556, 0.13524121291846142, 0.12120177924292354, 0.17107619568913765, 0.02146638867477356]",mixture,"{'mixture': 0.40562761102178846, 'uniform': 0.3019242904620139, 'top1': 0.3852192186783384, 'top2': 0.40122242697931965, 'top3': 0.40084130714444877}",0.6526551922105499,0.48330637836874146,0.01201782947328647,0.3206224561686304,0.31636739169836314,student,0.0,
+4,196,10.7,123.8,"[0.3669733595372897, 0.3639770152858429, 0.10806844150474738, 0.25307971683616926, 0.22584242770873106, 0.320093417456483, 0.38747785383420064, 0.3311470636749818]","[0.42873382874298777, 0.4201025611225556, 0.1465960471130387, 0.27571882770491135, 0.3152241379813523, 0.4274018046756167, 0.5223327910218584, 0.3509337191657171]",2546814,151,False,,,,"[0.020389457321379638, 0.6647963031539222, 0.0, 0.0, 0.0, 0.0, 0.31481423952469817, 0.0]",top3,"{'mixture': 0.3754659094523485, 'uniform': 0.31416604860326686, 'top1': 0.38747785383420064, 'top2': 0.3897138845510255, 'top3': 0.39374676780961954}",0.71405290677071,0.3943821589053846,0.0050473695531796405,0.3715079220082423,0.24320091987757675,student,0.0,
+5,245,10.6,138.8,"[0.3502083366531371, 0.34845351115348644, 0.11604729844495462, 0.2621757924425775, 0.22619772716440484, 0.33203357450989535, 0.3840164960993779, 0.313400181441156]","[0.37548772339205577, 0.3833013808355072, 0.1600921317997613, 0.25321508996470743, 0.31990754606797517, 0.4370099709404248, 0.5179453131615449, 0.280774759856714]",3577893,189,False,,,,"[0.01936135902517909, 0.6391615107247445, 0.0, 0.0, 0.0, 0.0, 0.3414771302500763, 0.0]",top3,"{'mixture': 0.3599807604482505, 'uniform': 0.3211029686375517, 'top1': 0.3840164960993779, 'top2': 0.38579360695658055, 'top3': 0.3835683473682102}",0.7095045778301132,0.3975579165871835,0.0022580185247888613,0.5884572297615537,0.2919429375125578,student,0.295,
+6,294,10.6,145.1,"[0.3652716200083902, 0.3585971398683318, 0.10992564142649246, 0.2578325740522286, 0.22277450818970798, 0.33973973894072296, 0.3846495074209926, 0.3305959501728769]","[0.37545897608442663, 0.36517019215223734, 0.15792099420152225, 0.2615905184780217, 0.330487145457808, 0.46403301342316666, 0.532613356475908, 0.28124866584265834]",4756463,232,False,,,,"[0.020966771040059683, 0.6589319110869489, 0.0, 0.0, 0.0, 0.0, 0.3201013178729914, 0.0]",top3,"{'mixture': 0.3652920718151623, 'uniform': 0.3223976533149485, 'top1': 0.3846495074209926, 'top2': 0.38678028399483, 'top3': 0.39248325722874766}",0.7143403163792835,0.40830035269899356,0.007214722442033303,0.5404097731059258,0.5780078399479068,student,0.135,
+7,294,0.0,56.6,,,,,,,,,,,,0.7309256869658374,0.3713934767695636,0.015823260301908625,,,,,True

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-3/log.txt

@@ -0,0 +1,107 @@
+[11:20:15][      0s][START] target=drosophila-vnc variant=macs-v11-norestart-acq-laplace budget=5.0% seed=3 | {"task": "mitochondria", "target": "drosophila-vnc", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 3, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "laplace", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[11:20:15][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[11:20:15][      0s][DATA] drosophila-vnc: train=686 val=147 test=147 patches
+[11:20:16][      1s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, multiclass=2.60
+[11:20:16][      1s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s
+[11:20:17][      2s][WEIGHTS] mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020 | eff.sources=1.77 | MMD2 mix=0.1975 uniform=0.3743
+[11:20:17][      2s][WEIGHTS] per-source: fly-mito=0.039, kasthuri=0.943, lucchi=0.014, multiclass=0.004 | min-MMD source (NeuroADDA rule) = kasthuri
+[11:20:21][      5s][TEACHERS] target-val Dice: fly-mito=0.278, kasthuri=0.183, lucchi=0.058, multiclass=0.017
+[11:20:21][      5s][TEACHERS] ensemble target-val Dice: mixture=0.187, mixture@raw=0.185, mixture@logit=0.193, per_source=0.184, per_source@raw=0.184, per_source@logit=0.189, uniform=0.070, uniform@raw=0.088, uniform@logit=0.065
+[11:20:21][      5s][WEIGHTS] using mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020
+[11:20:21][      5s][INIT] student initialised from teacher kasthuri (weight 0.838)
+[11:20:22][      6s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[11:20:22][      6s][KD] distilling the mixture ensemble (prob space) into the student
+[11:20:26][     10s][KD] soft labels for 686 patches from 4 teachers in 4s
+[11:20:32][     17s][KD] epoch 1/5 | patches=686 | loss=0.1985 | 6s
+[11:20:38][     23s][KD] epoch 2/5 | patches=686 | loss=0.1315 | 6s
+[11:20:44][     29s][KD] epoch 3/5 | patches=686 | loss=0.1251 | 6s
+[11:20:51][     36s][KD] epoch 4/5 | patches=686 | loss=0.1224 | 6s
+[11:20:57][     42s][KD] epoch 5/5 | patches=686 | loss=0.1199 | 6s
+[11:20:57][     42s][ROUND0] before any target label: val_dice=0.2226 val_vi=0.2217
+[11:20:57][     42s][AL] patches=686 (pool 686 after dropping blank; free blank negatives: True) | budget=34 (5.0%) | rounds=6 | per_round=5 | acquisition=laplace | first_round=stratified
+[11:21:00][     44s][AL] round-1 uncertainty quality on 686 patches: {'spearman_uncertainty_error': 0.10850904905588832, 'auroc_top20_error': 0.5691314926478366}
+[11:21:04][     48s][AL] round 1: stratified cold start: 3 patches from the membrane-rich half (ensemble prevalence >= 0.047) + 2 from the membrane-poor half of 686 candidates
+[11:21:04][     49s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.573, uniform=0.044, top1=0.572, top2=0.573, top3=0.573 -> using mixture
+[11:21:04][     49s][WEIGHTS] round 1: student kept (labeled Dice 0.659 vs best teacher kasthuri 0.572); distilling the corrected ensemble into it
+[11:21:04][     49s][WEIGHTS] round 1: teacher Dice on 5 labeled patches + 0 blank (17096 fg px): fly-mito=0.327, kasthuri=0.572, lucchi=0.000, multiclass=0.001 | weights: fly-mito=0.012, kasthuri=0.988, lucchi=0.000, multiclass=0.000
+[11:21:08][     53s][KD] soft labels for 681 patches from 4 teachers in 4s
+[11:21:15][     59s][KD] epoch 1/2 | patches=681 | loss=0.0830 | 6s
+[11:21:21][     66s][KD] epoch 2/2 | patches=681 | loss=0.0810 | 6s
+[11:21:25][     70s][CONS] pseudo-labels for 681 unlabeled patches from the current ensemble in 4s
+[11:21:25][     70s][FT] round 1/6 epoch 1/5 | patches=5 | loss=0.8492
+[11:21:26][     71s][FT] round 1/6 epoch 2/5 | patches=5 | loss=0.8436
+[11:21:26][     71s][FT] round 1/6 epoch 3/5 | patches=5 | loss=0.8554
+[11:21:26][     71s][FT] round 1/6 epoch 4/5 | patches=5 | loss=0.7927
+[11:21:27][     72s][FT] round 1/6 epoch 5/5 | patches=5 | loss=0.8922
+[11:21:27][     72s][AL] round 1/6 | labeled=5 | selection_seconds=6.4000 round_seconds=29.6000 restart_student_dice=0.6587 restart_teacher_dice=0.5724 val_dice=0.2651 val_vi=0.2537 val_ece=0.0339
+[11:21:30][     74s][AL] round 2: 126/681 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:21:30][     75s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.291, uniform=0.022, top1=0.344, top2=0.291, top3=0.291 -> using top1
+[11:21:30][     75s][WEIGHTS] round 2: teacher Dice on 10 labeled patches + 0 blank (49513 fg px): fly-mito=0.344, kasthuri=0.288, lucchi=0.002, multiclass=0.003 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[11:21:31][     76s][CONS] round 2: on the 5 new patches student Dice=0.049 vs ensemble 0.355 -> pseudo-labels from the ensemble | consistency x1.00
+[11:21:32][     77s][FT] round 2/6 epoch 1/5 | patches=10 | loss=1.2737
+[11:21:32][     77s][FT] round 2/6 epoch 2/5 | patches=10 | loss=1.1118
+[11:21:33][     78s][FT] round 2/6 epoch 3/5 | patches=10 | loss=0.9654
+[11:21:34][     78s][FT] round 2/6 epoch 4/5 | patches=10 | loss=0.9057
+[11:21:34][     79s][FT] round 2/6 epoch 5/5 | patches=10 | loss=0.8844
+[11:21:35][     79s][AL] round 2/6 | labeled=10 | selection_seconds=2.3000 round_seconds=6.8000 pseudo_student_dice=0.0485 pseudo_ensemble_dice=0.3552 consistency_scale=1.0000 val_dice=0.5304 val_vi=0.1854 val_ece=0.0069
+[11:21:37][     82s][AL] round 3: 113/676 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:21:38][     83s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.228, uniform=0.020, top1=0.313, top2=0.229, top3=0.229 -> using top1
+[11:21:38][     83s][WEIGHTS] round 3: teacher Dice on 15 labeled patches + 0 blank (67007 fg px): fly-mito=0.313, kasthuri=0.224, lucchi=0.002, multiclass=0.005 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[11:21:40][     85s][CONS] round 3: on the 5 new patches student Dice=0.371 vs ensemble 0.254 -> pseudo-labels from the student | consistency x0.00
+[11:21:40][     85s][FT] round 3/6 epoch 1/5 | patches=15 | loss=0.7781
+[11:21:41][     85s][FT] round 3/6 epoch 2/5 | patches=15 | loss=0.7260
+[11:21:41][     86s][FT] round 3/6 epoch 3/5 | patches=15 | loss=0.6272
+[11:21:41][     86s][FT] round 3/6 epoch 4/5 | patches=15 | loss=0.5273
+[11:21:42][     87s][FT] round 3/6 epoch 5/5 | patches=15 | loss=0.4273
+[11:21:42][     87s][AL] round 3/6 | labeled=15 | selection_seconds=2.4000 round_seconds=7.2000 pseudo_student_dice=0.3711 pseudo_ensemble_dice=0.2543 consistency_scale=0.0000 val_dice=0.5918 val_vi=0.1755 val_ece=0.0097
+[11:21:45][     89s][AL] round 4: 47/671 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:21:54][     99s][WEIGHTS] round 4: teacher agreement with the student on 666 unlabeled patches: fly-mito=0.343, kasthuri=0.444, lucchi=0.010, multiclass=0.002 | blended with labeled Dice at 0.50
+[11:22:01][    106s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.341, uniform=0.035, top1=0.341, top2=0.341, top3=0.341 -> using mixture
+[11:22:01][    106s][WEIGHTS] round 4: teacher Dice on 20 labeled patches + 0 blank (74131 fg px): fly-mito=0.315, kasthuri=0.341, lucchi=0.006, multiclass=0.003 | weights: fly-mito=0.097, kasthuri=0.901, lucchi=0.001, multiclass=0.001
+[11:22:03][    108s][CONS] round 4: on the 5 new patches student Dice=0.436 vs ensemble 0.357 -> pseudo-labels from the student | consistency x0.00
+[11:22:04][    109s][FT] round 4/6 epoch 1/5 | patches=20 | loss=0.4633
+[11:22:04][    109s][FT] round 4/6 epoch 2/5 | patches=20 | loss=0.4212
+[11:22:05][    110s][FT] round 4/6 epoch 3/5 | patches=20 | loss=0.3607
+[11:22:05][    110s][FT] round 4/6 epoch 4/5 | patches=20 | loss=0.3613
+[11:22:06][    111s][FT] round 4/6 epoch 5/5 | patches=20 | loss=0.3089
+[11:22:06][    111s][AL] round 4/6 | labeled=20 | selection_seconds=2.3000 round_seconds=23.4000 pseudo_student_dice=0.4365 pseudo_ensemble_dice=0.3567 consistency_scale=0.0000 val_dice=0.7202 val_vi=0.1155 val_ece=0.0083
+[11:22:08][    113s][AL] round 5: 104/666 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:22:17][    122s][WEIGHTS] round 5: teacher agreement with the student on 661 unlabeled patches: fly-mito=0.382, kasthuri=0.466, lucchi=0.009, multiclass=0.001 | blended with labeled Dice at 0.50
+[11:22:25][    130s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.374, uniform=0.044, top1=0.373, top2=0.374, top3=0.374 -> using mixture
+[11:22:25][    130s][WEIGHTS] round 5: teacher Dice on 25 labeled patches + 0 blank (82397 fg px): fly-mito=0.327, kasthuri=0.373, lucchi=0.012, multiclass=0.003 | weights: fly-mito=0.081, kasthuri=0.918, lucchi=0.001, multiclass=0.001
+[11:22:27][    132s][CONS] round 5: on the 5 new patches student Dice=0.860 vs ensemble 0.498 -> pseudo-labels from the student | consistency x1.00
+[11:22:28][    133s][FT] round 5/6 epoch 1/5 | patches=25 | loss=0.3228
+[11:22:29][    134s][FT] round 5/6 epoch 2/5 | patches=25 | loss=0.4003
+[11:22:29][    134s][FT] round 5/6 epoch 3/5 | patches=25 | loss=0.3622
+[11:22:30][    135s][FT] round 5/6 epoch 4/5 | patches=25 | loss=0.2707
+[11:22:31][    136s][FT] round 5/6 epoch 5/5 | patches=25 | loss=0.2498
+[11:22:32][    136s][AL] round 5/6 | labeled=25 | selection_seconds=2.3000 round_seconds=25.1000 pseudo_student_dice=0.8602 pseudo_ensemble_dice=0.4984 consistency_scale=1.0000 val_dice=0.7097 val_vi=0.1201 val_ece=0.0088
+[11:22:34][    139s][AL] round 6: 111/661 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[11:22:43][    148s][WEIGHTS] round 6: teacher agreement with the student on 656 unlabeled patches: fly-mito=0.391, kasthuri=0.449, lucchi=0.009, multiclass=0.001 | blended with labeled Dice at 0.50
+[11:22:51][    156s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.394, uniform=0.048, top1=0.394, top2=0.394, top3=0.394 -> using mixture
+[11:22:51][    156s][WEIGHTS] round 6: teacher Dice on 30 labeled patches + 0 blank (91279 fg px): fly-mito=0.327, kasthuri=0.394, lucchi=0.011, multiclass=0.002 | weights: fly-mito=0.067, kasthuri=0.932, lucchi=0.001, multiclass=0.000
+[11:22:53][    158s][CONS] round 6: on the 5 new patches student Dice=0.907 vs ensemble 0.639 -> pseudo-labels from the student | consistency x1.00
+[11:22:54][    158s][FT] round 6/6 epoch 1/5 | patches=30 | loss=0.2780
+[11:22:54][    159s][FT] round 6/6 epoch 2/5 | patches=30 | loss=0.2534
+[11:22:55][    160s][FT] round 6/6 epoch 3/5 | patches=30 | loss=0.2623
+[11:22:56][    161s][FT] round 6/6 epoch 4/5 | patches=30 | loss=0.2548
+[11:22:57][    162s][FT] round 6/6 epoch 5/5 | patches=30 | loss=0.2277
+[11:22:58][    162s][AL] round 6/6 | labeled=30 | selection_seconds=2.4000 round_seconds=25.5000 pseudo_student_dice=0.9074 pseudo_ensemble_dice=0.6390 consistency_scale=1.0000 val_dice=0.7320 val_vi=0.1066 val_ece=0.0094
+[11:22:59][    163s][CONS] final consolidation: pseudo-labels refreshed from the student
+[11:22:59][    164s][FT] final 1/10 epoch 1/1 | patches=30 | loss=0.2279
+[11:23:00][    165s][FT] final 2/10 epoch 1/1 | patches=30 | loss=0.1906
+[11:23:01][    166s][FT] final 3/10 epoch 1/1 | patches=30 | loss=0.1898
+[11:23:02][    167s][FT] final 4/10 epoch 1/1 | patches=30 | loss=0.1942
+[11:23:03][    168s][FT] final 5/10 epoch 1/1 | patches=30 | loss=0.1911
+[11:23:04][    169s][FT] final 6/10 epoch 1/1 | patches=30 | loss=0.1783
+[11:23:05][    169s][FT] final 7/10 epoch 1/1 | patches=30 | loss=0.1626
+[11:23:05][    170s][FT] final 8/10 epoch 1/1 | patches=30 | loss=0.2015
+[11:23:06][    171s][FT] final 9/10 epoch 1/1 | patches=30 | loss=0.1667
+[11:23:07][    172s][FT] final 10/10 epoch 1/1 | patches=30 | loss=0.1817
+[11:23:08][    172s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=9.6000 val_dice=0.7444 val_vi=0.1016 val_ece=0.0095
+[11:23:08][    173s][WEIGHTS] final: fly-mito=0.067, kasthuri=0.932, lucchi=0.001, multiclass=0.000
+[11:23:10][    175s][TEST] raw student (no EMA, no TTA): patch_dice=0.7977 patch_vi=0.1139
+[11:23:11][    175s][TEST] final model without TTA: patch_dice=0.7977 patch_vi=0.1139
+[11:23:11][    175s][RESOURCES] {"phases": {"setup": {"seconds": 0.64, "peak_cpu_rss_mb": 1025.3, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 5.64, "peak_cpu_rss_mb": 1806.2, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 35.29, "peak_cpu_rss_mb": 2104.5, "peak_gpu_alloc_mb": 3829.8, "peak_gpu_reserved_mb": 4168.0, "kd_epochs": 5}, "active_learning": {"seconds": 130.37, "peak_cpu_rss_mb": 2410.7, "peak_gpu_alloc_mb": 3950.0, "peak_gpu_reserved_mb": 4176.0, "rounds": 6}, "inference": {"seconds": 2.88, "peak_cpu_rss_mb": 2410.7, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4176.0, "patches": 147, "images": 3, "patches_per_second": 51.0, "seconds_per_image": 0.96}}, "peak_gpu_alloc_mb": 3950.0, "peak_cpu_rss_mb": 2410.7, "total_seconds": 174.8}
+[11:23:11][    175s][TEST] patch_dice=0.8348 patch_iou=0.7589 patch_precision=0.8195 patch_recall=0.9316 patch_vi=0.0987 image_dice=0.9122 image_iou=0.8396 image_precision=0.8615 image_recall=0.9710 image_vi=0.1005 cal_ece=0.0037 cal_nll=0.0333 cal_brier=0.0097 cal_mean_confidence=0.9907 cal_accuracy=0.9874 cal_predicted_prevalence=0.0674 cal_true_prevalence=0.0594 cal_mean_probability=0.0674
+[11:23:11][    176s][DONE] drosophila-vnc macs-v11-norestart-acq-laplace b5 s3 | test patch Dice=0.8348 VI=0.0987 | image Dice=0.9122 VI=0.1005 | 175s

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-3/result.json

@@ -0,0 +1,662 @@
+{
+  "target": "drosophila-vnc",
+  "variant": "macs-v11-norestart-acq-laplace",
+  "budget": 5.0,
+  "seed": 3,
+  "config": {
+    "task": "mitochondria",
+    "target": "drosophila-vnc",
+    "variant": "macs-v11-norestart-acq-laplace",
+    "budget": 5.0,
+    "seed": 3,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "laplace",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "drosophila-vnc",
+      "--variants",
+      "macs-v11-norestart-acq-laplace",
+      "--budgets",
+      "5",
+      "--seeds",
+      "3",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--task",
+      "mitochondria",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "fly-brain-mito-benchmark",
+    "kasthuri-pp",
+    "lucchi-pp",
+    "multiclass"
+  ],
+  "alphas": [
+    0.11690531690536884,
+    0.8375898290916663,
+    0.02560239431814963,
+    0.01990245968481518
+  ],
+  "weights_all": {
+    "sources": [
+      "fly-brain-mito-benchmark",
+      "kasthuri-pp",
+      "lucchi-pp",
+      "multiclass"
+    ],
+    "mixture": [
+      0.11690531690536884,
+      0.8375898290916663,
+      0.02560239431814963,
+      0.01990245968481518
+    ],
+    "per_source": [
+      0.038599381604120485,
+      0.943285983973492,
+      0.013712262708195049,
+      0.004402371714192633
+    ],
+    "uniform": [
+      0.25,
+      0.25,
+      0.25,
+      0.25
+    ],
+    "mixture_info": {
+      "objective": 0.1917814945701421,
+      "mmd2_mixture": 0.19749308036725388,
+      "mmd2_uniform": 0.3743480458439246,
+      "mmd2_single_source": [
+        0.7042945447726864,
+        0.2070680114579429,
+        0.6307945959829456,
+        0.6534353423959632
+      ],
+      "median_squared_distance": 0.1964580928585633,
+      "gammas": [
+        1.2725360221224549,
+        2.5450720442449097,
+        5.0901440884898195,
+        10.180288176979639,
+        20.360576353959278
+      ],
+      "max_weight": 0.8375898290916663,
+      "entropy": 0.5711585797111782,
+      "effective_sources": 1.7703169170005038,
+      "min_weight": 0.01990245968481518
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.6050756121533214,
+        0.3110169227475098,
+        0.7002952924749369,
+        0.8048259551971948
+      ],
+      "standardized": [
+        -0.0012381647254944845,
+        -1.5993047903844642,
+        0.5162347028009382,
+        1.0843082523090217
+      ],
+      "tau": 0.5,
+      "max_weight": 0.943285983973492,
+      "entropy": 0.2634007298678481,
+      "effective_sources": 1.3013481035311127,
+      "min_weight": 0.004402371714192633
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "kasthuri-pp",
+    "seconds": 1.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.2784035818310388,
+      0.18341782772166676,
+      0.057702367443012846,
+      0.01681466600081078
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.186522633793688,
+      "mixture@raw": 0.18548325188563394,
+      "mixture@logit": 0.19296315829990748,
+      "per_source": 0.18415887092754818,
+      "per_source@raw": 0.18350562678267812,
+      "per_source@logit": 0.1890579186582894,
+      "uniform": 0.06967222752121038,
+      "uniform@raw": 0.0884686387663988,
+      "uniform@logit": 0.06545324142703345
+    },
+    "val_patches": 147
+  },
+  "temperatures": [
+    1.3819772005081177,
+    2.5080416202545166,
+    3.7969236373901367,
+    2.597255229949951
+  ],
+  "top_source": "kasthuri-pp",
+  "weight_stats": {
+    "max_weight": 0.8375898290916663,
+    "entropy": 0.5711585797111782,
+    "effective_sources": 1.7703169170005038,
+    "min_weight": 0.01990245968481518
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.7999999999999999,
+    "spearman_weight_taxonomy": -0.21081851067789195,
+    "top_source": "kasthuri-pp",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "kasthuri-pp",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 35.3,
+  "round0_val": {
+    "val_dice": 0.22259710119393636,
+    "val_vi": 0.2217486787180137,
+    "val_ece": 0.019118359085155114
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 5,
+      "selection_seconds": 6.4,
+      "round_seconds": 29.6,
+      "teacher_dice_labeled": [
+        0.3265584832904884,
+        0.5723588778124132,
+        5.7760064691272456e-05,
+        0.0008705240554813998
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 17096,
+      "evidence_nonempty_patches": 4,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.6586710683430498,
+      "restart_teacher_dice": 0.5723588778124132,
+      "weights": [
+        0.011805122497730317,
+        0.9880187339743336,
+        9.875115690298976e-05,
+        7.739237103305626e-05
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.5725535002933638,
+        "uniform": 0.04434476493300023,
+        "top1": 0.5723588778124132,
+        "top2": 0.5725535002933638,
+        "top3": 0.5725535002933638
+      },
+      "val_dice": 0.2651183621363361,
+      "val_vi": 0.2536928580768912,
+      "val_ece": 0.03392489668719592
+    },
+    {
+      "round": 2,
+      "n_labeled": 10,
+      "selection_seconds": 2.3,
+      "round_seconds": 6.8,
+      "teacher_dice_labeled": [
+        0.34446258487008174,
+        0.2877706560379188,
+        0.0021410276932927806,
+        0.002911120480234496
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 49513,
+      "evidence_nonempty_patches": 9,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.2912965056802886,
+        "uniform": 0.02191753601589667,
+        "top1": 0.34446258487008174,
+        "top2": 0.2913448340824071,
+        "top3": 0.2913145287183076
+      },
+      "pseudo_student_dice": 0.048524089711711266,
+      "pseudo_ensemble_dice": 0.35520279127149357,
+      "pseudo_source": "ensemble",
+      "consistency_scale": 1.0,
+      "val_dice": 0.5303885061058371,
+      "val_vi": 0.1853514846468518,
+      "val_ece": 0.006875929179194766
+    },
+    {
+      "round": 3,
+      "n_labeled": 15,
+      "selection_seconds": 2.4,
+      "round_seconds": 7.2,
+      "teacher_dice_labeled": [
+        0.3131746656176239,
+        0.22434563791634468,
+        0.002236142580005035,
+        0.005071563756801514
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 67007,
+      "evidence_nonempty_patches": 14,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.22849699359095008,
+        "uniform": 0.01984394708280778,
+        "top1": 0.3131746656176239,
+        "top2": 0.22876961744740645,
+        "top3": 0.2286186652938789
+      },
+      "pseudo_student_dice": 0.3711265455655625,
+      "pseudo_ensemble_dice": 0.25430492104338365,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.5918325005506857,
+      "val_vi": 0.17548227962258417,
+      "val_ece": 0.009676093882371277
+    },
+    {
+      "round": 4,
+      "n_labeled": 20,
+      "selection_seconds": 2.3,
+      "round_seconds": 23.4,
+      "teacher_dice_labeled": [
+        0.31547631096594764,
+        0.34123593983097356,
+        0.0059793031811133535,
+        0.0034405810135199576
+      ],
+      "teacher_dice_pool": [
+        0.3425381911105748,
+        0.4438936034888215,
+        0.009888876495357073,
+        0.002299520650677445
+      ],
+      "evidence_fg_pixels": 74131,
+      "evidence_nonempty_patches": 18,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.09720801201042216,
+        0.9010976140529733,
+        0.0009638733077765422,
+        0.0007305006288280354
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.3409487021488296,
+        "uniform": 0.03477595874011665,
+        "top1": 0.34123593983097356,
+        "top2": 0.34102760479558736,
+        "top3": 0.34098090067463205
+      },
+      "pseudo_student_dice": 0.43649569353387324,
+      "pseudo_ensemble_dice": 0.3566652909616178,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.7202478275643855,
+      "val_vi": 0.11551100599924623,
+      "val_ece": 0.00827971477000551
+    },
+    {
+      "round": 5,
+      "n_labeled": 25,
+      "selection_seconds": 2.3,
+      "round_seconds": 25.1,
+      "teacher_dice_labeled": [
+        0.32748527014276285,
+        0.3732444408424384,
+        0.01181775174042917,
+        0.002690437438876882
+      ],
+      "teacher_dice_pool": [
+        0.3817607993473308,
+        0.4657078120427067,
+        0.009169410998209292,
+        0.001259368545293597
+      ],
+      "evidence_fg_pixels": 82397,
+      "evidence_nonempty_patches": 23,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.08105022278551019,
+        0.917658090242581,
+        0.0007555686292442745,
+        0.0005361183426645543
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.3736519199765924,
+        "uniform": 0.04405944383878502,
+        "top1": 0.3732444408424384,
+        "top2": 0.3737204245680453,
+        "top3": 0.373694044966715
+      },
+      "pseudo_student_dice": 0.8601795540110049,
+      "pseudo_ensemble_dice": 0.4983764769549923,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.7097438014736671,
+      "val_vi": 0.12009500484663031,
+      "val_ece": 0.00879034593597929
+    },
+    {
+      "round": 6,
+      "n_labeled": 30,
+      "selection_seconds": 2.4,
+      "round_seconds": 25.5,
+      "teacher_dice_labeled": [
+        0.32742006938767654,
+        0.39371978128758006,
+        0.010862133549017123,
+        0.002418636813878437
+      ],
+      "teacher_dice_pool": [
+        0.39091620078222555,
+        0.4491281687389267,
+        0.008656109805333422,
+        0.0011212111372366253
+      ],
+      "evidence_fg_pixels": 91279,
+      "evidence_nonempty_patches": 28,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.06702579943774015,
+        0.9319124729363013,
+        0.0006192908650021591,
+        0.00044243676095642704
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.39385881615174234,
+        "uniform": 0.048451003797465905,
+        "top1": 0.39371978128758006,
+        "top2": 0.39387514465174567,
+        "top3": 0.39385988022438434
+      },
+      "pseudo_student_dice": 0.9074005794047932,
+      "pseudo_ensemble_dice": 0.638978505072492,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.7320486463671512,
+      "val_vi": 0.10657063095909225,
+      "val_ece": 0.009392135846348293
+    },
+    {
+      "round": 7,
+      "n_labeled": 30,
+      "selection_seconds": 0.0,
+      "round_seconds": 9.6,
+      "final_consolidation": true,
+      "val_dice": 0.7444199834311062,
+      "val_vi": 0.10161531775765006,
+      "val_ece": 0.009509370836124294
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.10850904905588832,
+    "auroc_top20_error": 0.5691314926478366
+  },
+  "al_seconds": 130.4,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.06702579943774015,
+    0.9319124729363013,
+    0.0006192908650021591,
+    0.00044243676095642704
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9319124729363013,
+    "entropy": 0.2548559221187666,
+    "effective_sources": 1.2902757072900808,
+    "min_weight": 0.00044243676095642704
+  },
+  "top_source_final": "kasthuri-pp",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7977066471645559,
+    "patch_iou": 0.712263446875584,
+    "patch_precision": 0.762104894454457,
+    "patch_recall": 0.9284656145665409,
+    "patch_vi": 0.11394406286768838,
+    "n_patches": 147,
+    "cal_ece": 0.010102019099312812,
+    "cal_nll": 0.04980973616501659,
+    "cal_brier": 0.012402466143411325,
+    "cal_mean_confidence": 0.9944937530522531,
+    "cal_accuracy": 0.9843949298469388,
+    "cal_predicted_prevalence": 0.06852545705782313,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06730707713274209
+  },
+  "test_notta": {
+    "patch_dice": 0.7977066471645559,
+    "patch_iou": 0.712263446875584,
+    "patch_precision": 0.762104894454457,
+    "patch_recall": 0.9284656145665409,
+    "patch_vi": 0.11394406286768838,
+    "n_patches": 147,
+    "cal_ece": 0.010102019099312812,
+    "cal_nll": 0.04980973616501659,
+    "cal_brier": 0.012402466143411325,
+    "cal_mean_confidence": 0.9944937530522531,
+    "cal_accuracy": 0.9843949298469388,
+    "cal_predicted_prevalence": 0.06852545705782313,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06730707713274209
+  },
+  "test": {
+    "patch_dice": 0.83483573452591,
+    "patch_iou": 0.7589463120306933,
+    "patch_precision": 0.8194718578465112,
+    "patch_recall": 0.931564815747692,
+    "patch_vi": 0.09867003215150039,
+    "n_patches": 147,
+    "image_dice": 0.9122375170485886,
+    "image_iou": 0.8396129565317924,
+    "image_precision": 0.8614710995782565,
+    "image_recall": 0.9710319246167654,
+    "image_vi": 0.10048469250582326,
+    "n_images": 3,
+    "cal_ece": 0.003700974915991743,
+    "cal_nll": 0.03327139559211665,
+    "cal_brier": 0.009700747389481646,
+    "cal_mean_confidence": 0.9907116323509012,
+    "cal_accuracy": 0.9874043367346939,
+    "cal_predicted_prevalence": 0.06740274234693877,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06744815259594922
+  },
+  "eval_seconds": 2.9,
+  "total_seconds": 175.4,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 0.64,
+        "peak_cpu_rss_mb": 1025.3,
+        "peak_gpu_alloc_mb": 592.4,
+        "peak_gpu_reserved_mb": 594.0
+      },
+      "features_and_weights": {
+        "seconds": 5.64,
+        "peak_cpu_rss_mb": 1806.2,
+        "peak_gpu_alloc_mb": 2298.2,
+        "peak_gpu_reserved_mb": 2640.0
+      },
+      "distillation": {
+        "seconds": 35.29,
+        "peak_cpu_rss_mb": 2104.5,
+        "peak_gpu_alloc_mb": 3829.8,
+        "peak_gpu_reserved_mb": 4168.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 130.37,
+        "peak_cpu_rss_mb": 2410.7,
+        "peak_gpu_alloc_mb": 3950.0,
+        "peak_gpu_reserved_mb": 4176.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 2.88,
+        "peak_cpu_rss_mb": 2410.7,
+        "peak_gpu_alloc_mb": 2149.5,
+        "peak_gpu_reserved_mb": 4176.0,
+        "patches": 147,
+        "images": 3,
+        "patches_per_second": 51.0,
+        "seconds_per_image": 0.96
+      }
+    },
+    "peak_gpu_alloc_mb": 3950.0,
+    "peak_cpu_rss_mb": 2410.7,
+    "total_seconds": 174.8
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/drosophila-vnc/macs-v11-norestart-acq-laplace_b5_s3.pth"
+}

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-3/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,5,6.4,29.6,"[0.3265584832904884, 0.5723588778124132, 5.7760064691272456e-05, 0.0008705240554813998]",,17096,4,True,kept,0.6586710683430498,0.5723588778124132,"[0.011805122497730317, 0.9880187339743336, 9.875115690298976e-05, 7.739237103305626e-05]",mixture,"{'mixture': 0.5725535002933638, 'uniform': 0.04434476493300023, 'top1': 0.5723588778124132, 'top2': 0.5725535002933638, 'top3': 0.5725535002933638}",0.2651183621363361,0.2536928580768912,0.03392489668719592,,,,,
+2,10,2.3,6.8,"[0.34446258487008174, 0.2877706560379188, 0.0021410276932927806, 0.002911120480234496]",,49513,9,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.2912965056802886, 'uniform': 0.02191753601589667, 'top1': 0.34446258487008174, 'top2': 0.2913448340824071, 'top3': 0.2913145287183076}",0.5303885061058371,0.1853514846468518,0.006875929179194766,0.048524089711711266,0.35520279127149357,ensemble,1.0,
+3,15,2.4,7.2,"[0.3131746656176239, 0.22434563791634468, 0.002236142580005035, 0.005071563756801514]",,67007,14,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.22849699359095008, 'uniform': 0.01984394708280778, 'top1': 0.3131746656176239, 'top2': 0.22876961744740645, 'top3': 0.2286186652938789}",0.5918325005506857,0.17548227962258417,0.009676093882371277,0.3711265455655625,0.25430492104338365,student,0.0,
+4,20,2.3,23.4,"[0.31547631096594764, 0.34123593983097356, 0.0059793031811133535, 0.0034405810135199576]","[0.3425381911105748, 0.4438936034888215, 0.009888876495357073, 0.002299520650677445]",74131,18,False,,,,"[0.09720801201042216, 0.9010976140529733, 0.0009638733077765422, 0.0007305006288280354]",mixture,"{'mixture': 0.3409487021488296, 'uniform': 0.03477595874011665, 'top1': 0.34123593983097356, 'top2': 0.34102760479558736, 'top3': 0.34098090067463205}",0.7202478275643855,0.11551100599924623,0.00827971477000551,0.43649569353387324,0.3566652909616178,student,0.0,
+5,25,2.3,25.1,"[0.32748527014276285, 0.3732444408424384, 0.01181775174042917, 0.002690437438876882]","[0.3817607993473308, 0.4657078120427067, 0.009169410998209292, 0.001259368545293597]",82397,23,False,,,,"[0.08105022278551019, 0.917658090242581, 0.0007555686292442745, 0.0005361183426645543]",mixture,"{'mixture': 0.3736519199765924, 'uniform': 0.04405944383878502, 'top1': 0.3732444408424384, 'top2': 0.3737204245680453, 'top3': 0.373694044966715}",0.7097438014736671,0.12009500484663031,0.00879034593597929,0.8601795540110049,0.4983764769549923,student,1.0,
+6,30,2.4,25.5,"[0.32742006938767654, 0.39371978128758006, 0.010862133549017123, 0.002418636813878437]","[0.39091620078222555, 0.4491281687389267, 0.008656109805333422, 0.0011212111372366253]",91279,28,False,,,,"[0.06702579943774015, 0.9319124729363013, 0.0006192908650021591, 0.00044243676095642704]",mixture,"{'mixture': 0.39385881615174234, 'uniform': 0.048451003797465905, 'top1': 0.39371978128758006, 'top2': 0.39387514465174567, 'top3': 0.39385988022438434}",0.7320486463671512,0.10657063095909225,0.009392135846348293,0.9074005794047932,0.638978505072492,student,1.0,
+7,30,0.0,9.6,,,,,,,,,,,,0.7444199834311062,0.10161531775765006,0.009509370836124294,,,,,True

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-1/log.txt

@@ -0,0 +1,108 @@
+[04:57:02][      0s][START] target=drosophila-vnc variant=macs-v11-norestart-acq-margin budget=5.0% seed=1 | {"task": "mitochondria", "target": "drosophila-vnc", "variant": "macs-v11-norestart-acq-margin", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "margin", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[04:57:02][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[04:57:02][      0s][DATA] drosophila-vnc: train=686 val=147 test=147 patches
+[04:57:03][      1s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, multiclass=2.60
+[04:57:03][      1s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s
+[04:57:04][      2s][WEIGHTS] mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020 | eff.sources=1.77 | MMD2 mix=0.1975 uniform=0.3743
+[04:57:04][      2s][WEIGHTS] per-source: fly-mito=0.039, kasthuri=0.943, lucchi=0.014, multiclass=0.004 | min-MMD source (NeuroADDA rule) = kasthuri
+[04:57:08][      6s][TEACHERS] target-val Dice: fly-mito=0.278, kasthuri=0.183, lucchi=0.058, multiclass=0.017
+[04:57:08][      6s][TEACHERS] ensemble target-val Dice: mixture=0.187, mixture@raw=0.185, mixture@logit=0.193, per_source=0.184, per_source@raw=0.184, per_source@logit=0.189, uniform=0.070, uniform@raw=0.088, uniform@logit=0.065
+[04:57:08][      6s][WEIGHTS] using mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020
+[04:57:08][      6s][INIT] student initialised from teacher kasthuri (weight 0.838)
+[04:57:09][      7s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[04:57:09][      7s][KD] distilling the mixture ensemble (prob space) into the student
+[04:57:13][     11s][KD] soft labels for 686 patches from 4 teachers in 4s
+[04:57:19][     17s][KD] epoch 1/5 | patches=686 | loss=0.2005 | 7s
+[04:57:26][     24s][KD] epoch 2/5 | patches=686 | loss=0.1322 | 7s
+[04:57:32][     30s][KD] epoch 3/5 | patches=686 | loss=0.1258 | 6s
+[04:57:39][     37s][KD] epoch 4/5 | patches=686 | loss=0.1216 | 6s
+[04:57:45][     43s][KD] epoch 5/5 | patches=686 | loss=0.1204 | 6s
+[04:57:46][     44s][ROUND0] before any target label: val_dice=0.1917 val_vi=0.2212
+[04:57:46][     44s][AL] patches=686 (pool 686 after dropping blank; free blank negatives: True) | budget=34 (5.0%) | rounds=6 | per_round=5 | acquisition=margin | first_round=stratified
+[04:57:47][     45s][AL] round-1 uncertainty quality on 686 patches: {'spearman_uncertainty_error': 0.08899991394046061, 'auroc_top20_error': 0.5872474346768222}
+[04:57:51][     49s][AL] round 1: stratified cold start: 3 patches from the membrane-rich half (ensemble prevalence >= 0.047) + 2 from the membrane-poor half of 686 candidates
+[04:57:51][     49s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.303, uniform=0.018, top1=0.312, top2=0.305, top3=0.304 -> using top1
+[04:57:52][     50s][WEIGHTS] round 1: student kept (labeled Dice 0.122 vs best teacher fly-mito 0.312); distilling the corrected ensemble into it
+[04:57:52][     50s][WEIGHTS] round 1: teacher Dice on 5 labeled patches + 0 blank (15368 fg px): fly-mito=0.312, kasthuri=0.097, lucchi=0.000, multiclass=0.003 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:57:53][     51s][KD] soft labels for 681 patches from 4 teachers in 1s
+[04:57:59][     57s][KD] epoch 1/2 | patches=681 | loss=0.2886 | 6s
+[04:58:06][     64s][KD] epoch 2/2 | patches=681 | loss=0.2383 | 6s
+[04:58:07][     65s][CONS] pseudo-labels for 681 unlabeled patches from the current ensemble in 1s
+[04:58:07][     65s][FT] round 1/6 epoch 1/5 | patches=5 | loss=1.2861
+[04:58:08][     66s][FT] round 1/6 epoch 2/5 | patches=5 | loss=1.2453
+[04:58:08][     66s][FT] round 1/6 epoch 3/5 | patches=5 | loss=1.1900
+[04:58:09][     67s][FT] round 1/6 epoch 4/5 | patches=5 | loss=1.1678
+[04:58:09][     67s][FT] round 1/6 epoch 5/5 | patches=5 | loss=1.1580
+[04:58:10][     68s][AL] round 1/6 | labeled=5 | selection_seconds=5.2000 round_seconds=23.5000 restart_student_dice=0.1222 restart_teacher_dice=0.3119 val_dice=0.2709 val_vi=0.3049 val_ece=0.0198
+[04:58:11][     69s][AL] round 2: 10/681 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:58:12][     70s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.186, uniform=0.049, top1=0.197, top2=0.186, top3=0.186 -> using top1
+[04:58:12][     70s][WEIGHTS] round 2: teacher Dice on 10 labeled patches + 0 blank (18558 fg px): fly-mito=0.197, kasthuri=0.181, lucchi=0.000, multiclass=0.002 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:58:14][     72s][CONS] round 2: on the 5 new patches student Dice=0.108 vs ensemble 0.085 -> pseudo-labels from the student | consistency x0.00
+[04:58:14][     72s][FT] round 2/6 epoch 1/5 | patches=10 | loss=1.0486
+[04:58:14][     72s][FT] round 2/6 epoch 2/5 | patches=10 | loss=1.0370
+[04:58:15][     73s][FT] round 2/6 epoch 3/5 | patches=10 | loss=1.0217
+[04:58:15][     73s][FT] round 2/6 epoch 4/5 | patches=10 | loss=0.9820
+[04:58:16][     74s][FT] round 2/6 epoch 5/5 | patches=10 | loss=0.9480
+[04:58:16][     74s][AL] round 2/6 | labeled=10 | selection_seconds=1.3000 round_seconds=6.0000 pseudo_student_dice=0.1078 pseudo_ensemble_dice=0.0846 consistency_scale=0.0000 val_dice=0.3910 val_vi=0.2760 val_ece=0.0339
+[04:58:17][     75s][AL] round 3: 72/676 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:58:26][     84s][WEIGHTS] round 3: teacher agreement with the student on 671 unlabeled patches: fly-mito=0.542, kasthuri=0.514, lucchi=0.004, multiclass=0.000 | blended with labeled Dice at 0.50
+[04:58:34][     92s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.340, uniform=0.047, top1=0.364, top2=0.340, top3=0.340 -> using top1
+[04:58:34][     92s][WEIGHTS] round 3: teacher Dice on 15 labeled patches + 0 blank (24653 fg px): fly-mito=0.364, kasthuri=0.332, lucchi=0.002, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:58:36][     94s][CONS] round 3: on the 5 new patches student Dice=0.221 vs ensemble 0.166 -> pseudo-labels from the student | consistency x0.00
+[04:58:36][     94s][FT] round 3/6 epoch 1/5 | patches=15 | loss=0.9594
+[04:58:37][     95s][FT] round 3/6 epoch 2/5 | patches=15 | loss=0.9165
+[04:58:37][     95s][FT] round 3/6 epoch 3/5 | patches=15 | loss=0.8829
+[04:58:38][     96s][FT] round 3/6 epoch 4/5 | patches=15 | loss=0.8287
+[04:58:38][     96s][FT] round 3/6 epoch 5/5 | patches=15 | loss=0.7922
+[04:58:38][     96s][AL] round 3/6 | labeled=15 | selection_seconds=1.3000 round_seconds=22.0000 pseudo_student_dice=0.2206 pseudo_ensemble_dice=0.1655 consistency_scale=0.0000 val_dice=0.4794 val_vi=0.2113 val_ece=0.0096
+[04:58:40][     98s][AL] round 4: 91/671 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:58:49][    107s][WEIGHTS] round 4: teacher agreement with the student on 666 unlabeled patches: fly-mito=0.488, kasthuri=0.518, lucchi=0.005, multiclass=0.000 | blended with labeled Dice at 0.50
+[04:58:57][    115s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.336, uniform=0.042, top1=0.339, top2=0.336, top3=0.336 -> using mixture
+[04:58:57][    115s][WEIGHTS] round 4: teacher Dice on 20 labeled patches + 0 blank (39174 fg px): fly-mito=0.339, kasthuri=0.332, lucchi=0.003, multiclass=0.001 | weights: fly-mito=0.130, kasthuri=0.868, lucchi=0.001, multiclass=0.001
+[04:58:59][    117s][CONS] round 4: on the 5 new patches student Dice=0.266 vs ensemble 0.144 -> pseudo-labels from the student | consistency x0.00
+[04:58:59][    117s][FT] round 4/6 epoch 1/5 | patches=20 | loss=0.7862
+[04:59:00][    118s][FT] round 4/6 epoch 2/5 | patches=20 | loss=0.7593
+[04:59:00][    118s][FT] round 4/6 epoch 3/5 | patches=20 | loss=0.6879
+[04:59:01][    119s][FT] round 4/6 epoch 4/5 | patches=20 | loss=0.6546
+[04:59:01][    119s][FT] round 4/6 epoch 5/5 | patches=20 | loss=0.6581
+[04:59:02][    120s][AL] round 4/6 | labeled=20 | selection_seconds=1.3000 round_seconds=22.6000 pseudo_student_dice=0.2664 pseudo_ensemble_dice=0.1440 consistency_scale=0.0000 val_dice=0.5896 val_vi=0.1470 val_ece=0.0053
+[04:59:03][    121s][AL] round 5: 110/666 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:59:12][    130s][WEIGHTS] round 5: teacher agreement with the student on 661 unlabeled patches: fly-mito=0.377, kasthuri=0.468, lucchi=0.009, multiclass=0.000 | blended with labeled Dice at 0.50
+[04:59:19][    137s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.298, uniform=0.040, top1=0.296, top2=0.298, top3=0.298 -> using mixture
+[04:59:19][    137s][WEIGHTS] round 5: teacher Dice on 25 labeled patches + 0 blank (52837 fg px): fly-mito=0.283, kasthuri=0.296, lucchi=0.005, multiclass=0.001 | weights: fly-mito=0.109, kasthuri=0.888, lucchi=0.001, multiclass=0.001
+[04:59:21][    139s][CONS] round 5: on the 5 new patches student Dice=0.449 vs ensemble 0.043 -> pseudo-labels from the student | consistency x0.00
+[04:59:22][    140s][FT] round 5/6 epoch 1/5 | patches=25 | loss=0.6253
+[04:59:22][    140s][FT] round 5/6 epoch 2/5 | patches=25 | loss=0.5868
+[04:59:23][    141s][FT] round 5/6 epoch 3/5 | patches=25 | loss=0.5319
+[04:59:23][    141s][FT] round 5/6 epoch 4/5 | patches=25 | loss=0.5841
+[04:59:24][    142s][FT] round 5/6 epoch 5/5 | patches=25 | loss=0.5223
+[04:59:24][    142s][AL] round 5/6 | labeled=25 | selection_seconds=1.3000 round_seconds=22.2000 pseudo_student_dice=0.4489 pseudo_ensemble_dice=0.0427 consistency_scale=0.0000 val_dice=0.7094 val_vi=0.1120 val_ece=0.0088
+[04:59:25][    143s][AL] round 6: 158/661 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[04:59:34][    152s][WEIGHTS] round 6: teacher agreement with the student on 656 unlabeled patches: fly-mito=0.350, kasthuri=0.455, lucchi=0.010, multiclass=0.001 | blended with labeled Dice at 0.50
+[04:59:42][    160s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.285, uniform=0.042, top1=0.283, top2=0.285, top3=0.285 -> using mixture
+[04:59:42][    160s][WEIGHTS] round 6: teacher Dice on 30 labeled patches + 0 blank (67074 fg px): fly-mito=0.272, kasthuri=0.283, lucchi=0.006, multiclass=0.001 | weights: fly-mito=0.111, kasthuri=0.886, lucchi=0.002, multiclass=0.001
+[04:59:44][    162s][CONS] round 6: on the 5 new patches student Dice=0.399 vs ensemble 0.059 -> pseudo-labels from the student | consistency x0.00
+[04:59:45][    163s][FT] round 6/6 epoch 1/5 | patches=30 | loss=0.5660
+[04:59:45][    163s][FT] round 6/6 epoch 2/5 | patches=30 | loss=0.6058
+[04:59:46][    164s][FT] round 6/6 epoch 3/5 | patches=30 | loss=0.5576
+[04:59:46][    164s][FT] round 6/6 epoch 4/5 | patches=30 | loss=0.5193
+[04:59:47][    165s][FT] round 6/6 epoch 5/5 | patches=30 | loss=0.5235
+[04:59:47][    165s][AL] round 6/6 | labeled=30 | selection_seconds=1.2000 round_seconds=22.6000 pseudo_student_dice=0.3987 pseudo_ensemble_dice=0.0586 consistency_scale=0.0000 val_dice=0.6698 val_vi=0.1306 val_ece=0.0119
+[04:59:48][    166s][CONS] final consolidation: pseudo-labels refreshed from the student
+[04:59:49][    167s][FT] final 1/10 epoch 1/1 | patches=30 | loss=0.5154
+[04:59:49][    167s][FT] final 2/10 epoch 1/1 | patches=30 | loss=0.4676
+[04:59:50][    168s][FT] final 3/10 epoch 1/1 | patches=30 | loss=0.4743
+[04:59:50][    168s][FT] final 4/10 epoch 1/1 | patches=30 | loss=0.4537
+[04:59:51][    169s][FT] final 5/10 epoch 1/1 | patches=30 | loss=0.4082
+[04:59:51][    169s][FT] final 6/10 epoch 1/1 | patches=30 | loss=0.4052
+[04:59:52][    170s][FT] final 7/10 epoch 1/1 | patches=30 | loss=0.4700
+[04:59:52][    170s][FT] final 8/10 epoch 1/1 | patches=30 | loss=0.4553
+[04:59:53][    171s][FT] final 9/10 epoch 1/1 | patches=30 | loss=0.3955
+[04:59:53][    171s][FT] final 10/10 epoch 1/1 | patches=30 | loss=0.4500
+[04:59:54][    172s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=6.2000 val_dice=0.7056 val_vi=0.1168 val_ece=0.0112
+[04:59:54][    172s][WEIGHTS] final: fly-mito=0.111, kasthuri=0.886, lucchi=0.002, multiclass=0.001
+[04:59:56][    174s][TEST] raw student (no EMA, no TTA): patch_dice=0.8127 patch_vi=0.1193
+[04:59:56][    174s][TEST] final model without TTA: patch_dice=0.8127 patch_vi=0.1193
+[04:59:56][    174s][RESOURCES] {"phases": {"setup": {"seconds": 0.61, "peak_cpu_rss_mb": 1026.1, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 5.97, "peak_cpu_rss_mb": 1806.7, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 36.39, "peak_cpu_rss_mb": 2105.8, "peak_gpu_alloc_mb": 3829.8, "peak_gpu_reserved_mb": 4168.0, "kd_epochs": 5}, "active_learning": {"seconds": 128.19, "peak_cpu_rss_mb": 2391.1, "peak_gpu_alloc_mb": 3948.5, "peak_gpu_reserved_mb": 4176.0, "rounds": 6}, "inference": {"seconds": 2.58, "peak_cpu_rss_mb": 2391.1, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4176.0, "patches": 147, "images": 3, "patches_per_second": 57.0, "seconds_per_image": 0.86}}, "peak_gpu_alloc_mb": 3948.5, "peak_cpu_rss_mb": 2391.1, "total_seconds": 173.7}
+[04:59:56][    174s][TEST] patch_dice=0.8160 patch_iou=0.7287 patch_precision=0.7998 patch_recall=0.9001 patch_vi=0.1139 image_dice=0.8812 image_iou=0.7880 image_precision=0.8454 image_recall=0.9222 image_vi=0.1310 cal_ece=0.0042 cal_nll=0.0412 cal_brier=0.0119 cal_mean_confidence=0.9879 cal_accuracy=0.9841 cal_predicted_prevalence=0.0646 cal_true_prevalence=0.0594 cal_mean_probability=0.0660
+[04:59:57][    175s][DONE] drosophila-vnc macs-v11-norestart-acq-margin b5 s1 | test patch Dice=0.8160 VI=0.1139 | image Dice=0.8812 VI=0.1310 | 174s

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-1/result.json

@@ -0,0 +1,667 @@
+{
+  "target": "drosophila-vnc",
+  "variant": "macs-v11-norestart-acq-margin",
+  "budget": 5.0,
+  "seed": 1,
+  "config": {
+    "task": "mitochondria",
+    "target": "drosophila-vnc",
+    "variant": "macs-v11-norestart-acq-margin",
+    "budget": 5.0,
+    "seed": 1,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "margin",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "drosophila-vnc",
+      "--variants",
+      "macs-v11-norestart-acq-margin",
+      "--budgets",
+      "5",
+      "--seeds",
+      "1",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--task",
+      "mitochondria",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "fly-brain-mito-benchmark",
+    "kasthuri-pp",
+    "lucchi-pp",
+    "multiclass"
+  ],
+  "alphas": [
+    0.11690531690536884,
+    0.8375898290916663,
+    0.02560239431814963,
+    0.01990245968481518
+  ],
+  "weights_all": {
+    "sources": [
+      "fly-brain-mito-benchmark",
+      "kasthuri-pp",
+      "lucchi-pp",
+      "multiclass"
+    ],
+    "mixture": [
+      0.11690531690536884,
+      0.8375898290916663,
+      0.02560239431814963,
+      0.01990245968481518
+    ],
+    "per_source": [
+      0.038599381604120485,
+      0.943285983973492,
+      0.013712262708195049,
+      0.004402371714192633
+    ],
+    "uniform": [
+      0.25,
+      0.25,
+      0.25,
+      0.25
+    ],
+    "mixture_info": {
+      "objective": 0.1917814945701421,
+      "mmd2_mixture": 0.19749308036725388,
+      "mmd2_uniform": 0.3743480458439246,
+      "mmd2_single_source": [
+        0.7042945447726864,
+        0.2070680114579429,
+        0.6307945959829456,
+        0.6534353423959632
+      ],
+      "median_squared_distance": 0.1964580928585633,
+      "gammas": [
+        1.2725360221224549,
+        2.5450720442449097,
+        5.0901440884898195,
+        10.180288176979639,
+        20.360576353959278
+      ],
+      "max_weight": 0.8375898290916663,
+      "entropy": 0.5711585797111782,
+      "effective_sources": 1.7703169170005038,
+      "min_weight": 0.01990245968481518
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.6050756121533214,
+        0.3110169227475098,
+        0.7002952924749369,
+        0.8048259551971948
+      ],
+      "standardized": [
+        -0.0012381647254944845,
+        -1.5993047903844642,
+        0.5162347028009382,
+        1.0843082523090217
+      ],
+      "tau": 0.5,
+      "max_weight": 0.943285983973492,
+      "entropy": 0.2634007298678481,
+      "effective_sources": 1.3013481035311127,
+      "min_weight": 0.004402371714192633
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "kasthuri-pp",
+    "seconds": 1.1
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.2784035818310388,
+      0.18341782772166676,
+      0.057702367443012846,
+      0.01681466600081078
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.186522633793688,
+      "mixture@raw": 0.18548325188563394,
+      "mixture@logit": 0.19296315829990748,
+      "per_source": 0.18415887092754818,
+      "per_source@raw": 0.18350562678267812,
+      "per_source@logit": 0.1890579186582894,
+      "uniform": 0.06967222752121038,
+      "uniform@raw": 0.0884686387663988,
+      "uniform@logit": 0.06545324142703345
+    },
+    "val_patches": 147
+  },
+  "temperatures": [
+    1.3819772005081177,
+    2.5080416202545166,
+    3.7969236373901367,
+    2.597255229949951
+  ],
+  "top_source": "kasthuri-pp",
+  "weight_stats": {
+    "max_weight": 0.8375898290916663,
+    "entropy": 0.5711585797111782,
+    "effective_sources": 1.7703169170005038,
+    "min_weight": 0.01990245968481518
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.7999999999999999,
+    "spearman_weight_taxonomy": -0.21081851067789195,
+    "top_source": "kasthuri-pp",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "kasthuri-pp",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 36.4,
+  "round0_val": {
+    "val_dice": 0.19168290697357765,
+    "val_vi": 0.22123099590519313,
+    "val_ece": 0.02235104280276007
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 5,
+      "selection_seconds": 5.2,
+      "round_seconds": 23.5,
+      "teacher_dice_labeled": [
+        0.3118739095489208,
+        0.0968435913812229,
+        0.0004512053628980276,
+        0.00288
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 15368,
+      "evidence_nonempty_patches": 5,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.12218059783032857,
+      "restart_teacher_dice": 0.3118739095489208,
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.3032701474337657,
+        "uniform": 0.018090264826557255,
+        "top1": 0.3118739095489208,
+        "top2": 0.3048324685603729,
+        "top3": 0.3042586054721977
+      },
+      "val_dice": 0.27092069976166006,
+      "val_vi": 0.3048773627261147,
+      "val_ece": 0.019770819577808685
+    },
+    {
+      "round": 2,
+      "n_labeled": 10,
+      "selection_seconds": 1.3,
+      "round_seconds": 6.0,
+      "teacher_dice_labeled": [
+        0.19717803802753986,
+        0.1805899733130004,
+        0.0003737319807794981,
+        0.0023752969121140144
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 18558,
+      "evidence_nonempty_patches": 8,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.18626228722128985,
+        "uniform": 0.049455026993990014,
+        "top1": 0.19717803802753986,
+        "top2": 0.18620293462910215,
+        "top3": 0.18620970445945298
+      },
+      "pseudo_student_dice": 0.10775674024954192,
+      "pseudo_ensemble_dice": 0.0846065037521647,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.39099553159399997,
+      "val_vi": 0.27599488257770194,
+      "val_ece": 0.03385775081323113
+    },
+    {
+      "round": 3,
+      "n_labeled": 15,
+      "selection_seconds": 1.3,
+      "round_seconds": 22.0,
+      "teacher_dice_labeled": [
+        0.3640246067431969,
+        0.3316745838639298,
+        0.0019141719919512118,
+        0.0010269183279976834
+      ],
+      "teacher_dice_pool": [
+        0.5423140882497323,
+        0.5143849788839585,
+        0.003546472356295111,
+        0.0002619360467491595
+      ],
+      "evidence_fg_pixels": 24653,
+      "evidence_nonempty_patches": 11,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.3395188668558075,
+        "uniform": 0.04689973786647725,
+        "top1": 0.3640246067431969,
+        "top2": 0.3395821856889235,
+        "top3": 0.33954572307904063
+      },
+      "pseudo_student_dice": 0.22056063804832277,
+      "pseudo_ensemble_dice": 0.16550525645658046,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.47937634338993435,
+      "val_vi": 0.21128977939475435,
+      "val_ece": 0.009649798385442703
+    },
+    {
+      "round": 4,
+      "n_labeled": 20,
+      "selection_seconds": 1.3,
+      "round_seconds": 22.6,
+      "teacher_dice_labeled": [
+        0.33894606300963487,
+        0.33203760786110254,
+        0.003406787294381386,
+        0.0008236433469535518
+      ],
+      "teacher_dice_pool": [
+        0.4882980688802892,
+        0.5177185039317563,
+        0.005067358057912947,
+        0.0002118076972313708
+      ],
+      "evidence_fg_pixels": 39174,
+      "evidence_nonempty_patches": 16,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.12987252082567358,
+        0.8683830988706824,
+        0.0009925103168831313,
+        0.0007518699867609326
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.33624660092618736,
+        "uniform": 0.04154563567677867,
+        "top1": 0.33894606300963487,
+        "top2": 0.33631352136850234,
+        "top3": 0.33625433868249766
+      },
+      "pseudo_student_dice": 0.2663955078828018,
+      "pseudo_ensemble_dice": 0.14395372526193248,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.5895830169053273,
+      "val_vi": 0.14701486319551668,
+      "val_ece": 0.005259925208484618
+    },
+    {
+      "round": 5,
+      "n_labeled": 25,
+      "selection_seconds": 1.3,
+      "round_seconds": 22.2,
+      "teacher_dice_labeled": [
+        0.28347866835979546,
+        0.29629355806286156,
+        0.005265936456685989,
+        0.0011409426433959182
+      ],
+      "teacher_dice_pool": [
+        0.37725700020977354,
+        0.46814559181560483,
+        0.009089817328683986,
+        0.0002059936819967136
+      ],
+      "evidence_fg_pixels": 52837,
+      "evidence_nonempty_patches": 20,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.10907589581677156,
+        0.8883421701062245,
+        0.0014788106325029009,
+        0.0011031234445009884
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.2981801868826781,
+        "uniform": 0.040404132361419064,
+        "top1": 0.29629355806286156,
+        "top2": 0.29825831529887437,
+        "top3": 0.2982222264110502
+      },
+      "pseudo_student_dice": 0.4488768227000044,
+      "pseudo_ensemble_dice": 0.042720602432902106,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.7094354984789913,
+      "val_vi": 0.11197468489454974,
+      "val_ece": 0.008760546333009399
+    },
+    {
+      "round": 6,
+      "n_labeled": 30,
+      "selection_seconds": 1.2,
+      "round_seconds": 22.6,
+      "teacher_dice_labeled": [
+        0.2724591555035009,
+        0.2829514427188626,
+        0.005805749140485103,
+        0.0011238169277639535
+      ],
+      "teacher_dice_pool": [
+        0.35042379468295565,
+        0.45451658602563566,
+        0.01047669215596468,
+        0.0006093541967623729
+      ],
+      "evidence_fg_pixels": 67074,
+      "evidence_nonempty_patches": 23,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.11131163266174687,
+        0.8857376563778061,
+        0.0016940485074520217,
+        0.0012566624529949536
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.284854314782343,
+        "uniform": 0.041772513841434294,
+        "top1": 0.2829514427188626,
+        "top2": 0.284905995057746,
+        "top3": 0.28487357388139867
+      },
+      "pseudo_student_dice": 0.39869975876027314,
+      "pseudo_ensemble_dice": 0.05862639635912288,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.6698150851720401,
+      "val_vi": 0.1305862633741253,
+      "val_ece": 0.011866765588902227
+    },
+    {
+      "round": 7,
+      "n_labeled": 30,
+      "selection_seconds": 0.0,
+      "round_seconds": 6.2,
+      "final_consolidation": true,
+      "val_dice": 0.7055667561203255,
+      "val_vi": 0.11683635188920495,
+      "val_ece": 0.011193001351058354
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": 0.08899991394046061,
+    "auroc_top20_error": 0.5872474346768222
+  },
+  "al_seconds": 128.2,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.11131163266174687,
+    0.8857376563778061,
+    0.0016940485074520217,
+    0.0012566624529949536
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.8857376563778061,
+    "entropy": 0.37104918671331083,
+    "effective_sources": 1.449254355692067,
+    "min_weight": 0.0012566624529949536
+  },
+  "top_source_final": "kasthuri-pp",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.812722205200734,
+    "patch_iou": 0.725063429876011,
+    "patch_precision": 0.7900888338957477,
+    "patch_recall": 0.905797142842391,
+    "patch_vi": 0.11934194083093992,
+    "n_patches": 147,
+    "cal_ece": 0.01095954651793581,
+    "cal_nll": 0.05754046776538521,
+    "cal_brier": 0.013845200752527742,
+    "cal_mean_confidence": 0.994078189838488,
+    "cal_accuracy": 0.9832456420068028,
+    "cal_predicted_prevalence": 0.06584821428571429,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06584790728883756
+  },
+  "test_notta": {
+    "patch_dice": 0.812722205200734,
+    "patch_iou": 0.725063429876011,
+    "patch_precision": 0.7900888338957477,
+    "patch_recall": 0.905797142842391,
+    "patch_vi": 0.11934194083093992,
+    "n_patches": 147,
+    "cal_ece": 0.01095954651793581,
+    "cal_nll": 0.05754046776538521,
+    "cal_brier": 0.013845200752527742,
+    "cal_mean_confidence": 0.994078189838488,
+    "cal_accuracy": 0.9832456420068028,
+    "cal_predicted_prevalence": 0.06584821428571429,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06584790728883756
+  },
+  "test": {
+    "patch_dice": 0.8160149021553218,
+    "patch_iou": 0.728680459250354,
+    "patch_precision": 0.7997959914735588,
+    "patch_recall": 0.9001494344816307,
+    "patch_vi": 0.11392365994158125,
+    "n_patches": 147,
+    "image_dice": 0.8812000396184795,
+    "image_iou": 0.7880471247693754,
+    "image_precision": 0.8454001509224097,
+    "image_recall": 0.9222251231022218,
+    "image_vi": 0.1309689113182238,
+    "n_images": 3,
+    "cal_ece": 0.004152246751486293,
+    "cal_nll": 0.04119385390428756,
+    "cal_brier": 0.011927609300364279,
+    "cal_mean_confidence": 0.987949771162949,
+    "cal_accuracy": 0.9841225552721088,
+    "cal_predicted_prevalence": 0.06462585034013606,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06603734781981793
+  },
+  "eval_seconds": 2.6,
+  "total_seconds": 174.4,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 0.61,
+        "peak_cpu_rss_mb": 1026.1,
+        "peak_gpu_alloc_mb": 592.4,
+        "peak_gpu_reserved_mb": 594.0
+      },
+      "features_and_weights": {
+        "seconds": 5.97,
+        "peak_cpu_rss_mb": 1806.7,
+        "peak_gpu_alloc_mb": 2298.2,
+        "peak_gpu_reserved_mb": 2640.0
+      },
+      "distillation": {
+        "seconds": 36.39,
+        "peak_cpu_rss_mb": 2105.8,
+        "peak_gpu_alloc_mb": 3829.8,
+        "peak_gpu_reserved_mb": 4168.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 128.19,
+        "peak_cpu_rss_mb": 2391.1,
+        "peak_gpu_alloc_mb": 3948.5,
+        "peak_gpu_reserved_mb": 4176.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 2.58,
+        "peak_cpu_rss_mb": 2391.1,
+        "peak_gpu_alloc_mb": 2149.5,
+        "peak_gpu_reserved_mb": 4176.0,
+        "patches": 147,
+        "images": 3,
+        "patches_per_second": 57.0,
+        "seconds_per_image": 0.86
+      }
+    },
+    "peak_gpu_alloc_mb": 3948.5,
+    "peak_cpu_rss_mb": 2391.1,
+    "total_seconds": 173.7
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/drosophila-vnc/macs-v11-norestart-acq-margin_b5_s1.pth"
+}

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-1/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,5,5.2,23.5,"[0.3118739095489208, 0.0968435913812229, 0.0004512053628980276, 0.00288]",,15368,5,True,kept,0.12218059783032857,0.3118739095489208,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.3032701474337657, 'uniform': 0.018090264826557255, 'top1': 0.3118739095489208, 'top2': 0.3048324685603729, 'top3': 0.3042586054721977}",0.27092069976166006,0.3048773627261147,0.019770819577808685,,,,,
+2,10,1.3,6.0,"[0.19717803802753986, 0.1805899733130004, 0.0003737319807794981, 0.0023752969121140144]",,18558,8,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.18626228722128985, 'uniform': 0.049455026993990014, 'top1': 0.19717803802753986, 'top2': 0.18620293462910215, 'top3': 0.18620970445945298}",0.39099553159399997,0.27599488257770194,0.03385775081323113,0.10775674024954192,0.0846065037521647,student,0.0,
+3,15,1.3,22.0,"[0.3640246067431969, 0.3316745838639298, 0.0019141719919512118, 0.0010269183279976834]","[0.5423140882497323, 0.5143849788839585, 0.003546472356295111, 0.0002619360467491595]",24653,11,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.3395188668558075, 'uniform': 0.04689973786647725, 'top1': 0.3640246067431969, 'top2': 0.3395821856889235, 'top3': 0.33954572307904063}",0.47937634338993435,0.21128977939475435,0.009649798385442703,0.22056063804832277,0.16550525645658046,student,0.0,
+4,20,1.3,22.6,"[0.33894606300963487, 0.33203760786110254, 0.003406787294381386, 0.0008236433469535518]","[0.4882980688802892, 0.5177185039317563, 0.005067358057912947, 0.0002118076972313708]",39174,16,False,,,,"[0.12987252082567358, 0.8683830988706824, 0.0009925103168831313, 0.0007518699867609326]",mixture,"{'mixture': 0.33624660092618736, 'uniform': 0.04154563567677867, 'top1': 0.33894606300963487, 'top2': 0.33631352136850234, 'top3': 0.33625433868249766}",0.5895830169053273,0.14701486319551668,0.005259925208484618,0.2663955078828018,0.14395372526193248,student,0.0,
+5,25,1.3,22.2,"[0.28347866835979546, 0.29629355806286156, 0.005265936456685989, 0.0011409426433959182]","[0.37725700020977354, 0.46814559181560483, 0.009089817328683986, 0.0002059936819967136]",52837,20,False,,,,"[0.10907589581677156, 0.8883421701062245, 0.0014788106325029009, 0.0011031234445009884]",mixture,"{'mixture': 0.2981801868826781, 'uniform': 0.040404132361419064, 'top1': 0.29629355806286156, 'top2': 0.29825831529887437, 'top3': 0.2982222264110502}",0.7094354984789913,0.11197468489454974,0.008760546333009399,0.4488768227000044,0.042720602432902106,student,0.0,
+6,30,1.2,22.6,"[0.2724591555035009, 0.2829514427188626, 0.005805749140485103, 0.0011238169277639535]","[0.35042379468295565, 0.45451658602563566, 0.01047669215596468, 0.0006093541967623729]",67074,23,False,,,,"[0.11131163266174687, 0.8857376563778061, 0.0016940485074520217, 0.0012566624529949536]",mixture,"{'mixture': 0.284854314782343, 'uniform': 0.041772513841434294, 'top1': 0.2829514427188626, 'top2': 0.284905995057746, 'top3': 0.28487357388139867}",0.6698150851720401,0.1305862633741253,0.011866765588902227,0.39869975876027314,0.05862639635912288,student,0.0,
+7,30,0.0,6.2,,,,,,,,,,,,0.7055667561203255,0.11683635188920495,0.011193001351058354,,,,,True

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt

@@ -0,0 +1,103 @@
+[04:25:06][      0s][START] target=drosophila-vnc variant=macs-v11-norestart-acq-random budget=5.0% seed=1 | {"task": "mitochondria", "target": "drosophila-vnc", "variant": "macs-v11-norestart-acq-random", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "random", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 0, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[04:25:06][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[04:25:06][      0s][DATA] drosophila-vnc: train=686 val=147 test=147 patches
+[04:25:08][      2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, multiclass=2.60
+[04:25:09][      2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s
+[04:25:10][      3s][WEIGHTS] mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020 | eff.sources=1.77 | MMD2 mix=0.1975 uniform=0.3743
+[04:25:10][      3s][WEIGHTS] per-source: fly-mito=0.039, kasthuri=0.943, lucchi=0.014, multiclass=0.004 | min-MMD source (NeuroADDA rule) = kasthuri
+[04:25:13][      7s][TEACHERS] target-val Dice: fly-mito=0.278, kasthuri=0.183, lucchi=0.058, multiclass=0.017
+[04:25:13][      7s][TEACHERS] ensemble target-val Dice: mixture=0.187, mixture@raw=0.185, mixture@logit=0.193, per_source=0.184, per_source@raw=0.184, per_source@logit=0.189, uniform=0.070, uniform@raw=0.088, uniform@logit=0.065
+[04:25:13][      7s][WEIGHTS] using mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020
+[04:25:13][      7s][INIT] student initialised from teacher kasthuri (weight 0.838)
+[04:25:15][      8s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[04:25:15][      8s][KD] distilling the mixture ensemble (prob space) into the student
+[04:25:19][     12s][KD] soft labels for 686 patches from 4 teachers in 4s
+[04:25:25][     19s][KD] epoch 1/5 | patches=686 | loss=0.2005 | 6s
+[04:25:32][     25s][KD] epoch 2/5 | patches=686 | loss=0.1322 | 6s
+[04:25:38][     32s][KD] epoch 3/5 | patches=686 | loss=0.1258 | 6s
+[04:25:44][     38s][KD] epoch 4/5 | patches=686 | loss=0.1216 | 6s
+[04:25:51][     44s][KD] epoch 5/5 | patches=686 | loss=0.1204 | 6s
+[04:25:51][     45s][ROUND0] before any target label: val_dice=0.1917 val_vi=0.2213
+[04:25:51][     45s][AL] patches=686 (pool 686 after dropping blank; free blank negatives: True) | budget=34 (5.0%) | rounds=6 | per_round=5 | acquisition=random | first_round=stratified
+[04:25:52][     46s][AL] round-1 uncertainty quality on 686 patches: {'spearman_uncertainty_error': -0.03470887967342099, 'auroc_top20_error': 0.482987650099299}
+[04:25:56][     50s][AL] round 1: stratified cold start: 3 patches from the membrane-rich half (ensemble prevalence >= 0.047) + 2 from the membrane-poor half of 686 candidates
+[04:25:57][     50s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.303, uniform=0.018, top1=0.312, top2=0.305, top3=0.304 -> using top1
+[04:25:57][     51s][WEIGHTS] round 1: student kept (labeled Dice 0.122 vs best teacher fly-mito 0.312); distilling the corrected ensemble into it
+[04:25:57][     51s][WEIGHTS] round 1: teacher Dice on 5 labeled patches + 0 blank (15368 fg px): fly-mito=0.312, kasthuri=0.097, lucchi=0.000, multiclass=0.003 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:25:58][     52s][KD] soft labels for 681 patches from 4 teachers in 1s
+[04:26:05][     58s][KD] epoch 1/2 | patches=681 | loss=0.2886 | 7s
+[04:26:11][     65s][KD] epoch 2/2 | patches=681 | loss=0.2383 | 6s
+[04:26:12][     66s][CONS] pseudo-labels for 681 unlabeled patches from the current ensemble in 1s
+[04:26:13][     66s][FT] round 1/6 epoch 1/5 | patches=5 | loss=1.2861
+[04:26:13][     67s][FT] round 1/6 epoch 2/5 | patches=5 | loss=1.2453
+[04:26:13][     67s][FT] round 1/6 epoch 3/5 | patches=5 | loss=1.1900
+[04:26:14][     67s][FT] round 1/6 epoch 4/5 | patches=5 | loss=1.1676
+[04:26:14][     68s][FT] round 1/6 epoch 5/5 | patches=5 | loss=1.1582
+[04:26:15][     68s][AL] round 1/6 | labeled=5 | selection_seconds=5.3000 round_seconds=23.3000 restart_student_dice=0.1221 restart_teacher_dice=0.3119 val_dice=0.2704 val_vi=0.3045 val_ece=0.0198
+[04:26:16][     69s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.228, uniform=0.029, top1=0.337, top2=0.229, top3=0.229 -> using top1
+[04:26:16][     69s][WEIGHTS] round 2: teacher Dice on 10 labeled patches + 0 blank (29992 fg px): fly-mito=0.337, kasthuri=0.208, lucchi=0.000, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:26:17][     71s][CONS] round 2: on the 5 new patches student Dice=0.402 vs ensemble 0.361 -> pseudo-labels from the student | consistency x0.00
+[04:26:18][     71s][FT] round 2/6 epoch 1/5 | patches=10 | loss=0.9670
+[04:26:18][     72s][FT] round 2/6 epoch 2/5 | patches=10 | loss=0.9125
+[04:26:19][     72s][FT] round 2/6 epoch 3/5 | patches=10 | loss=0.8827
+[04:26:19][     73s][FT] round 2/6 epoch 4/5 | patches=10 | loss=0.8156
+[04:26:19][     73s][FT] round 2/6 epoch 5/5 | patches=10 | loss=0.7817
+[04:26:20][     74s][AL] round 2/6 | labeled=10 | selection_seconds=0.1000 round_seconds=4.7000 pseudo_student_dice=0.4019 pseudo_ensemble_dice=0.3610 consistency_scale=0.0000 val_dice=0.3955 val_vi=0.2676 val_ece=0.0101
+[04:26:29][     83s][WEIGHTS] round 3: teacher agreement with the student on 671 unlabeled patches: fly-mito=0.583, kasthuri=0.506, lucchi=0.004, multiclass=0.000 | blended with labeled Dice at 0.50
+[04:26:37][     91s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.379, uniform=0.031, top1=0.478, top2=0.380, top3=0.379 -> using top1
+[04:26:37][     91s][WEIGHTS] round 3: teacher Dice on 15 labeled patches + 0 blank (53089 fg px): fly-mito=0.478, kasthuri=0.362, lucchi=0.002, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:26:39][     93s][CONS] round 3: on the 5 new patches student Dice=0.659 vs ensemble 0.427 -> pseudo-labels from the student | consistency x0.53
+[04:26:39][     93s][FT] round 3/6 epoch 1/5 | patches=15 | loss=0.7948
+[04:26:40][     94s][FT] round 3/6 epoch 2/5 | patches=15 | loss=0.7941
+[04:26:41][     94s][FT] round 3/6 epoch 3/5 | patches=15 | loss=0.7283
+[04:26:41][     95s][FT] round 3/6 epoch 4/5 | patches=15 | loss=0.7077
+[04:26:42][     96s][FT] round 3/6 epoch 5/5 | patches=15 | loss=0.6730
+[04:26:42][     96s][AL] round 3/6 | labeled=15 | selection_seconds=0.1000 round_seconds=22.0000 pseudo_student_dice=0.6589 pseudo_ensemble_dice=0.4271 consistency_scale=0.5300 val_dice=0.5028 val_vi=0.1864 val_ece=0.0084
+[04:26:52][    105s][WEIGHTS] round 4: teacher agreement with the student on 666 unlabeled patches: fly-mito=0.484, kasthuri=0.500, lucchi=0.004, multiclass=0.000 | blended with labeled Dice at 0.50
+[04:26:59][    113s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.373, uniform=0.035, top1=0.431, top2=0.373, top3=0.373 -> using top1
+[04:26:59][    113s][WEIGHTS] round 4: teacher Dice on 20 labeled patches + 0 blank (74091 fg px): fly-mito=0.431, kasthuri=0.363, lucchi=0.006, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000
+[04:27:01][    115s][CONS] round 4: on the 5 new patches student Dice=0.649 vs ensemble 0.391 -> pseudo-labels from the student | consistency x0.50
+[04:27:02][    116s][FT] round 4/6 epoch 1/5 | patches=20 | loss=0.5785
+[04:27:03][    116s][FT] round 4/6 epoch 2/5 | patches=20 | loss=0.5677
+[04:27:04][    117s][FT] round 4/6 epoch 3/5 | patches=20 | loss=0.4914
+[04:27:04][    118s][FT] round 4/6 epoch 4/5 | patches=20 | loss=0.5238
+[04:27:05][    119s][FT] round 4/6 epoch 5/5 | patches=20 | loss=0.4591
+[04:27:05][    119s][AL] round 4/6 | labeled=20 | selection_seconds=0.1000 round_seconds=22.6000 pseudo_student_dice=0.6493 pseudo_ensemble_dice=0.3907 consistency_scale=0.4980 val_dice=0.6575 val_vi=0.1353 val_ece=0.0086
+[04:27:14][    128s][WEIGHTS] round 5: teacher agreement with the student on 661 unlabeled patches: fly-mito=0.371, kasthuri=0.460, lucchi=0.009, multiclass=0.001 | blended with labeled Dice at 0.50
+[04:27:22][    136s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.385, uniform=0.054, top1=0.382, top2=0.385, top3=0.385 -> using mixture
+[04:27:22][    136s][WEIGHTS] round 5: teacher Dice on 25 labeled patches + 0 blank (90327 fg px): fly-mito=0.372, kasthuri=0.382, lucchi=0.008, multiclass=0.004 | weights: fly-mito=0.111, kasthuri=0.888, lucchi=0.001, multiclass=0.000
+[04:27:24][    137s][CONS] round 5: on the 5 new patches student Dice=0.829 vs ensemble 0.593 -> pseudo-labels from the student | consistency x1.00
+[04:27:25][    138s][FT] round 5/6 epoch 1/5 | patches=25 | loss=0.4421
+[04:27:25][    139s][FT] round 5/6 epoch 2/5 | patches=25 | loss=0.3746
+[04:27:26][    140s][FT] round 5/6 epoch 3/5 | patches=25 | loss=0.3486
+[04:27:27][    141s][FT] round 5/6 epoch 4/5 | patches=25 | loss=0.3556
+[04:27:28][    142s][FT] round 5/6 epoch 5/5 | patches=25 | loss=0.3895
+[04:27:29][    142s][AL] round 5/6 | labeled=25 | selection_seconds=0.1000 round_seconds=22.6000 pseudo_student_dice=0.8290 pseudo_ensemble_dice=0.5930 consistency_scale=1.0000 val_dice=0.6519 val_vi=0.1272 val_ece=0.0108
+[04:27:38][    151s][WEIGHTS] round 6: teacher agreement with the student on 656 unlabeled patches: fly-mito=0.399, kasthuri=0.472, lucchi=0.010, multiclass=0.001 | blended with labeled Dice at 0.50
+[04:27:45][    159s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.376, uniform=0.049, top1=0.376, top2=0.376, top3=0.376 -> using mixture
+[04:27:45][    159s][WEIGHTS] round 6: teacher Dice on 30 labeled patches + 0 blank (102597 fg px): fly-mito=0.376, kasthuri=0.374, lucchi=0.008, multiclass=0.003 | weights: fly-mito=0.125, kasthuri=0.873, lucchi=0.001, multiclass=0.001
+[04:27:47][    161s][CONS] round 6: on the 5 new patches student Dice=0.783 vs ensemble 0.000 -> pseudo-labels from the student | consistency x0.94
+[04:27:48][    162s][FT] round 6/6 epoch 1/5 | patches=30 | loss=0.3306
+[04:27:49][    163s][FT] round 6/6 epoch 2/5 | patches=30 | loss=0.3561
+[04:27:50][    163s][FT] round 6/6 epoch 3/5 | patches=30 | loss=0.3140
+[04:27:51][    164s][FT] round 6/6 epoch 4/5 | patches=30 | loss=0.3094
+[04:27:52][    165s][FT] round 6/6 epoch 5/5 | patches=30 | loss=0.3001
+[04:27:52][    166s][AL] round 6/6 | labeled=30 | selection_seconds=0.1000 round_seconds=23.1000 pseudo_student_dice=0.7834 pseudo_ensemble_dice=0.0001 consistency_scale=0.9450 val_dice=0.6547 val_vi=0.1273 val_ece=0.0109
+[04:27:53][    167s][CONS] final consolidation: pseudo-labels refreshed from the student
+[04:27:54][    168s][FT] final 1/10 epoch 1/1 | patches=30 | loss=0.2706
+[04:27:55][    169s][FT] final 2/10 epoch 1/1 | patches=30 | loss=0.2997
+[04:27:56][    169s][FT] final 3/10 epoch 1/1 | patches=30 | loss=0.3040
+[04:27:57][    170s][FT] final 4/10 epoch 1/1 | patches=30 | loss=0.2837
+[04:27:58][    171s][FT] final 5/10 epoch 1/1 | patches=30 | loss=0.2883
+[04:27:58][    172s][FT] final 6/10 epoch 1/1 | patches=30 | loss=0.2767
+[04:27:59][    173s][FT] final 7/10 epoch 1/1 | patches=30 | loss=0.2819
+[04:28:00][    174s][FT] final 8/10 epoch 1/1 | patches=30 | loss=0.2961
+[04:28:01][    175s][FT] final 9/10 epoch 1/1 | patches=30 | loss=0.2780
+[04:28:02][    176s][FT] final 10/10 epoch 1/1 | patches=30 | loss=0.3077
+[04:28:02][    176s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=9.9000 val_dice=0.6747 val_vi=0.1178 val_ece=0.0097
+[04:28:02][    176s][WEIGHTS] final: fly-mito=0.125, kasthuri=0.873, lucchi=0.001, multiclass=0.001
+[04:28:05][    178s][TEST] raw student (no EMA, no TTA): patch_dice=0.7961 patch_vi=0.1242
+[04:28:05][    179s][TEST] final model without TTA: patch_dice=0.7961 patch_vi=0.1242
+[04:28:05][    179s][RESOURCES] {"phases": {"setup": {"seconds": 1.6, "peak_cpu_rss_mb": 1025.6, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 6.31, "peak_cpu_rss_mb": 1806.9, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 36.04, "peak_cpu_rss_mb": 2105.9, "peak_gpu_alloc_mb": 3829.8, "peak_gpu_reserved_mb": 4168.0, "kd_epochs": 5}, "active_learning": {"seconds": 131.45, "peak_cpu_rss_mb": 2382.2, "peak_gpu_alloc_mb": 3950.0, "peak_gpu_reserved_mb": 4178.0, "rounds": 6}, "inference": {"seconds": 2.73, "peak_cpu_rss_mb": 2382.2, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4178.0, "patches": 147, "images": 3, "patches_per_second": 53.8, "seconds_per_image": 0.91}}, "peak_gpu_alloc_mb": 3950.0, "peak_cpu_rss_mb": 2382.2, "total_seconds": 178.1}
+[04:28:05][    179s][TEST] patch_dice=0.7957 patch_iou=0.7131 patch_precision=0.7525 patch_recall=0.9245 patch_vi=0.1160 image_dice=0.8688 image_iou=0.7684 image_precision=0.7991 image_recall=0.9525 image_vi=0.1422 cal_ece=0.0046 cal_nll=0.0431 cal_brier=0.0125 cal_mean_confidence=0.9878 cal_accuracy=0.9832 cal_predicted_prevalence=0.0705 cal_true_prevalence=0.0594 cal_mean_probability=0.0712
+[04:28:05][    179s][DONE] drosophila-vnc macs-v11-norestart-acq-random b5 s1 | test patch Dice=0.7957 VI=0.1160 | image Dice=0.8688 VI=0.1422 | 179s

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-random/budget-5/seed-1/result.json

@@ -0,0 +1,667 @@
+{
+  "target": "drosophila-vnc",
+  "variant": "macs-v11-norestart-acq-random",
+  "budget": 5.0,
+  "seed": 1,
+  "config": {
+    "task": "mitochondria",
+    "target": "drosophila-vnc",
+    "variant": "macs-v11-norestart-acq-random",
+    "budget": 5.0,
+    "seed": 1,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "random",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 0,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "drosophila-vnc",
+      "--variants",
+      "macs-v11-norestart-acq-random",
+      "--budgets",
+      "5",
+      "--seeds",
+      "1",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--task",
+      "mitochondria",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "fly-brain-mito-benchmark",
+    "kasthuri-pp",
+    "lucchi-pp",
+    "multiclass"
+  ],
+  "alphas": [
+    0.11690531690536884,
+    0.8375898290916663,
+    0.02560239431814963,
+    0.01990245968481518
+  ],
+  "weights_all": {
+    "sources": [
+      "fly-brain-mito-benchmark",
+      "kasthuri-pp",
+      "lucchi-pp",
+      "multiclass"
+    ],
+    "mixture": [
+      0.11690531690536884,
+      0.8375898290916663,
+      0.02560239431814963,
+      0.01990245968481518
+    ],
+    "per_source": [
+      0.038599381604120485,
+      0.943285983973492,
+      0.013712262708195049,
+      0.004402371714192633
+    ],
+    "uniform": [
+      0.25,
+      0.25,
+      0.25,
+      0.25
+    ],
+    "mixture_info": {
+      "objective": 0.1917814945701421,
+      "mmd2_mixture": 0.19749308036725388,
+      "mmd2_uniform": 0.3743480458439246,
+      "mmd2_single_source": [
+        0.7042945447726864,
+        0.2070680114579429,
+        0.6307945959829456,
+        0.6534353423959632
+      ],
+      "median_squared_distance": 0.1964580928585633,
+      "gammas": [
+        1.2725360221224549,
+        2.5450720442449097,
+        5.0901440884898195,
+        10.180288176979639,
+        20.360576353959278
+      ],
+      "max_weight": 0.8375898290916663,
+      "entropy": 0.5711585797111782,
+      "effective_sources": 1.7703169170005038,
+      "min_weight": 0.01990245968481518
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.6050756121533214,
+        0.3110169227475098,
+        0.7002952924749369,
+        0.8048259551971948
+      ],
+      "standardized": [
+        -0.0012381647254944845,
+        -1.5993047903844642,
+        0.5162347028009382,
+        1.0843082523090217
+      ],
+      "tau": 0.5,
+      "max_weight": 0.943285983973492,
+      "entropy": 0.2634007298678481,
+      "effective_sources": 1.3013481035311127,
+      "min_weight": 0.004402371714192633
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "kasthuri-pp",
+    "seconds": 1.4
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.2784035818310388,
+      0.18341782772166676,
+      0.057702367443012846,
+      0.01681466600081078
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.186522633793688,
+      "mixture@raw": 0.18548325188563394,
+      "mixture@logit": 0.19296315829990748,
+      "per_source": 0.18415887092754818,
+      "per_source@raw": 0.18350562678267812,
+      "per_source@logit": 0.1890579186582894,
+      "uniform": 0.06967222752121038,
+      "uniform@raw": 0.0884686387663988,
+      "uniform@logit": 0.06545324142703345
+    },
+    "val_patches": 147
+  },
+  "temperatures": [
+    1.3819772005081177,
+    2.5080416202545166,
+    3.7969236373901367,
+    2.597255229949951
+  ],
+  "top_source": "kasthuri-pp",
+  "weight_stats": {
+    "max_weight": 0.8375898290916663,
+    "entropy": 0.5711585797111782,
+    "effective_sources": 1.7703169170005038,
+    "min_weight": 0.01990245968481518
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.7999999999999999,
+    "spearman_weight_taxonomy": -0.21081851067789195,
+    "top_source": "kasthuri-pp",
+    "top_source_is_best_teacher": false
+  },
+  "init_teacher": "kasthuri-pp",
+  "student_init": "top-source",
+  "blank_fraction": 0.0,
+  "kd_seconds": 36.0,
+  "round0_val": {
+    "val_dice": 0.1916880550436862,
+    "val_vi": 0.22128545106748224,
+    "val_ece": 0.022343801622776858
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 5,
+      "selection_seconds": 5.3,
+      "round_seconds": 23.3,
+      "teacher_dice_labeled": [
+        0.3118739095489208,
+        0.0968435913812229,
+        0.0004512053628980276,
+        0.00288
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 15368,
+      "evidence_nonempty_patches": 5,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.12206444565811032,
+      "restart_teacher_dice": 0.3118739095489208,
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.3032701474337657,
+        "uniform": 0.018090264826557255,
+        "top1": 0.3118739095489208,
+        "top2": 0.3048324685603729,
+        "top3": 0.3042586054721977
+      },
+      "val_dice": 0.27044419233563916,
+      "val_vi": 0.304520179729652,
+      "val_ece": 0.019762184201704298
+    },
+    {
+      "round": 2,
+      "n_labeled": 10,
+      "selection_seconds": 0.1,
+      "round_seconds": 4.7,
+      "teacher_dice_labeled": [
+        0.3370766113617249,
+        0.20785448050939587,
+        0.00023106885851983892,
+        0.001479435841798994
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 29992,
+      "evidence_nonempty_patches": 10,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.2284639770019887,
+        "uniform": 0.028564987762462964,
+        "top1": 0.3370766113617249,
+        "top2": 0.2287742341421212,
+        "top3": 0.22868552543573198
+      },
+      "pseudo_student_dice": 0.4019437835222003,
+      "pseudo_ensemble_dice": 0.36104428904428904,
+      "pseudo_source": "student",
+      "consistency_scale": 0.0,
+      "val_dice": 0.395526844758151,
+      "val_vi": 0.2676356453097999,
+      "val_ece": 0.01013508053922452
+    },
+    {
+      "round": 3,
+      "n_labeled": 15,
+      "selection_seconds": 0.1,
+      "round_seconds": 22.0,
+      "teacher_dice_labeled": [
+        0.47845312015423197,
+        0.36233065584123325,
+        0.0023601996475788446,
+        0.001159658464534431
+      ],
+      "teacher_dice_pool": [
+        0.5826728360901532,
+        0.5055378509654219,
+        0.0036168323423117956,
+        0.00028675870280224237
+      ],
+      "evidence_fg_pixels": 53089,
+      "evidence_nonempty_patches": 15,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.3793255215978283,
+        "uniform": 0.03080981377991152,
+        "top1": 0.47845312015423197,
+        "top2": 0.3795297534477732,
+        "top3": 0.37938025962169
+      },
+      "pseudo_student_dice": 0.6588509676155887,
+      "pseudo_ensemble_dice": 0.42708432425086634,
+      "pseudo_source": "student",
+      "consistency_scale": 0.53,
+      "val_dice": 0.5028134313637753,
+      "val_vi": 0.18638684223811713,
+      "val_ece": 0.008429511681865535
+    },
+    {
+      "round": 4,
+      "n_labeled": 20,
+      "selection_seconds": 0.1,
+      "round_seconds": 22.6,
+      "teacher_dice_labeled": [
+        0.43103000518725343,
+        0.36277141788004924,
+        0.005746551389353614,
+        0.0008508898528234805
+      ],
+      "teacher_dice_pool": [
+        0.4837028345554021,
+        0.49975188054577785,
+        0.004083124007295061,
+        0.00024411224105223264
+      ],
+      "evidence_fg_pixels": 74091,
+      "evidence_nonempty_patches": 20,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        1.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "ensemble_choice": "top1",
+      "candidate_dice": {
+        "mixture": 0.373334630363959,
+        "uniform": 0.03538747088305604,
+        "top1": 0.43103000518725343,
+        "top2": 0.37340299844945096,
+        "top3": 0.37335681419917816
+      },
+      "pseudo_student_dice": 0.6492734286625866,
+      "pseudo_ensemble_dice": 0.3906538327265473,
+      "pseudo_source": "student",
+      "consistency_scale": 0.498,
+      "val_dice": 0.6574593702393732,
+      "val_vi": 0.13533919737072217,
+      "val_ece": 0.008640088643453371
+    },
+    {
+      "round": 5,
+      "n_labeled": 25,
+      "selection_seconds": 0.1,
+      "round_seconds": 22.6,
+      "teacher_dice_labeled": [
+        0.3717128901337308,
+        0.3823762412712888,
+        0.007664640876645464,
+        0.0036914687359017786
+      ],
+      "teacher_dice_pool": [
+        0.3705947151012748,
+        0.45955121435857227,
+        0.009201777039825764,
+        0.0009343618515797726
+      ],
+      "evidence_fg_pixels": 90327,
+      "evidence_nonempty_patches": 25,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.11134689678576087,
+        0.8875352175350215,
+        0.0006398570780882527,
+        0.0004780286011293887
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.3852127875716316,
+        "uniform": 0.054485108960520695,
+        "top1": 0.3823762412712888,
+        "top2": 0.385270066563859,
+        "top3": 0.38524165066760796
+      },
+      "pseudo_student_dice": 0.8290223463687151,
+      "pseudo_ensemble_dice": 0.5929518365361226,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.6519417313245782,
+      "val_vi": 0.12718020187778287,
+      "val_ece": 0.010796660762073279
+    },
+    {
+      "round": 6,
+      "n_labeled": 30,
+      "selection_seconds": 0.1,
+      "round_seconds": 23.1,
+      "teacher_dice_labeled": [
+        0.37641538950337,
+        0.37352917480538395,
+        0.008009229163548612,
+        0.003443806587578693
+      ],
+      "teacher_dice_pool": [
+        0.39919716938468497,
+        0.47226146654321266,
+        0.009783755993179285,
+        0.0012095196263971187
+      ],
+      "evidence_fg_pixels": 102597,
+      "evidence_nonempty_patches": 28,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.12546358417446743,
+        0.8733335622071413,
+        0.0006902340557613062,
+        0.0005126195626298668
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.37627428788395567,
+        "uniform": 0.04938321963255888,
+        "top1": 0.37641538950337,
+        "top2": 0.37633076172426516,
+        "top3": 0.376285460732609
+      },
+      "pseudo_student_dice": 0.7833869929169349,
+      "pseudo_ensemble_dice": 5.972288580984233e-05,
+      "pseudo_source": "student",
+      "consistency_scale": 0.945,
+      "val_dice": 0.6547061788185442,
+      "val_vi": 0.12729252241884847,
+      "val_ece": 0.01092049264393179
+    },
+    {
+      "round": 7,
+      "n_labeled": 30,
+      "selection_seconds": 0.0,
+      "round_seconds": 9.9,
+      "final_consolidation": true,
+      "val_dice": 0.6747453638476713,
+      "val_vi": 0.11778407187789014,
+      "val_ece": 0.009716465947843022
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.03470887967342099,
+    "auroc_top20_error": 0.482987650099299
+  },
+  "al_seconds": 131.5,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.12546358417446743,
+    0.8733335622071413,
+    0.0006902340557613062,
+    0.0005126195626298668
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.8733335622071413,
+    "entropy": 0.38761949128992906,
+    "effective_sources": 1.4734690098128682,
+    "min_weight": 0.0005126195626298668
+  },
+  "top_source_final": "kasthuri-pp",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.7961039268265155,
+    "patch_iou": 0.7097687229145664,
+    "patch_precision": 0.7584498601881322,
+    "patch_recall": 0.9158267507576621,
+    "patch_vi": 0.1241707856182842,
+    "n_patches": 147,
+    "cal_ece": 0.01067422514685362,
+    "cal_nll": 0.05508691845768595,
+    "cal_brier": 0.014360713568413223,
+    "cal_mean_confidence": 0.9927506494665814,
+    "cal_accuracy": 0.9820764243197279,
+    "cal_predicted_prevalence": 0.0682530824829932,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06837096344262414
+  },
+  "test_notta": {
+    "patch_dice": 0.7961039268265155,
+    "patch_iou": 0.7097687229145664,
+    "patch_precision": 0.7584498601881322,
+    "patch_recall": 0.9158267507576621,
+    "patch_vi": 0.1241707856182842,
+    "n_patches": 147,
+    "cal_ece": 0.01067422514685362,
+    "cal_nll": 0.05508691845768595,
+    "cal_brier": 0.014360713568413223,
+    "cal_mean_confidence": 0.9927506494665814,
+    "cal_accuracy": 0.9820764243197279,
+    "cal_predicted_prevalence": 0.0682530824829932,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.06837096344262414
+  },
+  "test": {
+    "patch_dice": 0.795692102630151,
+    "patch_iou": 0.7131124461490234,
+    "patch_precision": 0.752512679245492,
+    "patch_recall": 0.9245430931406642,
+    "patch_vi": 0.11597981232523923,
+    "n_patches": 147,
+    "image_dice": 0.8687959458062932,
+    "image_iou": 0.7683731492929882,
+    "image_precision": 0.79911133488317,
+    "image_recall": 0.952452859311531,
+    "image_vi": 0.14220201389233786,
+    "n_images": 3,
+    "cal_ece": 0.004598624949523295,
+    "cal_nll": 0.04306924903348975,
+    "cal_brier": 0.012508827748498425,
+    "cal_mean_confidence": 0.9878069916271571,
+    "cal_accuracy": 0.9832389987244898,
+    "cal_predicted_prevalence": 0.07049186862244898,
+    "cal_true_prevalence": 0.05939094387755102,
+    "cal_mean_probability": 0.07124218746769279
+  },
+  "eval_seconds": 2.7,
+  "total_seconds": 178.8,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 1.6,
+        "peak_cpu_rss_mb": 1025.6,
+        "peak_gpu_alloc_mb": 592.4,
+        "peak_gpu_reserved_mb": 594.0
+      },
+      "features_and_weights": {
+        "seconds": 6.31,
+        "peak_cpu_rss_mb": 1806.9,
+        "peak_gpu_alloc_mb": 2298.2,
+        "peak_gpu_reserved_mb": 2640.0
+      },
+      "distillation": {
+        "seconds": 36.04,
+        "peak_cpu_rss_mb": 2105.9,
+        "peak_gpu_alloc_mb": 3829.8,
+        "peak_gpu_reserved_mb": 4168.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 131.45,
+        "peak_cpu_rss_mb": 2382.2,
+        "peak_gpu_alloc_mb": 3950.0,
+        "peak_gpu_reserved_mb": 4178.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 2.73,
+        "peak_cpu_rss_mb": 2382.2,
+        "peak_gpu_alloc_mb": 2149.5,
+        "peak_gpu_reserved_mb": 4178.0,
+        "patches": 147,
+        "images": 3,
+        "patches_per_second": 53.8,
+        "seconds_per_image": 0.91
+      }
+    },
+    "peak_gpu_alloc_mb": 3950.0,
+    "peak_cpu_rss_mb": 2382.2,
+    "total_seconds": 178.1
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/drosophila-vnc/macs-v11-norestart-acq-random_b5_s1.pth"
+}

macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-random/budget-5/seed-1/rounds.csv

@@ -0,0 +1,8 @@
+round,n_labeled,selection_seconds,round_seconds,teacher_dice_labeled,teacher_dice_pool,evidence_fg_pixels,evidence_nonempty_patches,first_correction,restart,restart_student_dice,restart_teacher_dice,weights,ensemble_choice,candidate_dice,val_dice,val_vi,val_ece,pseudo_student_dice,pseudo_ensemble_dice,pseudo_source,consistency_scale,final_consolidation
+1,5,5.3,23.3,"[0.3118739095489208, 0.0968435913812229, 0.0004512053628980276, 0.00288]",,15368,5,True,kept,0.12206444565811032,0.3118739095489208,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.3032701474337657, 'uniform': 0.018090264826557255, 'top1': 0.3118739095489208, 'top2': 0.3048324685603729, 'top3': 0.3042586054721977}",0.27044419233563916,0.304520179729652,0.019762184201704298,,,,,
+2,10,0.1,4.7,"[0.3370766113617249, 0.20785448050939587, 0.00023106885851983892, 0.001479435841798994]",,29992,10,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.2284639770019887, 'uniform': 0.028564987762462964, 'top1': 0.3370766113617249, 'top2': 0.2287742341421212, 'top3': 0.22868552543573198}",0.395526844758151,0.2676356453097999,0.01013508053922452,0.4019437835222003,0.36104428904428904,student,0.0,
+3,15,0.1,22.0,"[0.47845312015423197, 0.36233065584123325, 0.0023601996475788446, 0.001159658464534431]","[0.5826728360901532, 0.5055378509654219, 0.0036168323423117956, 0.00028675870280224237]",53089,15,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.3793255215978283, 'uniform': 0.03080981377991152, 'top1': 0.47845312015423197, 'top2': 0.3795297534477732, 'top3': 0.37938025962169}",0.5028134313637753,0.18638684223811713,0.008429511681865535,0.6588509676155887,0.42708432425086634,student,0.53,
+4,20,0.1,22.6,"[0.43103000518725343, 0.36277141788004924, 0.005746551389353614, 0.0008508898528234805]","[0.4837028345554021, 0.49975188054577785, 0.004083124007295061, 0.00024411224105223264]",74091,20,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.373334630363959, 'uniform': 0.03538747088305604, 'top1': 0.43103000518725343, 'top2': 0.37340299844945096, 'top3': 0.37335681419917816}",0.6574593702393732,0.13533919737072217,0.008640088643453371,0.6492734286625866,0.3906538327265473,student,0.498,
+5,25,0.1,22.6,"[0.3717128901337308, 0.3823762412712888, 0.007664640876645464, 0.0036914687359017786]","[0.3705947151012748, 0.45955121435857227, 0.009201777039825764, 0.0009343618515797726]",90327,25,False,,,,"[0.11134689678576087, 0.8875352175350215, 0.0006398570780882527, 0.0004780286011293887]",mixture,"{'mixture': 0.3852127875716316, 'uniform': 0.054485108960520695, 'top1': 0.3823762412712888, 'top2': 0.385270066563859, 'top3': 0.38524165066760796}",0.6519417313245782,0.12718020187778287,0.010796660762073279,0.8290223463687151,0.5929518365361226,student,1.0,
+6,30,0.1,23.1,"[0.37641538950337, 0.37352917480538395, 0.008009229163548612, 0.003443806587578693]","[0.39919716938468497, 0.47226146654321266, 0.009783755993179285, 0.0012095196263971187]",102597,28,False,,,,"[0.12546358417446743, 0.8733335622071413, 0.0006902340557613062, 0.0005126195626298668]",mixture,"{'mixture': 0.37627428788395567, 'uniform': 0.04938321963255888, 'top1': 0.37641538950337, 'top2': 0.37633076172426516, 'top3': 0.376285460732609}",0.6547061788185442,0.12729252241884847,0.01092049264393179,0.7833869929169349,5.972288580984233e-05,student,0.945,
+7,30,0.0,9.9,,,,,,,,,,,,0.6747453638476713,0.11778407187789014,0.009716465947843022,,,,,True

macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt

@@ -0,0 +1,108 @@
+[07:27:23][      0s][START] target=kasthuri-pp variant=macs-v11-norestart-acq-laplace budget=5.0% seed=2 | {"task": "mitochondria", "target": "kasthuri-pp", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 2, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 5, "batch_size": 8, "eval_batch_size": 16, "amp": true, "weighting": "mixture", "feature_n": 512, "feature_seed": 0, "source_split": "train", "kernel_scales": "0.25,0.5,1,2,4", "lam": 0.01, "tau": 0.5, "calibration": "source-val", "ensemble_space": "prob", "student_init": "top-source", "init_compatible": false, "weight_update": "labeled", "reliability_beta": 10.0, "update_cap": 512, "kd_after_update": 2, "reinit_after_update": true, "min_consistency": 64, "consistency_cap": 2000, "select_ensemble": true, "pseudo_source": "auto", "final_epochs": 10, "evidence": "pooled", "min_evidence_fg": 2000, "restart_rule": "evidence", "ema": 0.0, "conf_weight": false, "skip_blank": true, "free_blank": true, "aug_intensity": false, "final_pool": false, "final_pool_cap": 8000, "tta": true, "blank_std": 0.01, "kd": true, "kd_epochs": 5, "kd_lr": 0.0001, "kd_cap": 6000, "kd_kl_weight": 0.0, "rounds": 6, "label_schedule": "rounds", "eval_test_rounds": "", "acquisition": "laplace", "first_round": "stratified", "cold_cap": 2000, "recover": false, "recover_margin": 0.1, "recover_max": 2, "recover_big": 0.3, "recover_epochs": 3, "adabn": true, "adabn_cap": 512, "restart_margin": 9.0, "skip_empty": true, "empty_prevalence": 0.005, "score_pool_cap": 8000, "lap_precision": "pool", "lap_score": "entropy+var", "lap_scale": 1.0, "deconf_prev": false, "diversity": "none", "div_mult": 4, "evidence_explore_only": false, "pseudo_switch_margin": 0.05, "cons_gate": true, "cons_gate_lo": 0.5, "cons_gate_hi": 0.8, "pool_evidence": 0.5, "pool_evidence_n": 2000, "oracle_evidence": false, "typicality": 0.0, "evidence_rounds": 0, "explore_frac": 0.0, "strat_bins": 10, "dump_signals": "", "mc_passes": 5, "mc_dropout": 0.2, "ft_epochs": 5, "ft_lr": 0.0001, "ft_cap": 10000, "ft_loss": "combined", "augment": true, "l2sp": 0.0, "consistency": 0.5, "eval_rounds": true, "val_cap": 1024, "image_level": true}
+[07:27:23][      0s][DATA] MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.
+[07:27:23][      0s][DATA] kasthuri-pp: train=12023 val=2376 test=2376 patches
+[07:27:24][      1s][CALIB] source-val temperatures: fly-mito=1.38, lucchi=3.80, dros-vnc=2.77, multiclass=2.60
+[07:27:24][      1s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s
+[07:27:25][      2s][WEIGHTS] mixture: fly-mito=0.055, lucchi=0.143, dros-vnc=0.678, multiclass=0.124 | eff.sources=2.61 | MMD2 mix=0.1257 uniform=0.2386
+[07:27:25][      2s][WEIGHTS] per-source: fly-mito=0.004, lucchi=0.028, dros-vnc=0.948, multiclass=0.020 | min-MMD source (NeuroADDA rule) = dros-vnc
+[07:27:38][     15s][TEACHERS] target-val Dice: fly-mito=0.194, lucchi=0.137, dros-vnc=0.349, multiclass=0.152
+[07:27:38][     15s][TEACHERS] ensemble target-val Dice: mixture=0.341, mixture@raw=0.346, mixture@logit=0.321, per_source=0.348, per_source@raw=0.348, per_source@logit=0.346, uniform=0.188, uniform@raw=0.183, uniform@logit=0.207
+[07:27:38][     15s][WEIGHTS] using mixture: fly-mito=0.055, lucchi=0.143, dros-vnc=0.678, multiclass=0.124
+[07:27:38][     15s][INIT] student initialised from teacher dros-vnc (weight 0.678)
+[07:27:39][     16s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches
+[07:27:39][     16s][KD] distilling the mixture ensemble (prob space) into the student
+[07:28:14][     51s][KD] soft labels for 6000 patches from 4 teachers in 34s
+[07:29:07][    104s][KD] epoch 1/5 | patches=6000 | loss=0.1778 | 53s
+[07:30:02][    159s][KD] epoch 2/5 | patches=6000 | loss=0.1510 | 54s
+[07:30:56][    213s][KD] epoch 3/5 | patches=6000 | loss=0.1461 | 54s
+[07:31:49][    266s][KD] epoch 4/5 | patches=6000 | loss=0.1440 | 53s
+[07:32:43][    320s][KD] epoch 5/5 | patches=6000 | loss=0.1430 | 54s
+[07:32:45][    322s][ROUND0] before any target label: val_dice=0.2118 val_vi=0.1496
+[07:32:45][    322s][AL] patches=12023 (pool 10327 after dropping blank; free blank negatives: True) | budget=601 (5.0%) | rounds=6 | per_round=100 | acquisition=laplace | first_round=stratified
+[07:33:10][    347s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': -0.36017774401252334, 'auroc_top20_error': 0.360394235162642}
+[07:33:22][    359s][AL] round 1: stratified cold start: 50 patches from the membrane-rich half (ensemble prevalence >= 0.034) + 50 from the membrane-poor half of 2000 candidates
+[07:33:28][    365s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.221, uniform=0.011, top1=0.223, top2=0.223, top3=0.222 -> using mixture
+[07:33:28][    365s][WEIGHTS] round 1: student kept (labeled Dice 0.013 vs best teacher dros-vnc 0.223); distilling the corrected ensemble into it
+[07:33:28][    365s][WEIGHTS] round 1: teacher Dice on 100 labeled patches + 100 blank (267058 fg px): fly-mito=0.038, lucchi=0.009, dros-vnc=0.223, multiclass=0.000 | weights: fly-mito=0.012, lucchi=0.023, dros-vnc=0.946, multiclass=0.019
+[07:34:03][    400s][KD] soft labels for 6000 patches from 4 teachers in 34s
+[07:34:57][    454s][KD] epoch 1/2 | patches=6000 | loss=0.0535 | 54s
+[07:35:51][    508s][KD] epoch 2/2 | patches=6000 | loss=0.0528 | 54s
+[07:36:03][    520s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 12s
+[07:36:07][    524s][FT] round 1/6 epoch 1/5 | patches=200 | loss=0.7622
+[07:36:11][    528s][FT] round 1/6 epoch 2/5 | patches=200 | loss=0.5386
+[07:36:15][    532s][FT] round 1/6 epoch 3/5 | patches=200 | loss=0.4307
+[07:36:19][    536s][FT] round 1/6 epoch 4/5 | patches=200 | loss=0.3971
+[07:36:23][    540s][FT] round 1/6 epoch 5/5 | patches=200 | loss=0.3589
+[07:36:25][    542s][AL] round 1/6 | labeled=100 | selection_seconds=37.5000 round_seconds=218.1000 restart_student_dice=0.0133 restart_teacher_dice=0.2227 val_dice=0.7285 val_vi=0.0893 val_ece=0.0073
+[07:36:50][    567s][AL] round 2: 2834/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:37:00][    577s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.130, uniform=0.008, top1=0.132, top2=0.133, top3=0.131 -> using mixture
+[07:37:00][    577s][WEIGHTS] round 2: teacher Dice on 200 labeled patches + 200 blank (1115399 fg px): fly-mito=0.084, lucchi=0.012, dros-vnc=0.132, multiclass=0.007 | weights: fly-mito=0.043, lucchi=0.054, dros-vnc=0.858, multiclass=0.045
+[07:37:08][    585s][CONS] round 2: on the 100 new patches student Dice=0.552 vs ensemble 0.100 -> pseudo-labels from the student | consistency x0.17
+[07:37:15][    592s][FT] round 2/6 epoch 1/5 | patches=400 | loss=0.2396
+[07:37:23][    600s][FT] round 2/6 epoch 2/5 | patches=400 | loss=0.2152
+[07:37:30][    607s][FT] round 2/6 epoch 3/5 | patches=400 | loss=0.2085
+[07:37:37][    614s][FT] round 2/6 epoch 4/5 | patches=400 | loss=0.1984
+[07:37:44][    621s][FT] round 2/6 epoch 5/5 | patches=400 | loss=0.2184
+[07:37:46][    623s][AL] round 2/6 | labeled=200 | selection_seconds=25.5000 round_seconds=79.6000 pseudo_student_dice=0.5515 pseudo_ensemble_dice=0.0995 consistency_scale=0.1720 val_dice=0.8179 val_vi=0.0591 val_ece=0.0027
+[07:38:12][    649s][AL] round 3: 2464/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:38:44][    681s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.112, lucchi=0.015, dros-vnc=0.285, multiclass=0.001 | blended with labeled Dice at 0.50
+[07:39:13][    710s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.207, uniform=0.012, top1=0.208, top2=0.209, top3=0.208 -> using mixture
+[07:39:13][    710s][WEIGHTS] round 3: teacher Dice on 300 labeled patches + 300 blank (1632966 fg px): fly-mito=0.098, lucchi=0.021, dros-vnc=0.208, multiclass=0.004 | weights: fly-mito=0.025, lucchi=0.030, dros-vnc=0.923, multiclass=0.022
+[07:39:21][    718s][CONS] round 3: on the 100 new patches student Dice=0.781 vs ensemble 0.130 -> pseudo-labels from the student | consistency x0.94
+[07:39:31][    728s][FT] round 3/6 epoch 1/5 | patches=600 | loss=0.2248
+[07:39:42][    739s][FT] round 3/6 epoch 2/5 | patches=600 | loss=0.1817
+[07:39:53][    750s][FT] round 3/6 epoch 3/5 | patches=600 | loss=0.1835
+[07:40:03][    760s][FT] round 3/6 epoch 4/5 | patches=600 | loss=0.1582
+[07:40:14][    771s][FT] round 3/6 epoch 5/5 | patches=600 | loss=0.1784
+[07:40:16][    773s][AL] round 3/6 | labeled=300 | selection_seconds=25.4000 round_seconds=147.7000 pseudo_student_dice=0.7809 pseudo_ensemble_dice=0.1295 consistency_scale=0.9360 val_dice=0.8398 val_vi=0.0550 val_ece=0.0030
+[07:40:41][    798s][AL] round 4: 2492/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:41:16][    833s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.113, lucchi=0.014, dros-vnc=0.303, multiclass=0.001 | blended with labeled Dice at 0.50
+[07:41:47][    864s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.222, uniform=0.013, top1=0.223, top2=0.224, top3=0.223 -> using mixture
+[07:41:47][    864s][WEIGHTS] round 4: teacher Dice on 400 labeled patches + 400 blank (2188516 fg px): fly-mito=0.097, lucchi=0.018, dros-vnc=0.223, multiclass=0.004 | weights: fly-mito=0.022, lucchi=0.025, dros-vnc=0.934, multiclass=0.019
+[07:41:55][    872s][CONS] round 4: on the 100 new patches student Dice=0.815 vs ensemble 0.176 -> pseudo-labels from the student | consistency x1.00
+[07:42:09][    886s][FT] round 4/6 epoch 1/5 | patches=800 | loss=0.1741
+[07:42:23][    900s][FT] round 4/6 epoch 2/5 | patches=800 | loss=0.1575
+[07:42:37][    914s][FT] round 4/6 epoch 3/5 | patches=800 | loss=0.1496
+[07:42:51][    928s][FT] round 4/6 epoch 4/5 | patches=800 | loss=0.1385
+[07:43:05][    942s][FT] round 4/6 epoch 5/5 | patches=800 | loss=0.1491
+[07:43:07][    944s][AL] round 4/6 | labeled=400 | selection_seconds=25.5000 round_seconds=169.0000 pseudo_student_dice=0.8146 pseudo_ensemble_dice=0.1760 consistency_scale=1.0000 val_dice=0.8727 val_vi=0.0459 val_ece=0.0024
+[07:43:32][    969s][AL] round 5: 2323/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:44:09][   1006s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.113, lucchi=0.015, dros-vnc=0.264, multiclass=0.001 | blended with labeled Dice at 0.50
+[07:44:42][   1039s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.207, uniform=0.013, top1=0.209, top2=0.209, top3=0.208 -> using mixture
+[07:44:42][   1039s][WEIGHTS] round 5: teacher Dice on 500 labeled patches + 500 blank (2518881 fg px): fly-mito=0.094, lucchi=0.018, dros-vnc=0.209, multiclass=0.003 | weights: fly-mito=0.024, lucchi=0.029, dros-vnc=0.925, multiclass=0.022
+[07:44:50][   1047s][CONS] round 5: on the 100 new patches student Dice=0.794 vs ensemble 0.214 -> pseudo-labels from the student | consistency x0.98
+[07:45:08][   1065s][FT] round 5/6 epoch 1/5 | patches=1000 | loss=0.1510
+[07:45:25][   1082s][FT] round 5/6 epoch 2/5 | patches=1000 | loss=0.1710
+[07:45:42][   1099s][FT] round 5/6 epoch 3/5 | patches=1000 | loss=0.1518
+[07:46:00][   1117s][FT] round 5/6 epoch 4/5 | patches=1000 | loss=0.1546
+[07:46:17][   1134s][FT] round 5/6 epoch 5/5 | patches=1000 | loss=0.1450
+[07:46:19][   1136s][AL] round 5/6 | labeled=500 | selection_seconds=25.5000 round_seconds=190.5000 pseudo_student_dice=0.7943 pseudo_ensemble_dice=0.2140 consistency_scale=0.9810 val_dice=0.8634 val_vi=0.0461 val_ece=0.0023
+[07:46:45][   1162s][AL] round 6: 2522/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005)
+[07:47:21][   1198s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.112, lucchi=0.012, dros-vnc=0.286, multiclass=0.002 | blended with labeled Dice at 0.50
+[07:47:55][   1232s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.235, uniform=0.012, top1=0.236, top2=0.236, top3=0.235 -> using mixture
+[07:47:55][   1232s][WEIGHTS] round 6: teacher Dice on 512 labeled patches + 512 blank (2489206 fg px): fly-mito=0.093, lucchi=0.015, dros-vnc=0.236, multiclass=0.004 | weights: fly-mito=0.018, lucchi=0.022, dros-vnc=0.943, multiclass=0.017
+[07:48:03][   1240s][CONS] round 6: on the 100 new patches student Dice=0.945 vs ensemble 0.367 -> pseudo-labels from the student | consistency x1.00
+[07:48:24][   1261s][FT] round 6/6 epoch 1/5 | patches=1200 | loss=0.1372
+[07:48:45][   1282s][FT] round 6/6 epoch 2/5 | patches=1200 | loss=0.1167
+[07:49:05][   1302s][FT] round 6/6 epoch 3/5 | patches=1200 | loss=0.1188
+[07:49:26][   1323s][FT] round 6/6 epoch 4/5 | patches=1200 | loss=0.1208
+[07:49:47][   1344s][FT] round 6/6 epoch 5/5 | patches=1200 | loss=0.1142
+[07:49:49][   1346s][AL] round 6/6 | labeled=600 | selection_seconds=25.4000 round_seconds=207.8000 pseudo_student_dice=0.9455 pseudo_ensemble_dice=0.3667 consistency_scale=1.0000 val_dice=0.8758 val_vi=0.0413 val_ece=0.0024
+[07:49:52][   1349s][CONS] final consolidation: pseudo-labels refreshed from the student
+[07:50:13][   1370s][FT] final 1/10 epoch 1/1 | patches=1200 | loss=0.1081
+[07:50:34][   1391s][FT] final 2/10 epoch 1/1 | patches=1200 | loss=0.1131
+[07:50:55][   1412s][FT] final 3/10 epoch 1/1 | patches=1200 | loss=0.1010
+[07:51:16][   1433s][FT] final 4/10 epoch 1/1 | patches=1200 | loss=0.1056
+[07:51:36][   1453s][FT] final 5/10 epoch 1/1 | patches=1200 | loss=0.0999
+[07:51:57][   1474s][FT] final 6/10 epoch 1/1 | patches=1200 | loss=0.0932
+[07:52:18][   1495s][FT] final 7/10 epoch 1/1 | patches=1200 | loss=0.0827
+[07:52:39][   1516s][FT] final 8/10 epoch 1/1 | patches=1200 | loss=0.0815
+[07:53:00][   1537s][FT] final 9/10 epoch 1/1 | patches=1200 | loss=0.0790
+[07:53:21][   1558s][FT] final 10/10 epoch 1/1 | patches=1200 | loss=0.0742
+[07:53:23][   1560s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=211.4000 val_dice=0.8811 val_vi=0.0406 val_ece=0.0026
+[07:53:23][   1560s][WEIGHTS] final: fly-mito=0.018, lucchi=0.022, dros-vnc=0.943, multiclass=0.017
+[07:53:55][   1592s][TEST] raw student (no EMA, no TTA): patch_dice=0.8717 patch_vi=0.0270
+[07:53:59][   1596s][TEST] final model without TTA: patch_dice=0.8717 patch_vi=0.0270
+[07:53:59][   1596s][RESOURCES] {"phases": {"setup": {"seconds": 0.59, "peak_cpu_rss_mb": 1028.3, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 15.46, "peak_cpu_rss_mb": 1793.0, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 303.55, "peak_cpu_rss_mb": 3685.8, "peak_gpu_alloc_mb": 3829.8, "peak_gpu_reserved_mb": 4128.0, "kd_epochs": 5}, "active_learning": {"seconds": 1237.79, "peak_cpu_rss_mb": 3685.8, "peak_gpu_alloc_mb": 3950.0, "peak_gpu_reserved_mb": 4170.0, "rounds": 6}, "inference": {"seconds": 36.81, "peak_cpu_rss_mb": 3685.8, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4172.0, "patches": 2376, "images": 24, "patches_per_second": 64.5, "seconds_per_image": 1.534}}, "peak_gpu_alloc_mb": 3950.0, "peak_cpu_rss_mb": 3685.8, "total_seconds": 1594.2}
+[07:53:59][   1596s][TEST] patch_dice=0.8806 patch_iou=0.8521 patch_precision=0.9547 patch_recall=0.8873 patch_vi=0.0259 image_dice=0.9483 image_iou=0.9021 image_precision=0.9598 image_recall=0.9376 image_vi=0.0308 cal_ece=0.0014 cal_nll=0.0117 cal_brier=0.0024 cal_mean_confidence=0.9983 cal_accuracy=0.9969 cal_predicted_prevalence=0.0262 cal_true_prevalence=0.0268 cal_mean_probability=0.0260
+[07:53:59][   1596s][DONE] kasthuri-pp macs-v11-norestart-acq-laplace b5 s2 | test patch Dice=0.8806 VI=0.0259 | image Dice=0.9483 VI=0.0308 | 1596s

macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json

@@ -0,0 +1,667 @@
+{
+  "target": "kasthuri-pp",
+  "variant": "macs-v11-norestart-acq-laplace",
+  "budget": 5.0,
+  "seed": 2,
+  "config": {
+    "task": "mitochondria",
+    "target": "kasthuri-pp",
+    "variant": "macs-v11-norestart-acq-laplace",
+    "budget": 5.0,
+    "seed": 2,
+    "sources": "",
+    "run_tag": "",
+    "arch": "unet",
+    "teacher_ckpt": "model_best.pth",
+    "models_dir": "/home/user/connectomics/MACS-2/models",
+    "cache_dir": "/home/user/connectomics/MACS-2/cache/macs_v2",
+    "output_dir": "experiments/macs_v2/outputs-r26",
+    "checkpoint_dir": "/home/user/connectomics/MACS-2/models/experiments/macs_v2",
+    "device": "cuda:0",
+    "num_workers": 5,
+    "batch_size": 8,
+    "eval_batch_size": 16,
+    "amp": true,
+    "weighting": "mixture",
+    "feature_n": 512,
+    "feature_seed": 0,
+    "source_split": "train",
+    "kernel_scales": "0.25,0.5,1,2,4",
+    "lam": 0.01,
+    "tau": 0.5,
+    "calibration": "source-val",
+    "ensemble_space": "prob",
+    "student_init": "top-source",
+    "init_compatible": false,
+    "weight_update": "labeled",
+    "reliability_beta": 10.0,
+    "update_cap": 512,
+    "kd_after_update": 2,
+    "reinit_after_update": true,
+    "min_consistency": 64,
+    "consistency_cap": 2000,
+    "select_ensemble": true,
+    "pseudo_source": "auto",
+    "final_epochs": 10,
+    "evidence": "pooled",
+    "min_evidence_fg": 2000,
+    "restart_rule": "evidence",
+    "ema": 0.0,
+    "conf_weight": false,
+    "skip_blank": true,
+    "free_blank": true,
+    "aug_intensity": false,
+    "final_pool": false,
+    "final_pool_cap": 8000,
+    "tta": true,
+    "blank_std": 0.01,
+    "kd": true,
+    "kd_epochs": 5,
+    "kd_lr": 0.0001,
+    "kd_cap": 6000,
+    "kd_kl_weight": 0.0,
+    "rounds": 6,
+    "label_schedule": "rounds",
+    "eval_test_rounds": "",
+    "acquisition": "laplace",
+    "first_round": "stratified",
+    "cold_cap": 2000,
+    "recover": false,
+    "recover_margin": 0.1,
+    "recover_max": 2,
+    "recover_big": 0.3,
+    "recover_epochs": 3,
+    "adabn": true,
+    "adabn_cap": 512,
+    "restart_margin": 9.0,
+    "skip_empty": true,
+    "empty_prevalence": 0.005,
+    "score_pool_cap": 8000,
+    "lap_precision": "pool",
+    "lap_score": "entropy+var",
+    "lap_scale": 1.0,
+    "deconf_prev": false,
+    "diversity": "none",
+    "div_mult": 4,
+    "evidence_explore_only": false,
+    "pseudo_switch_margin": 0.05,
+    "cons_gate": true,
+    "cons_gate_lo": 0.5,
+    "cons_gate_hi": 0.8,
+    "pool_evidence": 0.5,
+    "pool_evidence_n": 2000,
+    "oracle_evidence": false,
+    "typicality": 0.0,
+    "evidence_rounds": 0,
+    "explore_frac": 0.0,
+    "strat_bins": 10,
+    "dump_signals": "",
+    "mc_passes": 5,
+    "mc_dropout": 0.2,
+    "ft_epochs": 5,
+    "ft_lr": 0.0001,
+    "ft_cap": 10000,
+    "ft_loss": "combined",
+    "augment": true,
+    "l2sp": 0.0,
+    "consistency": 0.5,
+    "eval_rounds": true,
+    "val_cap": 1024,
+    "image_level": true
+  },
+  "commit": "aecf380",
+  "data_access": "MMD weights and distillation use the unlabeled target train pool only; teacher temperatures are fitted on each teacher's own source validation set; target validation labels are used only for reported diagnostics, per-round monitoring and the explicitly named best-single oracle; test labels are used once, for the final evaluation.",
+  "environment": {
+    "host": "pc2",
+    "python": "3.14.7",
+    "torch": "2.14.1+cu130",
+    "argv": [
+      "experiments/macs_v2/run.py",
+      "sweep",
+      "--targets",
+      "kasthuri-pp",
+      "--variants",
+      "macs-v11-norestart-acq-laplace",
+      "--budgets",
+      "5",
+      "--seeds",
+      "2",
+      "--num-workers",
+      "5",
+      "--output-dir",
+      "experiments/macs_v2/outputs-v3",
+      "--task",
+      "mitochondria",
+      "--output-dir",
+      "experiments/macs_v2/outputs-r26"
+    ],
+    "cwd": "/home/user/connectomics/MACS-2",
+    "gpu": "NVIDIA GeForce RTX 4090",
+    "cuda": "13.0",
+    "commit": "aecf380",
+    "dirty": true
+  },
+  "sources": [
+    "fly-brain-mito-benchmark",
+    "lucchi-pp",
+    "drosophila-vnc",
+    "multiclass"
+  ],
+  "alphas": [
+    0.0552621669329701,
+    0.14304343546870246,
+    0.6777952504109969,
+    0.12389914718733061
+  ],
+  "weights_all": {
+    "sources": [
+      "fly-brain-mito-benchmark",
+      "lucchi-pp",
+      "drosophila-vnc",
+      "multiclass"
+    ],
+    "mixture": [
+      0.0552621669329701,
+      0.14304343546870246,
+      0.6777952504109969,
+      0.12389914718733061
+    ],
+    "per_source": [
+      0.00394004854372213,
+      0.027549943457875075,
+      0.9484681548320345,
+      0.02004185316636836
+    ],
+    "uniform": [
+      0.25,
+      0.25,
+      0.25,
+      0.25
+    ],
+    "mixture_info": {
+      "objective": 0.1161079474391782,
+      "mmd2_mixture": 0.12571317215793731,
+      "mmd2_uniform": 0.23862471954971437,
+      "mmd2_single_source": [
+        0.7126139881719412,
+        0.4366429148044758,
+        0.1798181763334552,
+        0.4482686385790514
+      ],
+      "median_squared_distance": 0.1734670188179156,
+      "gammas": [
+        1.4411961518887884,
+        2.8823923037775767,
+        5.764784607555153,
+        11.529569215110307,
+        23.059138430220614
+      ],
+      "max_weight": 0.6777952504109969,
+      "entropy": 0.9605224718759118,
+      "effective_sources": 2.6130613679073416,
+      "min_weight": 0.0552621669329701
+    },
+    "per_source_info": {
+      "mmd2_own_rkhs": [
+        0.6491308049576032,
+        0.5568981487675557,
+        0.3890679978127308,
+        0.5719877547192425
+      ],
+      "standardized": [
+        1.1318865389009094,
+        0.1594828192631669,
+        -1.6099410463963124,
+        0.31857168823223964
+      ],
+      "tau": 0.5,
+      "max_weight": 0.9484681548320345,
+      "entropy": 0.24930992499892873,
+      "effective_sources": 1.2831396485053355,
+      "min_weight": 0.00394004854372213
+    },
+    "lam": 0.01,
+    "tau": 0.5,
+    "single_mmd_source": "drosophila-vnc",
+    "seconds": 1.0
+  },
+  "teacher_diagnostics": {
+    "teacher_val_dice": [
+      0.19408169161460087,
+      0.13693054106355262,
+      0.34856214071670877,
+      0.15175765833135202
+    ],
+    "ensemble_val_dice": {
+      "mixture": 0.3408904698785591,
+      "mixture@raw": 0.3461558372224687,
+      "mixture@logit": 0.3209424494814286,
+      "per_source": 0.3475944498943022,
+      "per_source@raw": 0.34822194749797575,
+      "per_source@logit": 0.3464301210911742,
+      "uniform": 0.18819756340126403,
+      "uniform@raw": 0.18329087899026675,
+      "uniform@logit": 0.2066047847592749
+    },
+    "val_patches": 1024
+  },
+  "temperatures": [
+    1.3819772005081177,
+    3.7969236373901367,
+    2.767761468887329,
+    2.597255229949951
+  ],
+  "top_source": "drosophila-vnc",
+  "weight_stats": {
+    "max_weight": 0.6777952504109969,
+    "entropy": 0.9605224718759118,
+    "effective_sources": 2.6130613679073416,
+    "min_weight": 0.0552621669329701
+  },
+  "correlations": {
+    "spearman_weight_teacher_dice": 0.19999999999999998,
+    "spearman_weight_taxonomy": 0.2581988897471611,
+    "top_source": "drosophila-vnc",
+    "top_source_is_best_teacher": true
+  },
+  "init_teacher": "drosophila-vnc",
+  "student_init": "top-source",
+  "blank_fraction": 0.14106296265491142,
+  "kd_seconds": 303.6,
+  "round0_val": {
+    "val_dice": 0.2118068600167677,
+    "val_vi": 0.14964552136814735,
+    "val_ece": 0.02150047831005253
+  },
+  "rounds": [
+    {
+      "round": 1,
+      "n_labeled": 100,
+      "selection_seconds": 37.5,
+      "round_seconds": 218.1,
+      "teacher_dice_labeled": [
+        0.03829273777525091,
+        0.008551357968456415,
+        0.22265360583181462,
+        0.00019730768085296483
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 267058,
+      "evidence_nonempty_patches": 74,
+      "first_correction": true,
+      "restart": "kept",
+      "restart_student_dice": 0.013334523583695058,
+      "restart_teacher_dice": 0.22265360583181462,
+      "weights": [
+        0.012200911710615017,
+        0.0234567119495013,
+        0.9456533644481262,
+        0.018689011891757392
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.22146026492223692,
+        "uniform": 0.011353831964774486,
+        "top1": 0.22265360583181462,
+        "top2": 0.22287410068430968,
+        "top3": 0.2221372032521384
+      },
+      "val_dice": 0.7284880907378732,
+      "val_vi": 0.08925383861862662,
+      "val_ece": 0.007318781441033324
+    },
+    {
+      "round": 2,
+      "n_labeled": 200,
+      "selection_seconds": 25.5,
+      "round_seconds": 79.6,
+      "teacher_dice_labeled": [
+        0.0843977025985284,
+        0.012172782655249998,
+        0.13224230106627227,
+        0.0066752619085354665
+      ],
+      "teacher_dice_pool": null,
+      "evidence_fg_pixels": 1115399,
+      "evidence_nonempty_patches": 172,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.043330876352989676,
+        0.054471414405189024,
+        0.8575402913759741,
+        0.04465741786584716
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.12992613885762752,
+        "uniform": 0.008046029319777243,
+        "top1": 0.13224230106627227,
+        "top2": 0.13316217464277688,
+        "top3": 0.13142508151376586
+      },
+      "pseudo_student_dice": 0.5515170063829185,
+      "pseudo_ensemble_dice": 0.09952690898409115,
+      "pseudo_source": "student",
+      "consistency_scale": 0.172,
+      "val_dice": 0.8179183255309906,
+      "val_vi": 0.05907835029705699,
+      "val_ece": 0.002730643722707998
+    },
+    {
+      "round": 3,
+      "n_labeled": 300,
+      "selection_seconds": 25.4,
+      "round_seconds": 147.7,
+      "teacher_dice_labeled": [
+        0.09774431100649172,
+        0.02100781303993087,
+        0.20822007653296637,
+        0.0037858675867297508
+      ],
+      "teacher_dice_pool": [
+        0.11166637835235022,
+        0.014642946673106777,
+        0.28479381985635444,
+        0.0009620469771313624
+      ],
+      "evidence_fg_pixels": 1632966,
+      "evidence_nonempty_patches": 266,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.02493772672979201,
+        0.02996639548739577,
+        0.9232464064547601,
+        0.021849471328052025
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.20698885628003225,
+        "uniform": 0.011817528245685245,
+        "top1": 0.20822007653296637,
+        "top2": 0.2087118275421432,
+        "top3": 0.20773206921661225
+      },
+      "pseudo_student_dice": 0.7809243690957529,
+      "pseudo_ensemble_dice": 0.1295378560580478,
+      "pseudo_source": "student",
+      "consistency_scale": 0.936,
+      "val_dice": 0.8397631826742316,
+      "val_vi": 0.05498086933118604,
+      "val_ece": 0.00298188285613413
+    },
+    {
+      "round": 4,
+      "n_labeled": 400,
+      "selection_seconds": 25.5,
+      "round_seconds": 169.0,
+      "teacher_dice_labeled": [
+        0.09741935827619103,
+        0.018257986269871015,
+        0.22313621046835663,
+        0.0035449807697112965
+      ],
+      "teacher_dice_pool": [
+        0.11298328490431746,
+        0.013772936029469702,
+        0.30295471895352916,
+        0.0013155154647071317
+      ],
+      "evidence_fg_pixels": 2188516,
+      "evidence_nonempty_patches": 363,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.021660446784121497,
+        0.025404693616115942,
+        0.9339408533881395,
+        0.018994006211623046
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.22211558136819576,
+        "uniform": 0.012848261537052798,
+        "top1": 0.22313621046835663,
+        "top2": 0.22353551488564827,
+        "top3": 0.22274276271997542
+      },
+      "pseudo_student_dice": 0.8146274749309655,
+      "pseudo_ensemble_dice": 0.17598222940226171,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.8727027758814474,
+      "val_vi": 0.04586085054725591,
+      "val_ece": 0.0023572107581083693
+    },
+    {
+      "round": 5,
+      "n_labeled": 500,
+      "selection_seconds": 25.5,
+      "round_seconds": 190.5,
+      "teacher_dice_labeled": [
+        0.09423271841560225,
+        0.01786067875926255,
+        0.20863073975641946,
+        0.0032668848418224545
+      ],
+      "teacher_dice_pool": [
+        0.11321393596398284,
+        0.015169177507659712,
+        0.2639679558408339,
+        0.0014020792250121128
+      ],
+      "evidence_fg_pixels": 2518881,
+      "evidence_nonempty_patches": 451,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.024031619475181773,
+        0.028983011307454676,
+        0.9252901596473508,
+        0.02169520957001273
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.20744316954220898,
+        "uniform": 0.01252951462867678,
+        "top1": 0.20863073975641946,
+        "top2": 0.2090463951847489,
+        "top3": 0.2081601937542924
+      },
+      "pseudo_student_dice": 0.7943479375839712,
+      "pseudo_ensemble_dice": 0.21403595365511507,
+      "pseudo_source": "student",
+      "consistency_scale": 0.981,
+      "val_dice": 0.8633937643280325,
+      "val_vi": 0.0460846659403376,
+      "val_ece": 0.0023077556545834403
+    },
+    {
+      "round": 6,
+      "n_labeled": 600,
+      "selection_seconds": 25.4,
+      "round_seconds": 207.8,
+      "teacher_dice_labeled": [
+        0.09276645918051137,
+        0.014889898090616777,
+        0.23552871081062293,
+        0.0035663265755556104
+      ],
+      "teacher_dice_pool": [
+        0.11218089728921189,
+        0.011678871931780969,
+        0.28592648339309323,
+        0.0019156104138320128
+      ],
+      "evidence_fg_pixels": 2489206,
+      "evidence_nonempty_patches": 468,
+      "first_correction": false,
+      "restart": "",
+      "weights": [
+        0.018437500364764308,
+        0.021904253452186863,
+        0.9427167721353997,
+        0.01694147404764911
+      ],
+      "ensemble_choice": "mixture",
+      "candidate_dice": {
+        "mixture": 0.23450584653940748,
+        "uniform": 0.011503959466380224,
+        "top1": 0.23552871081062293,
+        "top2": 0.23586462884397835,
+        "top3": 0.23511377156938024
+      },
+      "pseudo_student_dice": 0.9454673857874526,
+      "pseudo_ensemble_dice": 0.36671536038318975,
+      "pseudo_source": "student",
+      "consistency_scale": 1.0,
+      "val_dice": 0.8758495345285738,
+      "val_vi": 0.041280698120005795,
+      "val_ece": 0.00239807538318637
+    },
+    {
+      "round": 7,
+      "n_labeled": 600,
+      "selection_seconds": 0.0,
+      "round_seconds": 211.4,
+      "final_consolidation": true,
+      "val_dice": 0.8810958794996602,
+      "val_vi": 0.04057494837156638,
+      "val_ece": 0.0026333786376941874
+    }
+  ],
+  "uncertainty_quality": {
+    "spearman_uncertainty_error": -0.36017774401252334,
+    "auroc_top20_error": 0.360394235162642
+  },
+  "al_seconds": 1237.8,
+  "source_archs": [
+    "unet",
+    "unet",
+    "unet",
+    "unet"
+  ],
+  "alphas_final": [
+    0.018437500364764308,
+    0.021904253452186863,
+    0.9427167721353997,
+    0.01694147404764911
+  ],
+  "weight_stats_final": {
+    "max_weight": 0.9427167721353997,
+    "entropy": 0.2820229765274581,
+    "effective_sources": 1.325809182127911,
+    "min_weight": 0.01694147404764911
+  },
+  "top_source_final": "drosophila-vnc",
+  "ensemble_choice_final": "",
+  "restart": "kept",
+  "test_raw_student": {
+    "patch_dice": 0.8716612471825786,
+    "patch_iou": 0.8425408245737226,
+    "patch_precision": 0.9459390937256462,
+    "patch_recall": 0.8854203269001327,
+    "patch_vi": 0.027044653248263214,
+    "n_patches": 2376,
+    "cal_ece": 0.0019236981596546764,
+    "cal_nll": 0.01505363365568678,
+    "cal_brier": 0.0027411552110365505,
+    "cal_mean_confidence": 0.9986372529786782,
+    "cal_accuracy": 0.9967135548190236,
+    "cal_predicted_prevalence": 0.026243473142887205,
+    "cal_true_prevalence": 0.026826286958122895,
+    "cal_mean_probability": 0.025931264006348577
+  },
+  "test_notta": {
+    "patch_dice": 0.8716612471825786,
+    "patch_iou": 0.8425408245737226,
+    "patch_precision": 0.9459390937256462,
+    "patch_recall": 0.8854203269001327,
+    "patch_vi": 0.027044653248263214,
+    "n_patches": 2376,
+    "cal_ece": 0.0019236981596546764,
+    "cal_nll": 0.01505363365568678,
+    "cal_brier": 0.0027411552110365505,
+    "cal_mean_confidence": 0.9986372529786782,
+    "cal_accuracy": 0.9967135548190236,
+    "cal_predicted_prevalence": 0.026243473142887205,
+    "cal_true_prevalence": 0.026826286958122895,
+    "cal_mean_probability": 0.025931264006348577
+  },
+  "test": {
+    "patch_dice": 0.8806073948945004,
+    "patch_iou": 0.8520803541413138,
+    "patch_precision": 0.9547011432703486,
+    "patch_recall": 0.887258433513542,
+    "patch_vi": 0.025885128225993523,
+    "n_patches": 2376,
+    "image_dice": 0.948303957395134,
+    "image_iou": 0.9020892014257859,
+    "image_precision": 0.9597966206412246,
+    "image_recall": 0.9375817043623197,
+    "image_vi": 0.03077187452953857,
+    "n_images": 24,
+    "cal_ece": 0.0014176467257451166,
+    "cal_nll": 0.011710549894120235,
+    "cal_brier": 0.0024250120921702435,
+    "cal_mean_confidence": 0.9983486264448106,
+    "cal_accuracy": 0.9969309797190656,
+    "cal_predicted_prevalence": 0.026230731797138047,
+    "cal_true_prevalence": 0.026826286958122895,
+    "cal_mean_probability": 0.02602820112381996
+  },
+  "eval_seconds": 36.8,
+  "total_seconds": 1596.3,
+  "resources": {
+    "phases": {
+      "setup": {
+        "seconds": 0.59,
+        "peak_cpu_rss_mb": 1028.3,
+        "peak_gpu_alloc_mb": 592.4,
+        "peak_gpu_reserved_mb": 594.0
+      },
+      "features_and_weights": {
+        "seconds": 15.46,
+        "peak_cpu_rss_mb": 1793.0,
+        "peak_gpu_alloc_mb": 2298.2,
+        "peak_gpu_reserved_mb": 2640.0
+      },
+      "distillation": {
+        "seconds": 303.55,
+        "peak_cpu_rss_mb": 3685.8,
+        "peak_gpu_alloc_mb": 3829.8,
+        "peak_gpu_reserved_mb": 4128.0,
+        "kd_epochs": 5
+      },
+      "active_learning": {
+        "seconds": 1237.79,
+        "peak_cpu_rss_mb": 3685.8,
+        "peak_gpu_alloc_mb": 3950.0,
+        "peak_gpu_reserved_mb": 4170.0,
+        "rounds": 6
+      },
+      "inference": {
+        "seconds": 36.81,
+        "peak_cpu_rss_mb": 3685.8,
+        "peak_gpu_alloc_mb": 2149.5,
+        "peak_gpu_reserved_mb": 4172.0,
+        "patches": 2376,
+        "images": 24,
+        "patches_per_second": 64.5,
+        "seconds_per_image": 1.534
+      }
+    },
+    "peak_gpu_alloc_mb": 3950.0,
+    "peak_cpu_rss_mb": 3685.8,
+    "total_seconds": 1594.2
+  },
+  "student_parameters": 31042369,
+  "teacher_parameters": [
+    31042369,
+    31042369,
+    31042369,
+    31042369
+  ],
+  "checkpoint": "/home/user/connectomics/MACS-2/models/experiments/macs_v2/kasthuri-pp/macs-v11-norestart-acq-laplace_b5_s2.pth"
+}