diff --git a/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt b/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..7099e19201b0c2d4845486ee087c6b359343f451 --- /dev/null +++ b/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt @@ -0,0 +1,107 @@ +[03:49:50][ 0s][START] target=dhanyasi-P14-mouse-cerebellum variant=macs-v11-norestart-acq-entropy budget=5.0% seed=1 | {"task": "membrane", "target": "dhanyasi-P14-mouse-cerebellum", "variant": "macs-v11-norestart-acq-entropy", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[03:49:50][ 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. +[03:49:51][ 0s][DATA] dhanyasi-P14-mouse-cerebellum: train=8649 val=1922 test=2883 patches +[03:49:53][ 2s][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=4.07 +[03:49:53][ 3s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[03:50:00][ 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 +[03:50:00][ 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 +[03:51:14][ 84s][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 +[03:51:14][ 84s][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 +[03:51:14][ 84s][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 +[03:51:14][ 84s][INIT] student initialised from teacher microns (weight 0.417) +[03:51:18][ 88s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[03:51:19][ 88s][KD] distilling the mixture ensemble (prob space) into the student +[03:55:24][ 333s][KD] soft labels for 6000 patches from 8 teachers in 245s +[03:58:22][ 512s][KD] epoch 1/5 | patches=6000 | loss=0.5816 | 179s +[04:01:21][ 691s][KD] epoch 2/5 | patches=6000 | loss=0.4493 | 179s +[04:04:20][ 870s][KD] epoch 3/5 | patches=6000 | loss=0.4312 | 179s +[04:07:19][ 1048s][KD] epoch 4/5 | patches=6000 | loss=0.4244 | 179s +[04:10:18][ 1227s][KD] epoch 5/5 | patches=6000 | loss=0.4237 | 179s +[04:10:24][ 1233s][ROUND0] before any target label: val_dice=0.1287 val_vi=0.5096 +[04:10:24][ 1233s][AL] patches=8649 (pool 8647 after dropping blank; free blank negatives: True) | budget=432 (5.0%) | rounds=6 | per_round=72 | acquisition=entropy | first_round=stratified +[04:11:06][ 1276s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': -0.5377973366328569, 'auroc_top20_error': 0.224696484375} +[04:12:28][ 1358s][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 +[04:12:42][ 1372s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.548, uniform=0.270, top1=0.481, top2=0.492, top3=0.499 -> using mixture +[04:12:45][ 1374s][WEIGHTS] round 1: student kept (labeled Dice 0.242 vs best teacher mouse-brain 0.481); distilling the corrected ensemble into it +[04:12:45][ 1374s][WEIGHTS] round 1: teacher Dice on 72 labeled patches + 2 blank (1139017 fg px): c-elegans=0.423, flywire=0.157, berghia=0.194, microns=0.305, oct-glia=0.168, oct-sfl=0.436, snemi=0.411, mouse-brain=0.481 | weights: c-elegans=0.079, flywire=0.017, berghia=0.002, microns=0.200, oct-glia=0.005, oct-sfl=0.095, snemi=0.147, mouse-brain=0.456 +[04:16:49][ 1619s][KD] soft labels for 6000 patches from 8 teachers in 245s +[04:19:48][ 1798s][KD] epoch 1/2 | patches=6000 | loss=0.5616 | 179s +[04:22:47][ 1977s][KD] epoch 2/2 | patches=6000 | loss=0.5540 | 179s +[04:24:09][ 2059s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[04:24:16][ 2066s][FT] round 1/6 epoch 1/5 | patches=74 | loss=1.3421 +[04:24:23][ 2073s][FT] round 1/6 epoch 2/5 | patches=74 | loss=1.3098 +[04:24:30][ 2079s][FT] round 1/6 epoch 3/5 | patches=74 | loss=1.2440 +[04:24:37][ 2086s][FT] round 1/6 epoch 4/5 | patches=74 | loss=1.2534 +[04:24:43][ 2093s][FT] round 1/6 epoch 5/5 | patches=74 | loss=1.2370 +[04:24:49][ 2099s][AL] round 1/6 | labeled=72 | selection_seconds=124.6000 round_seconds=859.5000 restart_student_dice=0.2421 restart_teacher_dice=0.4806 val_dice=0.6204 val_vi=0.6482 val_ece=0.0404 +[04:25:31][ 2141s][AL] round 2: 824/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[04:25:56][ 2166s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.388, uniform=0.184, top1=0.368, top2=0.399, top3=0.425 -> using top3 +[04:25:56][ 2166s][WEIGHTS] round 2: teacher Dice on 144 labeled patches + 2 blank (1961159 fg px): c-elegans=0.312, flywire=0.118, berghia=0.162, microns=0.220, oct-glia=0.106, oct-sfl=0.316, snemi=0.360, mouse-brain=0.368 | weights: c-elegans=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.108, snemi=0.335, mouse-brain=0.557 +[04:26:16][ 2186s][CONS] round 2: on the 72 new patches student Dice=0.507 vs ensemble 0.287 -> pseudo-labels from the student | consistency x0.02 +[04:26:28][ 2197s][FT] round 2/6 epoch 1/5 | patches=146 | loss=1.1134 +[04:26:38][ 2208s][FT] round 2/6 epoch 2/5 | patches=146 | loss=0.9894 +[04:26:50][ 2219s][FT] round 2/6 epoch 3/5 | patches=146 | loss=0.9750 +[04:27:00][ 2230s][FT] round 2/6 epoch 4/5 | patches=146 | loss=0.9333 +[04:27:11][ 2241s][FT] round 2/6 epoch 5/5 | patches=146 | loss=0.9466 +[04:27:17][ 2247s][AL] round 2/6 | labeled=144 | selection_seconds=42.1000 round_seconds=142.0000 pseudo_student_dice=0.5069 pseudo_ensemble_dice=0.2872 consistency_scale=0.0230 val_dice=0.5836 val_vi=0.6030 val_ece=0.0512 +[04:27:59][ 2289s][AL] round 3: 1378/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[04:30:53][ 2463s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.380, flywire=0.188, berghia=0.176, microns=0.335, oct-glia=0.214, oct-sfl=0.499, snemi=0.431, mouse-brain=0.452 | blended with labeled Dice at 0.50 +[04:33:37][ 2626s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.455, uniform=0.218, top1=0.465, top2=0.474, top3=0.473 -> using top3 +[04:33:37][ 2626s][WEIGHTS] round 3: teacher Dice on 216 labeled patches + 2 blank (3836186 fg px): c-elegans=0.368, flywire=0.136, berghia=0.159, microns=0.249, oct-glia=0.137, oct-sfl=0.363, snemi=0.372, mouse-brain=0.465 | weights: c-elegans=0.085, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.000, snemi=0.186, mouse-brain=0.729 +[04:33:47][ 2636s][CONS] round 3: on the 72 new patches student Dice=0.584 vs ensemble 0.653 -> pseudo-labels from the ensemble | consistency x1.00 +[04:34:01][ 2651s][FT] round 3/6 epoch 1/5 | patches=218 | loss=1.4325 +[04:34:16][ 2666s][FT] round 3/6 epoch 2/5 | patches=218 | loss=1.4260 +[04:34:32][ 2681s][FT] round 3/6 epoch 3/5 | patches=218 | loss=1.2945 +[04:34:47][ 2696s][FT] round 3/6 epoch 4/5 | patches=218 | loss=1.2789 +[04:35:02][ 2712s][FT] round 3/6 epoch 5/5 | patches=218 | loss=1.2626 +[04:35:08][ 2718s][AL] round 3/6 | labeled=216 | selection_seconds=42.1000 round_seconds=464.8000 pseudo_student_dice=0.5842 pseudo_ensemble_dice=0.6535 consistency_scale=1.0000 val_dice=0.6468 val_vi=0.5581 val_ece=0.0517 +[04:35:50][ 2760s][AL] round 4: 1289/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[04:36:36][ 2806s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.434, uniform=0.133, top1=0.451, top2=0.451, top3=0.450 -> using top3 +[04:36:36][ 2806s][WEIGHTS] round 4: teacher Dice on 288 labeled patches + 2 blank (4894951 fg px): c-elegans=0.358, flywire=0.098, berghia=0.171, microns=0.172, oct-glia=0.055, oct-sfl=0.259, snemi=0.338, mouse-brain=0.451 | weights: c-elegans=0.092, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.000, snemi=0.158, mouse-brain=0.750 +[04:36:56][ 2826s][CONS] round 4: on the 72 new patches student Dice=0.644 vs ensemble 0.369 -> pseudo-labels from the student | consistency x0.48 +[04:37:15][ 2845s][FT] round 4/6 epoch 1/5 | patches=290 | loss=0.9723 +[04:37:34][ 2864s][FT] round 4/6 epoch 2/5 | patches=290 | loss=0.9621 +[04:37:54][ 2883s][FT] round 4/6 epoch 3/5 | patches=290 | loss=0.9220 +[04:38:13][ 2903s][FT] round 4/6 epoch 4/5 | patches=290 | loss=0.9362 +[04:38:33][ 2922s][FT] round 4/6 epoch 5/5 | patches=290 | loss=0.9013 +[04:38:39][ 2929s][AL] round 4/6 | labeled=288 | selection_seconds=42.1000 round_seconds=204.8000 pseudo_student_dice=0.6440 pseudo_ensemble_dice=0.3688 consistency_scale=0.4800 val_dice=0.6459 val_vi=0.5368 val_ece=0.0249 +[04:39:21][ 2971s][AL] round 5: 1384/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[04:42:25][ 3155s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.388, flywire=0.168, berghia=0.173, microns=0.320, oct-glia=0.212, oct-sfl=0.460, snemi=0.435, mouse-brain=0.471 | blended with labeled Dice at 0.50 +[04:45:20][ 3329s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.473, uniform=0.222, top1=0.462, top2=0.473, top3=0.472 -> using mixture +[04:45:20][ 3329s][WEIGHTS] round 5: teacher Dice on 360 labeled patches + 2 blank (6110421 fg px): c-elegans=0.380, flywire=0.136, berghia=0.177, microns=0.254, oct-glia=0.133, oct-sfl=0.368, snemi=0.396, mouse-brain=0.462 | weights: c-elegans=0.069, flywire=0.018, berghia=0.002, microns=0.161, oct-glia=0.005, oct-sfl=0.065, snemi=0.170, mouse-brain=0.510 +[04:45:41][ 3350s][CONS] round 5: on the 72 new patches student Dice=0.607 vs ensemble 0.484 -> pseudo-labels from the student | consistency x0.36 +[04:46:04][ 3374s][FT] round 5/6 epoch 1/5 | patches=362 | loss=0.8871 +[04:46:28][ 3397s][FT] round 5/6 epoch 2/5 | patches=362 | loss=0.8923 +[04:46:51][ 3421s][FT] round 5/6 epoch 3/5 | patches=362 | loss=0.8688 +[04:47:14][ 3444s][FT] round 5/6 epoch 4/5 | patches=362 | loss=0.8893 +[04:47:38][ 3467s][FT] round 5/6 epoch 5/5 | patches=362 | loss=0.8411 +[04:47:44][ 3473s][AL] round 5/6 | labeled=360 | selection_seconds=42.1000 round_seconds=538.7000 pseudo_student_dice=0.6073 pseudo_ensemble_dice=0.4837 consistency_scale=0.3580 val_dice=0.6549 val_vi=0.5238 val_ece=0.0051 +[04:48:26][ 3515s][AL] round 6: 1422/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[04:51:36][ 3705s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.400, flywire=0.171, berghia=0.184, microns=0.326, oct-glia=0.207, oct-sfl=0.477, snemi=0.435, mouse-brain=0.478 | blended with labeled Dice at 0.50 +[04:54:35][ 3885s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.480, uniform=0.220, top1=0.469, top2=0.481, top3=0.481 -> using mixture +[04:54:35][ 3885s][WEIGHTS] round 6: teacher Dice on 432 labeled patches + 2 blank (7377505 fg px): c-elegans=0.392, flywire=0.135, berghia=0.185, microns=0.255, oct-glia=0.129, oct-sfl=0.373, snemi=0.403, mouse-brain=0.469 | weights: c-elegans=0.074, flywire=0.017, berghia=0.002, microns=0.153, oct-glia=0.004, oct-sfl=0.064, snemi=0.173, mouse-brain=0.513 +[04:54:56][ 3906s][CONS] round 6: on the 72 new patches student Dice=0.659 vs ensemble 0.462 -> pseudo-labels from the student | consistency x0.53 +[04:55:24][ 3933s][FT] round 6/6 epoch 1/5 | patches=434 | loss=0.8836 +[04:55:51][ 3961s][FT] round 6/6 epoch 2/5 | patches=434 | loss=0.8524 +[04:56:19][ 3988s][FT] round 6/6 epoch 3/5 | patches=434 | loss=0.8450 +[04:56:47][ 4016s][FT] round 6/6 epoch 4/5 | patches=434 | loss=0.8465 +[04:57:14][ 4044s][FT] round 6/6 epoch 5/5 | patches=434 | loss=0.8346 +[04:57:20][ 4050s][AL] round 6/6 | labeled=432 | selection_seconds=42.2000 round_seconds=570.5000 pseudo_student_dice=0.6586 pseudo_ensemble_dice=0.4617 consistency_scale=0.5290 val_dice=0.6593 val_vi=0.5305 val_ece=0.0107 +[04:57:31][ 4060s][CONS] final consolidation: pseudo-labels refreshed from the student +[04:57:58][ 4088s][FT] final 1/10 epoch 1/1 | patches=434 | loss=0.8176 +[04:58:26][ 4116s][FT] final 2/10 epoch 1/1 | patches=434 | loss=0.8093 +[04:58:54][ 4143s][FT] final 3/10 epoch 1/1 | patches=434 | loss=0.7822 +[04:59:22][ 4171s][FT] final 4/10 epoch 1/1 | patches=434 | loss=0.7432 +[04:59:49][ 4199s][FT] final 5/10 epoch 1/1 | patches=434 | loss=0.7414 +[05:00:17][ 4227s][FT] final 6/10 epoch 1/1 | patches=434 | loss=0.6925 +[05:00:45][ 4255s][FT] final 7/10 epoch 1/1 | patches=434 | loss=0.6934 +[05:01:12][ 4282s][FT] final 8/10 epoch 1/1 | patches=434 | loss=0.6712 +[05:01:41][ 4310s][FT] final 9/10 epoch 1/1 | patches=434 | loss=0.6439 +[05:02:08][ 4338s][FT] final 10/10 epoch 1/1 | patches=434 | loss=0.6502 +[05:02:14][ 4344s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=287.9000 val_dice=0.6657 val_vi=0.5277 val_ece=0.0145 +[05:02:14][ 4344s][WEIGHTS] final: c-elegans=0.074, flywire=0.017, berghia=0.002, microns=0.153, oct-glia=0.004, oct-sfl=0.064, snemi=0.173, mouse-brain=0.513 +[05:04:29][ 4479s][TEST] raw student (no EMA, no TTA): patch_dice=0.6755 patch_vi=0.4444 +[05:04:45][ 4495s][TEST] final model without TTA: patch_dice=0.6755 patch_vi=0.4444 +[05:04:45][ 4495s][RESOURCES] {"phases": {"setup": {"seconds": 1.57, "peak_cpu_rss_mb": 2160.6, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 85.57, "peak_cpu_rss_mb": 4184.1, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 1139.29, "peak_cpu_rss_mb": 5957.0, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4598.0, "kd_epochs": 5}, "active_learning": {"seconds": 3110.53, "peak_cpu_rss_mb": 5957.0, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4642.0, "rounds": 6}, "inference": {"seconds": 150.79, "peak_cpu_rss_mb": 5957.0, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4644.0, "patches": 2883, "images": 3, "patches_per_second": 19.1, "seconds_per_image": 50.263}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5957.0, "total_seconds": 4487.8} +[05:04:45][ 4495s][TEST] patch_dice=0.7316 patch_iou=0.6304 patch_precision=0.7851 patch_recall=0.7817 patch_vi=0.4133 image_dice=0.8190 image_iou=0.6967 image_precision=0.8642 image_recall=0.7816 image_vi=0.4620 cal_ece=0.0032 cal_nll=0.1825 cal_brier=0.0543 cal_mean_confidence=0.9228 cal_accuracy=0.9259 cal_predicted_prevalence=0.2157 cal_true_prevalence=0.2301 cal_mean_probability=0.2249 +[05:04:46][ 4495s][DONE] dhanyasi-P14-mouse-cerebellum macs-v11-norestart-acq-entropy b5 s1 | test patch Dice=0.7316 VI=0.4133 | image Dice=0.8190 VI=0.4620 | 4495s diff --git a/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json b/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..a610b9b7c1c33e5b247508535535d850e7a4b73b --- /dev/null +++ b/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json @@ -0,0 +1,776 @@ +{ + "target": "dhanyasi-P14-mouse-cerebellum", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 1, + "config": { + "task": "membrane", + "target": "dhanyasi-P14-mouse-cerebellum", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "dhanyasi-P14-mouse-cerebellum", + "--variants", + "macs-v11-norestart-acq-entropy", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.05083445087713844, + 0.15580217281625375, + 0.01024336472294703, + 0.4172897252912645, + 0.039642121317843636, + 0.05395053407082265, + 0.10715675218847985, + 0.1650808787152502 + ], + "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": [ + 0.05083445087713844, + 0.15580217281625375, + 0.01024336472294703, + 0.4172897252912645, + 0.039642121317843636, + 0.05395053407082265, + 0.10715675218847985, + 0.1650808787152502 + ], + "per_source": [ + 0.18317784495730796, + 0.0008117673075475437, + 0.07130128976808368, + 0.2901325749523702, + 0.19474134185881758, + 0.1442779689307106, + 0.11165161174087061, + 0.003905600484291838 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.04944686886108997, + "mmd2_mixture": 0.06619593169520155, + "mmd2_uniform": 0.10385689512746193, + "mmd2_single_source": [ + 0.3515701146669005, + 0.19141235024309128, + 0.2296475179170791, + 0.08478110169677128, + 0.27262177079818134, + 0.24630850341824578, + 0.1500490344627728, + 0.17311817557273296 + ], + "median_squared_distance": 0.18230580278935338, + "gammas": [ + 1.3713222298736392, + 2.7426444597472783, + 5.485288919494557, + 10.970577838989113, + 21.941155677978227 + ], + "max_weight": 0.4172897252912645, + "entropy": 1.6749062834111577, + "effective_sources": 5.338294838847327, + "min_weight": 0.01024336472294703 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.24983786823863818, + 0.6076264103713149, + 0.31213514946535237, + 0.2194743346098349, + 0.24579617653258468, + 0.265599036130135, + 0.28252505597774746, + 0.503904484908046 + ], + "standardized": [ + -0.6514551732920085, + 2.0580443558260613, + -0.17968362593338755, + -0.8813954044106882, + -0.6820625306688456, + -0.5320973069785436, + -0.4039181245665406, + 1.2725678100239537 + ], + "tau": 0.5, + "max_weight": 0.2901325749523702, + "entropy": 1.7283714073603158, + "effective_sources": 5.63147505732349, + "min_weight": 0.0008117673075475437 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "microns_basil", + "seconds": 6.6 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.4128695009061776, + 0.10650556878992745, + 0.15770951130321542, + 0.22267897752458055, + 0.11007038183068829, + 0.2906751954472746, + 0.2970256007290182, + 0.44753146510338837 + ], + "ensemble_val_dice": { + "mixture": 0.20976392808829658, + "mixture@raw": 0.2201044151609466, + "mixture@logit": 0.18899978060037625, + "per_source": 0.193177417318676, + "per_source@raw": 0.20883301715723235, + "per_source@logit": 0.15551528847303686, + "uniform": 0.19238477463512638, + "uniform@raw": 0.2103392831498283, + "uniform@logit": 0.15367871932302002 + }, + "val_patches": 1024 + }, + "temperatures": [ + 2.575598955154419, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 2.642308473587036, + 3.547258138656616, + 4.074584007263184 + ], + "top_source": "microns_basil", + "weight_stats": { + "max_weight": 0.4172897252912645, + "entropy": 1.6749062834111577, + "effective_sources": 5.338294838847327, + "min_weight": 0.01024336472294703 + }, + "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": 1139.3, + "round0_val": { + "val_dice": 0.12873302341577902, + "val_vi": 0.5095556457765907, + "val_ece": 0.1409334247270805 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 72, + "selection_seconds": 124.6, + "round_seconds": 859.5, + "teacher_dice_labeled": [ + 0.423321230947482, + 0.15747034038174015, + 0.1937080422460344, + 0.30535734387879976, + 0.16796457367489764, + 0.4355867030005887, + 0.4106488934816602, + 0.480569706780208 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1139017, + "evidence_nonempty_patches": 61, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.24208465449533073, + "restart_teacher_dice": 0.480569706780208, + "weights": [ + 0.07916149016146976, + 0.016996268628103182, + 0.0016054657774156285, + 0.19974819501005328, + 0.004803002789714615, + 0.09497730470154105, + 0.14700770716957254, + 0.4557005657621298 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.547725247586474, + "uniform": 0.2700913480165708, + "top1": 0.480569706780208, + "top2": 0.4919855915340812, + "top3": 0.4986229536904216 + }, + "val_dice": 0.6203974319004413, + "val_vi": 0.6482077865842528, + "val_ece": 0.04038987229168493 + }, + { + "round": 2, + "n_labeled": 144, + "selection_seconds": 42.1, + "round_seconds": 142.0, + "teacher_dice_labeled": [ + 0.3121220380549426, + 0.11784411213032195, + 0.16213463696136268, + 0.21973314935568444, + 0.10574009869144996, + 0.3155480158339058, + 0.3603575734043396, + 0.3678067279431099 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1961159, + "evidence_nonempty_patches": 104, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.10788468470323365, + 0.3354200554780413, + 0.5566952598187249 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.3879878504056142, + "uniform": 0.18391130588268922, + "top1": 0.3678067279431099, + "top2": 0.39948449736175295, + "top3": 0.4246588769388564 + }, + "pseudo_student_dice": 0.506870089555582, + "pseudo_ensemble_dice": 0.28715103968797784, + "pseudo_source": "student", + "consistency_scale": 0.023, + "val_dice": 0.5836245726576526, + "val_vi": 0.6030459611865171, + "val_ece": 0.05115153999860124 + }, + { + "round": 3, + "n_labeled": 216, + "selection_seconds": 42.1, + "round_seconds": 464.8, + "teacher_dice_labeled": [ + 0.3678614412663967, + 0.13648598384761942, + 0.15898055108815634, + 0.24919594486654167, + 0.13671349199434196, + 0.3634301167820999, + 0.37207817186759584, + 0.4654090843012749 + ], + "teacher_dice_pool": [ + 0.38011472038533395, + 0.18782671866421538, + 0.1762019359249727, + 0.335320945586393, + 0.2141001779587749, + 0.4985947426940896, + 0.43097701284546114, + 0.4516571637763781 + ], + "evidence_fg_pixels": 3836186, + "evidence_nonempty_patches": 172, + "first_correction": false, + "restart": "", + "weights": [ + 0.08465750023050535, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.18614007060340373, + 0.729202429166091 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.4550915721689207, + "uniform": 0.21818578097968497, + "top1": 0.4654090843012749, + "top2": 0.474151057114913, + "top3": 0.4729733104091749 + }, + "pseudo_student_dice": 0.5842168055752709, + "pseudo_ensemble_dice": 0.6534647165896713, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.6467763653471634, + "val_vi": 0.558056933432851, + "val_ece": 0.05167226492881565 + }, + { + "round": 4, + "n_labeled": 288, + "selection_seconds": 42.1, + "round_seconds": 204.8, + "teacher_dice_labeled": [ + 0.35821635711104094, + 0.09778125362066968, + 0.17137258333636496, + 0.1724433457499402, + 0.05513871750205011, + 0.2586880798396885, + 0.3382993663944764, + 0.4506664101801657 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 4894951, + "evidence_nonempty_patches": 240, + "first_correction": false, + "restart": "", + "weights": [ + 0.09163584339119596, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.1582808577862576, + 0.7500832988225464 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.43443073868488324, + "uniform": 0.1334549874746405, + "top1": 0.4506664101801657, + "top2": 0.4506745268961577, + "top3": 0.45005077329606585 + }, + "pseudo_student_dice": 0.6440377318350516, + "pseudo_ensemble_dice": 0.3687653529910971, + "pseudo_source": "student", + "consistency_scale": 0.48, + "val_dice": 0.6458938731498896, + "val_vi": 0.5368060920046469, + "val_ece": 0.024876896241767484 + }, + { + "round": 5, + "n_labeled": 360, + "selection_seconds": 42.1, + "round_seconds": 538.7, + "teacher_dice_labeled": [ + 0.38013299125106925, + 0.1355929442454303, + 0.17720215081782453, + 0.2539448023511426, + 0.13333927029771672, + 0.3683974998441432, + 0.39580615773577776, + 0.462093229389536 + ], + "teacher_dice_pool": [ + 0.38762516986152434, + 0.16847478642657696, + 0.17322085467147122, + 0.3200939826971463, + 0.211538626877524, + 0.4602015716310891, + 0.4350030382843811, + 0.4713753858291866 + ], + "evidence_fg_pixels": 6110421, + "evidence_nonempty_patches": 310, + "first_correction": false, + "restart": "", + "weights": [ + 0.06914616887588221, + 0.018372132368569834, + 0.0018311974724859205, + 0.16070128006889814, + 0.004570412095666965, + 0.06525880886321075, + 0.17048933831294702, + 0.5096306619423392 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.4731747180910907, + "uniform": 0.22201451964945562, + "top1": 0.462093229389536, + "top2": 0.47332982939518164, + "top3": 0.47240656144828863 + }, + "pseudo_student_dice": 0.6073316378424809, + "pseudo_ensemble_dice": 0.483669413330306, + "pseudo_source": "student", + "consistency_scale": 0.358, + "val_dice": 0.6548823077127274, + "val_vi": 0.5238367298145632, + "val_ece": 0.005050470148968892 + }, + { + "round": 6, + "n_labeled": 432, + "selection_seconds": 42.2, + "round_seconds": 570.5, + "teacher_dice_labeled": [ + 0.3924474982027606, + 0.13520125388940998, + 0.18469172221338315, + 0.25488349411404726, + 0.1294684251406094, + 0.3729135753452115, + 0.40291120307243233, + 0.46865716029759097 + ], + "teacher_dice_pool": [ + 0.3999496986695009, + 0.17062220618934187, + 0.18388714703370632, + 0.325641113187999, + 0.20701207561780072, + 0.47749391841243344, + 0.43473254729499794, + 0.4778808675544609 + ], + "evidence_fg_pixels": 7377505, + "evidence_nonempty_patches": 375, + "first_correction": false, + "restart": "", + "weights": [ + 0.07373789940615798, + 0.017254401160920353, + 0.001860815823363887, + 0.15294575848580463, + 0.004145585803265005, + 0.06437149923884866, + 0.17258198504016561, + 0.513102055041474 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.479814058220684, + "uniform": 0.2197636441812369, + "top1": 0.46865716029759097, + "top2": 0.48056416983213823, + "top3": 0.48105135534617427 + }, + "pseudo_student_dice": 0.6586200443659856, + "pseudo_ensemble_dice": 0.4617141635573474, + "pseudo_source": "student", + "consistency_scale": 0.529, + "val_dice": 0.6593283705355009, + "val_vi": 0.5304568455956489, + "val_ece": 0.01068904340930946 + }, + { + "round": 7, + "n_labeled": 432, + "selection_seconds": 0.0, + "round_seconds": 287.9, + "final_consolidation": true, + "val_dice": 0.6656920404478153, + "val_vi": 0.5277453584741619, + "val_ece": 0.014463817565412058 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.5377973366328569, + "auroc_top20_error": 0.224696484375 + }, + "al_seconds": 3110.5, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.07373789940615798, + 0.017254401160920353, + 0.001860815823363887, + 0.15294575848580463, + 0.004145585803265005, + 0.06437149923884866, + 0.17258198504016561, + 0.513102055041474 + ], + "weight_stats_final": { + "max_weight": 0.513102055041474, + "entropy": 1.4060874658688964, + "effective_sources": 4.079961148192515, + "min_weight": 0.001860815823363887 + }, + "top_source_final": "whole-mouse-brain", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.675490319683645, + "patch_iou": 0.568467878690645, + "patch_precision": 0.7172071866334969, + "patch_recall": 0.7691740397507477, + "patch_vi": 0.44435762694673336, + "n_patches": 2883, + "cal_ece": 0.011123024071693167, + "cal_nll": 0.20394394839630497, + "cal_brier": 0.060213044436333046, + "cal_mean_confidence": 0.9296774405866081, + "cal_accuracy": 0.918554416514915, + "cal_predicted_prevalence": 0.21542230315643426, + "cal_true_prevalence": 0.23013442215140478, + "cal_mean_probability": 0.22161547074418342 + }, + "test_notta": { + "patch_dice": 0.675490319683645, + "patch_iou": 0.568467878690645, + "patch_precision": 0.7172071866334969, + "patch_recall": 0.7691740397507477, + "patch_vi": 0.44435762694673336, + "n_patches": 2883, + "cal_ece": 0.011123024071693167, + "cal_nll": 0.20394394839630497, + "cal_brier": 0.060213044436333046, + "cal_mean_confidence": 0.9296774405866081, + "cal_accuracy": 0.918554416514915, + "cal_predicted_prevalence": 0.21542230315643426, + "cal_true_prevalence": 0.23013442215140478, + "cal_mean_probability": 0.22161547074418342 + }, + "test": { + "patch_dice": 0.7316377151657691, + "patch_iou": 0.6303840026333201, + "patch_precision": 0.7850571730891608, + "patch_recall": 0.7817328455606284, + "patch_vi": 0.4133368944747202, + "n_patches": 2883, + "image_dice": 0.8189727425328098, + "image_iou": 0.6966907438405338, + "image_precision": 0.8641969001002137, + "image_recall": 0.7815801609355528, + "image_vi": 0.4620376655571004, + "n_images": 3, + "cal_ece": 0.0031798602784341846, + "cal_nll": 0.18247887747829966, + "cal_brier": 0.05429650188685861, + "cal_mean_confidence": 0.9227941310674863, + "cal_accuracy": 0.9259042094823101, + "cal_predicted_prevalence": 0.21568583615157821, + "cal_true_prevalence": 0.23013442215140478, + "cal_mean_probability": 0.22487676152217542 + }, + "eval_seconds": 150.8, + "total_seconds": 4494.7, + "resources": { + "phases": { + "setup": { + "seconds": 1.57, + "peak_cpu_rss_mb": 2160.6, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 85.57, + "peak_cpu_rss_mb": 4184.1, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 1139.29, + "peak_cpu_rss_mb": 5957.0, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4598.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 3110.53, + "peak_cpu_rss_mb": 5957.0, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4642.0, + "rounds": 6 + }, + "inference": { + "seconds": 150.79, + "peak_cpu_rss_mb": 5957.0, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4644.0, + "patches": 2883, + "images": 3, + "patches_per_second": 19.1, + "seconds_per_image": 50.263 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5957.0, + "total_seconds": 4487.8 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..b0c7e214e663bc4fe9e0023d20b3db1ff980408d --- /dev/null +++ b/macs_v2_r26/runs/membrane/dhanyasi-P14-mouse-cerebellum/macs-v11-norestart-acq-entropy/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 +1,72,124.6,859.5,"[0.423321230947482, 0.15747034038174015, 0.1937080422460344, 0.30535734387879976, 0.16796457367489764, 0.4355867030005887, 0.4106488934816602, 0.480569706780208]",,1139017,61,True,kept,0.24208465449533073,0.480569706780208,"[0.07916149016146976, 0.016996268628103182, 0.0016054657774156285, 0.19974819501005328, 0.004803002789714615, 0.09497730470154105, 0.14700770716957254, 0.4557005657621298]",mixture,"{'mixture': 0.547725247586474, 'uniform': 0.2700913480165708, 'top1': 0.480569706780208, 'top2': 0.4919855915340812, 'top3': 0.4986229536904216}",0.6203974319004413,0.6482077865842528,0.04038987229168493,,,,, +2,144,42.1,142.0,"[0.3121220380549426, 0.11784411213032195, 0.16213463696136268, 0.21973314935568444, 0.10574009869144996, 0.3155480158339058, 0.3603575734043396, 0.3678067279431099]",,1961159,104,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.10788468470323365, 0.3354200554780413, 0.5566952598187249]",top3,"{'mixture': 0.3879878504056142, 'uniform': 0.18391130588268922, 'top1': 0.3678067279431099, 'top2': 0.39948449736175295, 'top3': 0.4246588769388564}",0.5836245726576526,0.6030459611865171,0.05115153999860124,0.506870089555582,0.28715103968797784,student,0.023, +3,216,42.1,464.8,"[0.3678614412663967, 0.13648598384761942, 0.15898055108815634, 0.24919594486654167, 0.13671349199434196, 0.3634301167820999, 0.37207817186759584, 0.4654090843012749]","[0.38011472038533395, 0.18782671866421538, 0.1762019359249727, 0.335320945586393, 0.2141001779587749, 0.4985947426940896, 0.43097701284546114, 0.4516571637763781]",3836186,172,False,,,,"[0.08465750023050535, 0.0, 0.0, 0.0, 0.0, 0.0, 0.18614007060340373, 0.729202429166091]",top3,"{'mixture': 0.4550915721689207, 'uniform': 0.21818578097968497, 'top1': 0.4654090843012749, 'top2': 0.474151057114913, 'top3': 0.4729733104091749}",0.6467763653471634,0.558056933432851,0.05167226492881565,0.5842168055752709,0.6534647165896713,ensemble,1.0, +4,288,42.1,204.8,"[0.35821635711104094, 0.09778125362066968, 0.17137258333636496, 0.1724433457499402, 0.05513871750205011, 0.2586880798396885, 0.3382993663944764, 0.4506664101801657]",,4894951,240,False,,,,"[0.09163584339119596, 0.0, 0.0, 0.0, 0.0, 0.0, 0.1582808577862576, 0.7500832988225464]",top3,"{'mixture': 0.43443073868488324, 'uniform': 0.1334549874746405, 'top1': 0.4506664101801657, 'top2': 0.4506745268961577, 'top3': 0.45005077329606585}",0.6458938731498896,0.5368060920046469,0.024876896241767484,0.6440377318350516,0.3687653529910971,student,0.48, +5,360,42.1,538.7,"[0.38013299125106925, 0.1355929442454303, 0.17720215081782453, 0.2539448023511426, 0.13333927029771672, 0.3683974998441432, 0.39580615773577776, 0.462093229389536]","[0.38762516986152434, 0.16847478642657696, 0.17322085467147122, 0.3200939826971463, 0.211538626877524, 0.4602015716310891, 0.4350030382843811, 0.4713753858291866]",6110421,310,False,,,,"[0.06914616887588221, 0.018372132368569834, 0.0018311974724859205, 0.16070128006889814, 0.004570412095666965, 0.06525880886321075, 0.17048933831294702, 0.5096306619423392]",mixture,"{'mixture': 0.4731747180910907, 'uniform': 0.22201451964945562, 'top1': 0.462093229389536, 'top2': 0.47332982939518164, 'top3': 0.47240656144828863}",0.6548823077127274,0.5238367298145632,0.005050470148968892,0.6073316378424809,0.483669413330306,student,0.358, +6,432,42.2,570.5,"[0.3924474982027606, 0.13520125388940998, 0.18469172221338315, 0.25488349411404726, 0.1294684251406094, 0.3729135753452115, 0.40291120307243233, 0.46865716029759097]","[0.3999496986695009, 0.17062220618934187, 0.18388714703370632, 0.325641113187999, 0.20701207561780072, 0.47749391841243344, 0.43473254729499794, 0.4778808675544609]",7377505,375,False,,,,"[0.07373789940615798, 0.017254401160920353, 0.001860815823363887, 0.15294575848580463, 0.004145585803265005, 0.06437149923884866, 0.17258198504016561, 0.513102055041474]",mixture,"{'mixture': 0.479814058220684, 'uniform': 0.2197636441812369, 'top1': 0.46865716029759097, 'top2': 0.48056416983213823, 'top3': 0.48105135534617427}",0.6593283705355009,0.5304568455956489,0.01068904340930946,0.6586200443659856,0.4617141635573474,student,0.529, +7,432,0.0,287.9,,,,,,,,,,,,0.6656920404478153,0.5277453584741619,0.014463817565412058,,,,,True diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..63e627d266f98620d7470521226d1526bc29831b --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt @@ -0,0 +1,108 @@ +[05:04:49][ 0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-entropy budget=5.0% seed=1 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-entropy", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[05:04:49][ 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. +[05:04:49][ 0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches +[05:04:51][ 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, snemi=3.55, mouse-brain=4.07 +[05:04:51][ 3s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[05:04:58][ 9s][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 +[05:04:58][ 9s][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 +[05:05:53][ 65s][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 +[05:05:53][ 65s][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 +[05:05:53][ 65s][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 +[05:05:53][ 65s][INIT] student initialised from teacher oct-glia (weight 0.636) +[05:05:57][ 68s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[05:05:57][ 68s][KD] distilling the mixture ensemble (prob space) into the student +[05:08:22][ 214s][KD] soft labels for 3584 patches from 8 teachers in 146s +[05:10:09][ 320s][KD] epoch 1/5 | patches=3584 | loss=0.5720 | 106s +[05:11:55][ 427s][KD] epoch 2/5 | patches=3584 | loss=0.4388 | 106s +[05:13:42][ 533s][KD] epoch 3/5 | patches=3584 | loss=0.4006 | 106s +[05:15:28][ 640s][KD] epoch 4/5 | patches=3584 | loss=0.3569 | 107s +[05:17:15][ 746s][KD] epoch 5/5 | patches=3584 | loss=0.3310 | 106s +[05:17:19][ 751s][ROUND0] before any target label: val_dice=0.8502 val_vi=0.1470 +[05:17:19][ 751s][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 +[05:17:32][ 763s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.44560086267226684, 'auroc_top20_error': 0.593673156908451} +[05:18:54][ 845s][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 +[05:19:03][ 854s][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 +[05:19:04][ 855s][WEIGHTS] round 1: student kept (labeled Dice 0.750 vs best teacher oct-glia 0.731); distilling the corrected ensemble into it +[05:19:04][ 855s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (237130 fg px): c-elegans=0.296, cerebellum=0.619, 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 +[05:20:38][ 949s][KD] soft labels for 2309 patches from 8 teachers in 94s +[05:21:47][ 1018s][KD] epoch 1/2 | patches=2309 | loss=0.7212 | 69s +[05:22:55][ 1087s][KD] epoch 2/2 | patches=2309 | loss=0.6145 | 69s +[05:24:17][ 1169s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[05:24:22][ 1173s][FT] round 1/6 epoch 1/5 | patches=58 | loss=1.4306 +[05:24:26][ 1177s][FT] round 1/6 epoch 2/5 | patches=58 | loss=1.3428 +[05:24:30][ 1182s][FT] round 1/6 epoch 3/5 | patches=58 | loss=1.3158 +[05:24:35][ 1186s][FT] round 1/6 epoch 4/5 | patches=58 | loss=1.2565 +[05:24:39][ 1190s][FT] round 1/6 epoch 5/5 | patches=58 | loss=1.3613 +[05:24:44][ 1195s][AL] round 1/6 | labeled=29 | selection_seconds=94.3000 round_seconds=439.9000 restart_student_dice=0.7496 restart_teacher_dice=0.7308 val_dice=0.3973 val_vi=0.1807 val_ece=0.0784 +[05:24:56][ 1207s][AL] round 2: 518/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[05:25:14][ 1225s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.588, uniform=0.611, top1=0.588, top2=0.587, top3=0.587 -> using uniform +[05:25:14][ 1225s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (646819 fg px): c-elegans=0.348, cerebellum=0.462, flywire=0.176, berghia=0.317, microns=0.481, oct-glia=0.588, snemi=0.179, mouse-brain=0.323 | 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 +[05:25:30][ 1242s][CONS] round 2: on the 29 new patches student Dice=0.611 vs ensemble 0.523 -> pseudo-labels from the student | consistency x0.37 +[05:25:38][ 1249s][FT] round 2/6 epoch 1/5 | patches=116 | loss=1.0340 +[05:25:46][ 1257s][FT] round 2/6 epoch 2/5 | patches=116 | loss=1.0324 +[05:25:54][ 1265s][FT] round 2/6 epoch 3/5 | patches=116 | loss=0.9938 +[05:26:01][ 1273s][FT] round 2/6 epoch 4/5 | patches=116 | loss=0.9853 +[05:26:09][ 1280s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.9211 +[05:26:14][ 1285s][AL] round 2/6 | labeled=58 | selection_seconds=12.5000 round_seconds=85.6000 pseudo_student_dice=0.6108 pseudo_ensemble_dice=0.5226 consistency_scale=0.3690 val_dice=0.6369 val_vi=0.1633 val_ece=0.0094 +[05:26:26][ 1297s][AL] round 3: 566/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[05:29:13][ 1465s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.441, cerebellum=0.666, flywire=0.237, berghia=0.399, microns=0.581, oct-glia=0.806, snemi=0.334, mouse-brain=0.310 | blended with labeled Dice at 0.50 +[05:31:51][ 1622s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.692, uniform=0.709, top1=0.692, top2=0.692, top3=0.691 -> using uniform +[05:31:51][ 1622s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (1020182 fg px): c-elegans=0.388, cerebellum=0.565, flywire=0.205, berghia=0.354, microns=0.525, oct-glia=0.692, snemi=0.257, mouse-brain=0.318 | 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 +[05:32:07][ 1639s][CONS] round 3: on the 29 new patches student Dice=0.685 vs ensemble 0.565 -> pseudo-labels from the student | consistency x0.62 +[05:32:18][ 1650s][FT] round 3/6 epoch 1/5 | patches=174 | loss=0.9333 +[05:32:29][ 1661s][FT] round 3/6 epoch 2/5 | patches=174 | loss=0.9096 +[05:32:41][ 1672s][FT] round 3/6 epoch 3/5 | patches=174 | loss=0.8885 +[05:32:52][ 1683s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.8886 +[05:33:03][ 1694s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.8893 +[05:33:07][ 1698s][AL] round 3/6 | labeled=87 | selection_seconds=12.3000 round_seconds=409.1000 pseudo_student_dice=0.6845 pseudo_ensemble_dice=0.5650 consistency_scale=0.6150 val_dice=0.8477 val_vi=0.1608 val_ece=0.0074 +[05:33:19][ 1710s][AL] round 4: 568/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[05:36:11][ 1882s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.447, cerebellum=0.668, flywire=0.234, berghia=0.398, microns=0.567, oct-glia=0.791, snemi=0.341, mouse-brain=0.320 | blended with labeled Dice at 0.50 +[05:38:54][ 2045s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.699, uniform=0.708, top1=0.698, top2=0.699, top3=0.698 -> using uniform +[05:38:54][ 2045s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1253998 fg px): c-elegans=0.367, cerebellum=0.569, flywire=0.206, berghia=0.348, microns=0.519, oct-glia=0.698, snemi=0.259, mouse-brain=0.300 | 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 +[05:39:10][ 2061s][CONS] round 4: on the 29 new patches student Dice=0.717 vs ensemble 0.649 -> pseudo-labels from the student | consistency x0.72 +[05:39:24][ 2076s][FT] round 4/6 epoch 1/5 | patches=232 | loss=0.8830 +[05:39:39][ 2090s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.8655 +[05:39:53][ 2104s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.8936 +[05:40:08][ 2119s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.8549 +[05:40:22][ 2133s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.7787 +[05:40:26][ 2138s][AL] round 4/6 | labeled=116 | selection_seconds=12.2000 round_seconds=434.8000 pseudo_student_dice=0.7174 pseudo_ensemble_dice=0.6485 consistency_scale=0.7250 val_dice=0.8907 val_vi=0.1518 val_ece=0.0106 +[05:40:38][ 2149s][AL] round 5: 572/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[05:43:34][ 2326s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.428, cerebellum=0.669, flywire=0.232, berghia=0.383, microns=0.566, oct-glia=0.815, snemi=0.326, mouse-brain=0.303 | blended with labeled Dice at 0.50 +[05:46:21][ 2492s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.698, uniform=0.702, top1=0.698, top2=0.699, top3=0.698 -> using mixture +[05:46:21][ 2492s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1654235 fg px): c-elegans=0.367, cerebellum=0.566, flywire=0.204, berghia=0.340, microns=0.509, oct-glia=0.698, snemi=0.254, mouse-brain=0.298 | weights: c-elegans=0.005, cerebellum=0.012, flywire=0.000, berghia=0.006, microns=0.007, oct-glia=0.968, snemi=0.000, mouse-brain=0.001 +[05:46:37][ 2508s][CONS] round 5: on the 29 new patches student Dice=0.783 vs ensemble 0.496 -> pseudo-labels from the student | consistency x0.94 +[05:46:55][ 2526s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.7943 +[05:47:13][ 2544s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.7149 +[05:47:31][ 2562s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.6657 +[05:47:49][ 2580s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.7386 +[05:48:07][ 2598s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.5905 +[05:48:11][ 2603s][AL] round 5/6 | labeled=145 | selection_seconds=12.0000 round_seconds=460.5000 pseudo_student_dice=0.7831 pseudo_ensemble_dice=0.4959 consistency_scale=0.9440 val_dice=0.8796 val_vi=0.1516 val_ece=0.0127 +[05:48:23][ 2614s][AL] round 6: 561/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[05:51:24][ 2796s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.432, cerebellum=0.673, flywire=0.240, berghia=0.391, microns=0.593, oct-glia=0.838, snemi=0.311, mouse-brain=0.294 | blended with labeled Dice at 0.50 +[05:54:16][ 2967s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.681, uniform=0.689, top1=0.680, top2=0.681, top3=0.680 -> using uniform +[05:54:16][ 2967s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2203858 fg px): c-elegans=0.356, cerebellum=0.551, flywire=0.203, berghia=0.333, microns=0.501, oct-glia=0.680, snemi=0.258, mouse-brain=0.289 | 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 +[05:54:32][ 2983s][CONS] round 6: on the 29 new patches student Dice=0.586 vs ensemble 0.339 -> pseudo-labels from the student | consistency x0.29 +[05:54:53][ 3004s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.6190 +[05:55:14][ 3026s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.5244 +[05:55:36][ 3047s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.6222 +[05:55:57][ 3068s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.4741 +[05:56:18][ 3090s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.5008 +[05:56:23][ 3094s][AL] round 6/6 | labeled=174 | selection_seconds=11.9000 round_seconds=487.1000 pseudo_student_dice=0.5858 pseudo_ensemble_dice=0.3394 consistency_scale=0.2860 val_dice=0.8889 val_vi=0.1497 val_ece=0.0118 +[05:56:33][ 3104s][CONS] final consolidation: pseudo-labels refreshed from the student +[05:56:54][ 3125s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.4606 +[05:57:15][ 3146s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.4388 +[05:57:36][ 3168s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.4371 +[05:57:58][ 3189s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.4126 +[05:58:19][ 3210s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.3581 +[05:58:40][ 3231s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.3504 +[05:59:01][ 3253s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.3270 +[05:59:23][ 3274s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.3318 +[05:59:44][ 3295s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.3389 +[06:00:05][ 3317s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.3184 +[06:00:10][ 3321s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=222.6000 val_dice=0.8926 val_vi=0.1435 val_ece=0.0113 +[06:00:10][ 3321s][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 +[06:00:46][ 3357s][TEST] raw student (no EMA, no TTA): patch_dice=0.8728 patch_vi=0.1771 +[06:00:50][ 3362s][TEST] final model without TTA: patch_dice=0.8728 patch_vi=0.1771 +[06:00:50][ 3362s][RESOURCES] {"phases": {"setup": {"seconds": 1.48, "peak_cpu_rss_mb": 2173.9, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 65.54, "peak_cpu_rss_mb": 4182.5, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 678.3, "peak_cpu_rss_mb": 5201.6, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4598.0, "kd_epochs": 5}, "active_learning": {"seconds": 2570.54, "peak_cpu_rss_mb": 5201.6, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4606.0, "rounds": 6}, "inference": {"seconds": 40.49, "peak_cpu_rss_mb": 5201.6, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4606.0, "patches": 768, "images": 6, "patches_per_second": 19.0, "seconds_per_image": 6.748}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5201.6, "total_seconds": 3356.3} +[06:00:50][ 3362s][TEST] patch_dice=0.8733 patch_iou=0.8121 patch_precision=0.8897 patch_recall=0.8868 patch_vi=0.1731 image_dice=0.8078 image_iou=0.6844 image_precision=0.7892 image_recall=0.8329 image_vi=0.1954 cal_ece=0.0053 cal_nll=0.0703 cal_brier=0.0211 cal_mean_confidence=0.9763 cal_accuracy=0.9711 cal_predicted_prevalence=0.0766 cal_true_prevalence=0.0718 cal_mean_probability=0.0757 +[06:00:51][ 3362s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-entropy b5 s1 | test patch Dice=0.8733 VI=0.1731 | image Dice=0.8078 VI=0.1954 | 3362s diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..a11aa98c539335b335786bb4b43fd3e9a7c92d8a --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json @@ -0,0 +1,785 @@ +{ + "target": "octopus-vulgaris-vertical-lobe-sfltract", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 1, + "config": { + "task": "membrane", + "target": "octopus-vulgaris-vertical-lobe-sfltract", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "octopus-vulgaris-vertical-lobe-sfltract", + "--variants", + "macs-v11-norestart-acq-entropy", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.09888716021456875, + 0.02852231397957182, + 0.006877159468562143, + 0.147623932500446, + 0.03162958721157525, + 0.636412589592954, + 0.019031438719961654, + 0.031015818312360402 + ], + "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": [ + 0.09888716021456875, + 0.02852231397957182, + 0.006877159468562143, + 0.147623932500446, + 0.03162958721157525, + 0.636412589592954, + 0.019031438719961654, + 0.031015818312360402 + ], + "per_source": [ + 0.3196471036584027, + 0.016524109438155077, + 0.0007715730533314212, + 0.17539928949217498, + 0.027481458110301667, + 0.4363205819091585, + 0.016643531248023022, + 0.007212353090452595 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.00048686699692031786, + "mmd2_mixture": 0.012755679331756475, + "mmd2_uniform": 0.09698314484068699, + "mmd2_single_source": [ + 0.07244562949164945, + 0.2568973175367495, + 0.39290896557558747, + 0.05287206442937192, + 0.2649845830320105, + 0.011934176639413208, + 0.33925886828153656, + 0.26844309711822195 + ], + "median_squared_distance": 0.1834397705913593, + "gammas": [ + 1.3628451409095685, + 2.725690281819137, + 5.451380563638274, + 10.902761127276548, + 21.805522254553097 + ], + "max_weight": 0.636412589592954, + "entropy": 1.2268812334836157, + "effective_sources": 3.4105761415276636, + "min_weight": 0.006877159468562143 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.049620160284728154, + 0.42663804362685775, + 0.8166050888075048, + 0.12600028070223912, + 0.36189803378480956, + 0.01001954120360149, + 0.42572157074448214, + 0.5321462950254623 + ], + "standardized": [ + -1.1548909896059252, + 0.32630755724364063, + 1.8583797643241364, + -0.8548147110811022, + 0.07196204012648774, + -1.3104708228948398, + 0.3227069898393355, + 0.740820172048267 + ], + "tau": 0.5, + "max_weight": 0.4363205819091585, + "entropy": 1.3076009419864412, + "effective_sources": 3.6972930433455575, + "min_weight": 0.0007715730533314212 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "seconds": 6.6 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.47571341080584045, + 0.6329400594773938, + 0.1419064245794933, + 0.4669233743357945, + 0.7079863797747427, + 0.8450584862058909, + 0.22902171788078127, + 0.11473171105409108 + ], + "ensemble_val_dice": { + "mixture": 0.8488016622672679, + "mixture@raw": 0.8468775415451569, + "mixture@logit": 0.8498181877900208, + "per_source": 0.8430718482924652, + "per_source@raw": 0.6192917168334541, + "per_source@logit": 0.8479512124278182, + "uniform": 0.7183086201637726, + "uniform@raw": 0.6771618878909939, + "uniform@logit": 0.8065091143369424 + }, + "val_patches": 768 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 3.547258138656616, + 4.074584007263184 + ], + "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "weight_stats": { + "max_weight": 0.636412589592954, + "entropy": 1.2268812334836157, + "effective_sources": 3.4105761415276636, + "min_weight": 0.006877159468562143 + }, + "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": 678.3, + "round0_val": { + "val_dice": 0.8502116365713305, + "val_vi": 0.1469760206899403, + "val_ece": 0.06606864095266592 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 29, + "selection_seconds": 94.3, + "round_seconds": 439.9, + "teacher_dice_labeled": [ + 0.29568020341690904, + 0.6193671974528233, + 0.19588368058977745, + 0.33035580668548686, + 0.5508702113809155, + 0.7308332381787659, + 0.15051414530012766, + 0.23402934928630323 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 237130, + "evidence_nonempty_patches": 23, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.749598658200551, + "restart_teacher_dice": 0.7308332381787659, + "weights": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "ensemble_choice": "uniform", + "candidate_dice": { + "mixture": 0.7308488562528247, + "uniform": 0.7388948429036649, + "top1": 0.7308332381787659, + "top2": 0.7311335062716425, + "top3": 0.7306909902224555 + }, + "val_dice": 0.39731069928710405, + "val_vi": 0.18067060453166736, + "val_ece": 0.07838129414619284 + }, + { + "round": 2, + "n_labeled": 58, + "selection_seconds": 12.5, + "round_seconds": 85.6, + "teacher_dice_labeled": [ + 0.3476127955077974, + 0.4623686045382735, + 0.1755232866300641, + 0.31664754608108453, + 0.48104076607485885, + 0.5880462888045375, + 0.1790241268093611, + 0.32257295870994507 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 646819, + "evidence_nonempty_patches": 51, + "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": 0.5883446508492198, + "uniform": 0.6113743713600025, + "top1": 0.5880462888045375, + "top2": 0.5871022090819282, + "top3": 0.587255051506554 + }, + "pseudo_student_dice": 0.6107994294456149, + "pseudo_ensemble_dice": 0.5225904590692428, + "pseudo_source": "student", + "consistency_scale": 0.369, + "val_dice": 0.6369395682370388, + "val_vi": 0.16326645853342284, + "val_ece": 0.009399857270229506 + }, + { + "round": 3, + "n_labeled": 87, + "selection_seconds": 12.3, + "round_seconds": 409.1, + "teacher_dice_labeled": [ + 0.3880088072732191, + 0.565341164182688, + 0.20471024332465787, + 0.35380388386182804, + 0.5253267179901164, + 0.6915026222402474, + 0.2565538563065365, + 0.3179727238418305 + ], + "teacher_dice_pool": [ + 0.4412735330757333, + 0.6659325718988698, + 0.2372073842200266, + 0.39943171418340423, + 0.5811530177876038, + 0.8057331885912369, + 0.33433901329823956, + 0.3096649357020063 + ], + "evidence_fg_pixels": 1020182, + "evidence_nonempty_patches": 80, + "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": 0.6919442120407975, + "uniform": 0.7091946157908361, + "top1": 0.6915026222402474, + "top2": 0.6921985279311229, + "top3": 0.6912061210138836 + }, + "pseudo_student_dice": 0.6845415260994039, + "pseudo_ensemble_dice": 0.5650033092900414, + "pseudo_source": "student", + "consistency_scale": 0.615, + "val_dice": 0.8476713037914542, + "val_vi": 0.1607710772373184, + "val_ece": 0.007436648340219965 + }, + { + "round": 4, + "n_labeled": 116, + "selection_seconds": 12.2, + "round_seconds": 434.8, + "teacher_dice_labeled": [ + 0.3669914562867691, + 0.5693699554122069, + 0.2055175935989612, + 0.347743730912508, + 0.5187721250757013, + 0.6984750534228014, + 0.2589420340565617, + 0.3004582926329548 + ], + "teacher_dice_pool": [ + 0.44670708753866184, + 0.667642860577924, + 0.23361163069008736, + 0.39822489431136593, + 0.5667288786894037, + 0.7912791062085153, + 0.3405437241505429, + 0.32008342376085336 + ], + "evidence_fg_pixels": 1253998, + "evidence_nonempty_patches": 108, + "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": 0.6988568524984016, + "uniform": 0.7081510561372487, + "top1": 0.6984750534228014, + "top2": 0.6991990657433986, + "top3": 0.6983183514463073 + }, + "pseudo_student_dice": 0.717398226943492, + "pseudo_ensemble_dice": 0.6485005341745833, + "pseudo_source": "student", + "consistency_scale": 0.725, + "val_dice": 0.8906511127825961, + "val_vi": 0.15183013964667516, + "val_ece": 0.010564493907424609 + }, + { + "round": 5, + "n_labeled": 145, + "selection_seconds": 12.0, + "round_seconds": 460.5, + "teacher_dice_labeled": [ + 0.3668099150383525, + 0.5657863433890937, + 0.20371422005742845, + 0.3401057868287816, + 0.508806826757005, + 0.6980405494858268, + 0.254115069915782, + 0.2978415150971249 + ], + "teacher_dice_pool": [ + 0.4282138051931642, + 0.668509289478364, + 0.23248299174518863, + 0.38337992924291425, + 0.5661442227675124, + 0.8152971118400434, + 0.32553445749354537, + 0.3033784254312041 + ], + "evidence_fg_pixels": 1654235, + "evidence_nonempty_patches": 137, + "first_correction": false, + "restart": "", + "weights": [ + 0.005480660578475023, + 0.011561692285882852, + 7.460818807273828e-05, + 0.006264338487527958, + 0.007252225134167767, + 0.9681622203186412, + 0.0003417725105441578, + 0.0008624824966881743 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.6984906391263901, + "uniform": 0.7020901946833366, + "top1": 0.6980405494858268, + "top2": 0.6987377441865222, + "top3": 0.6979569026901176 + }, + "pseudo_student_dice": 0.7831485684638488, + "pseudo_ensemble_dice": 0.4959069721765298, + "pseudo_source": "student", + "consistency_scale": 0.944, + "val_dice": 0.8796156527618478, + "val_vi": 0.15158004082917956, + "val_ece": 0.012681466394150536 + }, + { + "round": 6, + "n_labeled": 174, + "selection_seconds": 11.9, + "round_seconds": 487.1, + "teacher_dice_labeled": [ + 0.356145961605246, + 0.551214355755117, + 0.2025672126275297, + 0.3327589734690903, + 0.5010580750877105, + 0.6802621501411376, + 0.25820683333981315, + 0.28948928888955194 + ], + "teacher_dice_pool": [ + 0.431692128592248, + 0.6726023749138484, + 0.24020601597533023, + 0.3911778628654421, + 0.5933084714965258, + 0.8380355437195518, + 0.31126908803516595, + 0.2935730368159814 + ], + "evidence_fg_pixels": 2203858, + "evidence_nonempty_patches": 166, + "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": 0.6807795554829543, + "uniform": 0.68870702244862, + "top1": 0.6802621501411376, + "top2": 0.6810272289642794, + "top3": 0.6801802090041148 + }, + "pseudo_student_dice": 0.5857522693305887, + "pseudo_ensemble_dice": 0.33939174177213893, + "pseudo_source": "student", + "consistency_scale": 0.286, + "val_dice": 0.8888998968722844, + "val_vi": 0.14966503109444784, + "val_ece": 0.011809566908036034 + }, + { + "round": 7, + "n_labeled": 174, + "selection_seconds": 0.0, + "round_seconds": 222.6, + "final_consolidation": true, + "val_dice": 0.8925997424783859, + "val_vi": 0.1434862563966719, + "val_ece": 0.01128403642853145 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.44560086267226684, + "auroc_top20_error": 0.593673156908451 + }, + "al_seconds": 2570.5, + "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": 0.8727808053306628, + "patch_iou": 0.8096397365891068, + "patch_precision": 0.884013050033685, + "patch_recall": 0.8888230542795013, + "patch_vi": 0.1771062671777847, + "n_patches": 768, + "cal_ece": 0.006959328183459772, + "cal_nll": 0.07563690589200811, + "cal_brier": 0.02189482180963423, + "cal_mean_confidence": 0.9771665425389284, + "cal_accuracy": 0.9702072143554688, + "cal_predicted_prevalence": 0.07643890380859375, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.07524883666029399 + }, + "test_notta": { + "patch_dice": 0.8727808053306628, + "patch_iou": 0.8096397365891068, + "patch_precision": 0.884013050033685, + "patch_recall": 0.8888230542795013, + "patch_vi": 0.1771062671777847, + "n_patches": 768, + "cal_ece": 0.006959328183459772, + "cal_nll": 0.07563690589200811, + "cal_brier": 0.02189482180963423, + "cal_mean_confidence": 0.9771665425389284, + "cal_accuracy": 0.9702072143554688, + "cal_predicted_prevalence": 0.07643890380859375, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.07524883666029399 + }, + "test": { + "patch_dice": 0.8733466141713727, + "patch_iou": 0.8120811888493558, + "patch_precision": 0.889693337909526, + "patch_recall": 0.8867975267013571, + "patch_vi": 0.17312031856958507, + "n_patches": 768, + "image_dice": 0.8077534707124677, + "image_iou": 0.684351592405526, + "image_precision": 0.7892406397670163, + "image_recall": 0.8329104380944133, + "image_vi": 0.19538874245769558, + "n_images": 6, + "cal_ece": 0.005300650660982887, + "cal_nll": 0.0702677819857817, + "cal_brier": 0.02107877091438942, + "cal_mean_confidence": 0.9763216560626121, + "cal_accuracy": 0.9711151123046875, + "cal_predicted_prevalence": 0.07657623291015625, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.0756733806638027 + }, + "eval_seconds": 40.5, + "total_seconds": 3361.6, + "resources": { + "phases": { + "setup": { + "seconds": 1.48, + "peak_cpu_rss_mb": 2173.9, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 65.54, + "peak_cpu_rss_mb": 4182.5, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 678.3, + "peak_cpu_rss_mb": 5201.6, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4598.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2570.54, + "peak_cpu_rss_mb": 5201.6, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4606.0, + "rounds": 6 + }, + "inference": { + "seconds": 40.49, + "peak_cpu_rss_mb": 5201.6, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4606.0, + "patches": 768, + "images": 6, + "patches_per_second": 19.0, + "seconds_per_image": 6.748 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5201.6, + "total_seconds": 3356.3 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..46d5d827deb198e1063d1c7cc9c0dcd30ba12577 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-entropy/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 +1,29,94.3,439.9,"[0.29568020341690904, 0.6193671974528233, 0.19588368058977745, 0.33035580668548686, 0.5508702113809155, 0.7308332381787659, 0.15051414530012766, 0.23402934928630323]",,237130,23,True,kept,0.749598658200551,0.7308332381787659,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.7308488562528247, 'uniform': 0.7388948429036649, 'top1': 0.7308332381787659, 'top2': 0.7311335062716425, 'top3': 0.7306909902224555}",0.39731069928710405,0.18067060453166736,0.07838129414619284,,,,, +2,58,12.5,85.6,"[0.3476127955077974, 0.4623686045382735, 0.1755232866300641, 0.31664754608108453, 0.48104076607485885, 0.5880462888045375, 0.1790241268093611, 0.32257295870994507]",,646819,51,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.5883446508492198, 'uniform': 0.6113743713600025, 'top1': 0.5880462888045375, 'top2': 0.5871022090819282, 'top3': 0.587255051506554}",0.6369395682370388,0.16326645853342284,0.009399857270229506,0.6107994294456149,0.5225904590692428,student,0.369, +3,87,12.3,409.1,"[0.3880088072732191, 0.565341164182688, 0.20471024332465787, 0.35380388386182804, 0.5253267179901164, 0.6915026222402474, 0.2565538563065365, 0.3179727238418305]","[0.4412735330757333, 0.6659325718988698, 0.2372073842200266, 0.39943171418340423, 0.5811530177876038, 0.8057331885912369, 0.33433901329823956, 0.3096649357020063]",1020182,80,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6919442120407975, 'uniform': 0.7091946157908361, 'top1': 0.6915026222402474, 'top2': 0.6921985279311229, 'top3': 0.6912061210138836}",0.8476713037914542,0.1607710772373184,0.007436648340219965,0.6845415260994039,0.5650033092900414,student,0.615, +4,116,12.2,434.8,"[0.3669914562867691, 0.5693699554122069, 0.2055175935989612, 0.347743730912508, 0.5187721250757013, 0.6984750534228014, 0.2589420340565617, 0.3004582926329548]","[0.44670708753866184, 0.667642860577924, 0.23361163069008736, 0.39822489431136593, 0.5667288786894037, 0.7912791062085153, 0.3405437241505429, 0.32008342376085336]",1253998,108,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6988568524984016, 'uniform': 0.7081510561372487, 'top1': 0.6984750534228014, 'top2': 0.6991990657433986, 'top3': 0.6983183514463073}",0.8906511127825961,0.15183013964667516,0.010564493907424609,0.717398226943492,0.6485005341745833,student,0.725, +5,145,12.0,460.5,"[0.3668099150383525, 0.5657863433890937, 0.20371422005742845, 0.3401057868287816, 0.508806826757005, 0.6980405494858268, 0.254115069915782, 0.2978415150971249]","[0.4282138051931642, 0.668509289478364, 0.23248299174518863, 0.38337992924291425, 0.5661442227675124, 0.8152971118400434, 0.32553445749354537, 0.3033784254312041]",1654235,137,False,,,,"[0.005480660578475023, 0.011561692285882852, 7.460818807273828e-05, 0.006264338487527958, 0.007252225134167767, 0.9681622203186412, 0.0003417725105441578, 0.0008624824966881743]",mixture,"{'mixture': 0.6984906391263901, 'uniform': 0.7020901946833366, 'top1': 0.6980405494858268, 'top2': 0.6987377441865222, 'top3': 0.6979569026901176}",0.8796156527618478,0.15158004082917956,0.012681466394150536,0.7831485684638488,0.4959069721765298,student,0.944, +6,174,11.9,487.1,"[0.356145961605246, 0.551214355755117, 0.2025672126275297, 0.3327589734690903, 0.5010580750877105, 0.6802621501411376, 0.25820683333981315, 0.28948928888955194]","[0.431692128592248, 0.6726023749138484, 0.24020601597533023, 0.3911778628654421, 0.5933084714965258, 0.8380355437195518, 0.31126908803516595, 0.2935730368159814]",2203858,166,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6807795554829543, 'uniform': 0.68870702244862, 'top1': 0.6802621501411376, 'top2': 0.6810272289642794, 'top3': 0.6801802090041148}",0.8888998968722844,0.14966503109444784,0.011809566908036034,0.5857522693305887,0.33939174177213893,student,0.286, +7,174,0.0,222.6,,,,,,,,,,,,0.8925997424783859,0.1434862563966719,0.01128403642853145,,,,,True diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..1e6d402fc1b77011866295484ff1704388427127 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt @@ -0,0 +1,108 @@ +[00:56:03][ 0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-laplace budget=5.0% seed=1 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[00:56:03][ 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. +[00:56:04][ 1s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches +[00:56:22][ 19s][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=4.07 +[00:56:56][ 53s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 34s +[00:57:02][ 59s][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 +[00:57:02][ 59s][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 +[00:57:57][ 114s][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 +[00:57:57][ 114s][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 +[00:57:57][ 114s][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 +[00:57:57][ 114s][INIT] student initialised from teacher oct-glia (weight 0.636) +[00:58:01][ 117s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[00:58:01][ 117s][KD] distilling the mixture ensemble (prob space) into the student +[01:00:26][ 263s][KD] soft labels for 3584 patches from 8 teachers in 145s +[01:02:13][ 369s][KD] epoch 1/5 | patches=3584 | loss=0.5717 | 106s +[01:03:59][ 476s][KD] epoch 2/5 | patches=3584 | loss=0.4387 | 106s +[01:05:46][ 582s][KD] epoch 3/5 | patches=3584 | loss=0.4024 | 106s +[01:07:32][ 689s][KD] epoch 4/5 | patches=3584 | loss=0.3575 | 107s +[01:09:19][ 796s][KD] epoch 5/5 | patches=3584 | loss=0.3323 | 107s +[01:09:23][ 800s][ROUND0] before any target label: val_dice=0.4068 val_vi=0.1562 +[01:09:23][ 800s][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 +[01:09:49][ 825s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': -0.42973054159052243, 'auroc_top20_error': 0.01460704062277854} +[01:11:11][ 907s][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 +[01:11:20][ 917s][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 +[01:11:21][ 917s][WEIGHTS] round 1: student kept (labeled Dice 0.724 vs best teacher oct-glia 0.731); distilling the corrected ensemble into it +[01:11:21][ 917s][WEIGHTS] round 1: teacher Dice on 29 labeled patches + 29 blank (237130 fg px): c-elegans=0.296, cerebellum=0.619, 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 +[01:12:55][ 1012s][KD] soft labels for 2309 patches from 8 teachers in 94s +[01:14:04][ 1080s][KD] epoch 1/2 | patches=2309 | loss=0.7251 | 69s +[01:15:13][ 1149s][KD] epoch 2/2 | patches=2309 | loss=0.6221 | 69s +[01:16:34][ 1231s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[01:16:39][ 1235s][FT] round 1/6 epoch 1/5 | patches=58 | loss=1.3947 +[01:16:43][ 1240s][FT] round 1/6 epoch 2/5 | patches=58 | loss=1.3601 +[01:16:47][ 1244s][FT] round 1/6 epoch 3/5 | patches=58 | loss=1.3657 +[01:16:52][ 1248s][FT] round 1/6 epoch 4/5 | patches=58 | loss=1.2920 +[01:16:56][ 1253s][FT] round 1/6 epoch 5/5 | patches=58 | loss=1.3077 +[01:17:01][ 1257s][AL] round 1/6 | labeled=29 | selection_seconds=107.5000 round_seconds=452.9000 restart_student_dice=0.7237 restart_teacher_dice=0.7308 val_dice=0.8265 val_vi=0.1609 val_ece=0.0489 +[01:17:26][ 1282s][AL] round 2: 571/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[01:17:44][ 1300s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.601, uniform=0.616, top1=0.601, top2=0.600, top3=0.600 -> using uniform +[01:17:44][ 1300s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (672023 fg px): c-elegans=0.317, cerebellum=0.466, flywire=0.176, berghia=0.291, microns=0.517, oct-glia=0.601, snemi=0.186, mouse-brain=0.309 | 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 +[01:18:00][ 1316s][CONS] round 2: on the 29 new patches student Dice=0.614 vs ensemble 0.545 -> pseudo-labels from the student | consistency x0.38 +[01:18:07][ 1324s][FT] round 2/6 epoch 1/5 | patches=116 | loss=1.0646 +[01:18:15][ 1332s][FT] round 2/6 epoch 2/5 | patches=116 | loss=0.9956 +[01:18:23][ 1339s][FT] round 2/6 epoch 3/5 | patches=116 | loss=0.9742 +[01:18:30][ 1347s][FT] round 2/6 epoch 4/5 | patches=116 | loss=0.9444 +[01:18:38][ 1354s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.9440 +[01:18:42][ 1359s][AL] round 2/6 | labeled=58 | selection_seconds=25.1000 round_seconds=97.3000 pseudo_student_dice=0.6139 pseudo_ensemble_dice=0.5450 consistency_scale=0.3800 val_dice=0.8588 val_vi=0.1514 val_ece=0.0125 +[01:19:07][ 1384s][AL] round 3: 602/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[01:21:55][ 1551s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.371, cerebellum=0.616, flywire=0.239, berghia=0.365, microns=0.592, oct-glia=0.873, snemi=0.275, mouse-brain=0.247 | blended with labeled Dice at 0.50 +[01:24:33][ 1709s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.650, uniform=0.638, top1=0.649, top2=0.650, top3=0.649 -> using mixture +[01:24:33][ 1709s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (1597523 fg px): c-elegans=0.352, cerebellum=0.508, flywire=0.197, berghia=0.306, microns=0.487, oct-glia=0.649, snemi=0.272, mouse-brain=0.308 | weights: c-elegans=0.008, cerebellum=0.011, flywire=0.000, berghia=0.007, microns=0.009, oct-glia=0.963, snemi=0.001, mouse-brain=0.002 +[01:24:49][ 1725s][CONS] round 3: on the 29 new patches student Dice=0.383 vs ensemble 0.260 -> pseudo-labels from the student | consistency x0.00 +[01:24:54][ 1731s][FT] round 3/6 epoch 1/5 | patches=174 | loss=1.1783 +[01:25:00][ 1736s][FT] round 3/6 epoch 2/5 | patches=174 | loss=1.0471 +[01:25:05][ 1742s][FT] round 3/6 epoch 3/5 | patches=174 | loss=1.0678 +[01:25:11][ 1748s][FT] round 3/6 epoch 4/5 | patches=174 | loss=1.0251 +[01:25:17][ 1753s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.9424 +[01:25:21][ 1758s][AL] round 3/6 | labeled=87 | selection_seconds=24.9000 round_seconds=394.2000 pseudo_student_dice=0.3832 pseudo_ensemble_dice=0.2603 consistency_scale=0.0000 val_dice=0.8525 val_vi=0.1750 val_ece=0.0105 +[01:25:45][ 1782s][AL] round 4: 537/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[01:28:38][ 1954s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.443, cerebellum=0.666, flywire=0.236, berghia=0.392, microns=0.582, oct-glia=0.788, snemi=0.338, mouse-brain=0.314 | blended with labeled Dice at 0.50 +[01:31:20][ 2117s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.631, uniform=0.664, top1=0.630, top2=0.631, top3=0.630 -> using uniform +[01:31:20][ 2117s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1870770 fg px): c-elegans=0.372, cerebellum=0.550, flywire=0.199, berghia=0.328, microns=0.494, oct-glia=0.630, snemi=0.295, mouse-brain=0.318 | 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 +[01:31:36][ 2133s][CONS] round 4: on the 29 new patches student Dice=0.716 vs ensemble 0.677 -> pseudo-labels from the student | consistency x0.72 +[01:31:51][ 2147s][FT] round 4/6 epoch 1/5 | patches=232 | loss=1.0101 +[01:32:05][ 2162s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.9749 +[01:32:19][ 2176s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.9456 +[01:32:34][ 2190s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.9118 +[01:32:48][ 2204s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.9304 +[01:32:52][ 2209s][AL] round 4/6 | labeled=116 | selection_seconds=24.6000 round_seconds=447.0000 pseudo_student_dice=0.7162 pseudo_ensemble_dice=0.6767 consistency_scale=0.7210 val_dice=0.8798 val_vi=0.1542 val_ece=0.0075 +[01:33:16][ 2233s][AL] round 5: 569/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[01:36:13][ 2409s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.415, cerebellum=0.661, flywire=0.244, berghia=0.384, microns=0.603, oct-glia=0.851, snemi=0.310, mouse-brain=0.282 | blended with labeled Dice at 0.50 +[01:38:59][ 2576s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.671, uniform=0.681, top1=0.670, top2=0.671, top3=0.670 -> using uniform +[01:38:59][ 2576s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (2186789 fg px): c-elegans=0.366, cerebellum=0.549, flywire=0.204, berghia=0.329, microns=0.509, oct-glia=0.670, snemi=0.271, 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 +[01:39:15][ 2592s][CONS] round 5: on the 29 new patches student Dice=0.726 vs ensemble 0.597 -> pseudo-labels from the student | consistency x0.75 +[01:39:33][ 2609s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.9539 +[01:39:51][ 2627s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.9196 +[01:40:09][ 2645s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.9494 +[01:40:27][ 2663s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.8943 +[01:40:45][ 2681s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.8669 +[01:40:49][ 2686s][AL] round 5/6 | labeled=145 | selection_seconds=24.3000 round_seconds=472.3000 pseudo_student_dice=0.7263 pseudo_ensemble_dice=0.5966 consistency_scale=0.7540 val_dice=0.8752 val_vi=0.1569 val_ece=0.0086 +[01:41:13][ 2709s][AL] round 6: 553/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[01:44:14][ 2891s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.434, cerebellum=0.668, flywire=0.240, berghia=0.395, microns=0.591, oct-glia=0.824, snemi=0.320, mouse-brain=0.299 | blended with labeled Dice at 0.50 +[01:47:06][ 3063s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.665, uniform=0.683, top1=0.664, top2=0.665, top3=0.664 -> using uniform +[01:47:06][ 3063s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2478728 fg px): c-elegans=0.368, cerebellum=0.555, flywire=0.202, berghia=0.336, microns=0.505, oct-glia=0.664, snemi=0.273, 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 +[01:47:22][ 3078s][CONS] round 6: on the 29 new patches student Dice=0.793 vs ensemble 0.596 -> pseudo-labels from the student | consistency x0.98 +[01:47:43][ 3100s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.8826 +[01:48:04][ 3121s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.9147 +[01:48:26][ 3142s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.8845 +[01:48:47][ 3164s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.8723 +[01:49:08][ 3185s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.8485 +[01:49:13][ 3189s][AL] round 6/6 | labeled=174 | selection_seconds=24.0000 round_seconds=499.4000 pseudo_student_dice=0.7929 pseudo_ensemble_dice=0.5957 consistency_scale=0.9760 val_dice=0.8927 val_vi=0.1440 val_ece=0.0078 +[01:49:23][ 3199s][CONS] final consolidation: pseudo-labels refreshed from the student +[01:49:44][ 3220s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.8445 +[01:50:05][ 3242s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.8160 +[01:50:26][ 3263s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.7961 +[01:50:48][ 3284s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.7927 +[01:51:09][ 3305s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.7625 +[01:51:30][ 3327s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.7530 +[01:51:51][ 3348s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.7819 +[01:52:13][ 3369s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.7434 +[01:52:34][ 3391s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.7514 +[01:52:55][ 3412s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.7475 +[01:53:00][ 3416s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=222.6000 val_dice=0.8887 val_vi=0.1467 val_ece=0.0078 +[01:53:00][ 3416s][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 +[01:53:36][ 3453s][TEST] raw student (no EMA, no TTA): patch_dice=0.8645 patch_vi=0.1770 +[01:53:40][ 3457s][TEST] final model without TTA: patch_dice=0.8645 patch_vi=0.1770 +[01:53:40][ 3457s][RESOURCES] {"phases": {"setup": {"seconds": 16.85, "peak_cpu_rss_mb": 2160.3, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 98.94, "peak_cpu_rss_mb": 4183.7, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 678.09, "peak_cpu_rss_mb": 5202.8, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4598.0, "kd_epochs": 5}, "active_learning": {"seconds": 2616.44, "peak_cpu_rss_mb": 5202.8, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4606.0, "rounds": 6}, "inference": {"seconds": 40.52, "peak_cpu_rss_mb": 5202.8, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4606.0, "patches": 768, "images": 6, "patches_per_second": 19.0, "seconds_per_image": 6.753}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5202.8, "total_seconds": 3450.8} +[01:53:40][ 3457s][TEST] patch_dice=0.8750 patch_iou=0.8143 patch_precision=0.8867 patch_recall=0.8961 patch_vi=0.1732 image_dice=0.8095 image_iou=0.6867 image_precision=0.7810 image_recall=0.8463 image_vi=0.1962 cal_ece=0.0100 cal_nll=0.0766 cal_brier=0.0215 cal_mean_confidence=0.9717 cal_accuracy=0.9708 cal_predicted_prevalence=0.0786 cal_true_prevalence=0.0718 cal_mean_probability=0.0855 +[01:53:41][ 3457s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-laplace b5 s1 | test patch Dice=0.8750 VI=0.1732 | image Dice=0.8095 VI=0.1962 | 3457s diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..32710151377ae6a65293f57015de2cd91e7c5f40 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json @@ -0,0 +1,785 @@ +{ + "target": "octopus-vulgaris-vertical-lobe-sfltract", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "config": { + "task": "membrane", + "target": "octopus-vulgaris-vertical-lobe-sfltract", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "octopus-vulgaris-vertical-lobe-sfltract", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.09888716021456875, + 0.02852231397957182, + 0.006877159468562143, + 0.147623932500446, + 0.03162958721157525, + 0.636412589592954, + 0.019031438719961654, + 0.031015818312360402 + ], + "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": [ + 0.09888716021456875, + 0.02852231397957182, + 0.006877159468562143, + 0.147623932500446, + 0.03162958721157525, + 0.636412589592954, + 0.019031438719961654, + 0.031015818312360402 + ], + "per_source": [ + 0.3196471036584027, + 0.016524109438155077, + 0.0007715730533314212, + 0.17539928949217498, + 0.027481458110301667, + 0.4363205819091585, + 0.016643531248023022, + 0.007212353090452595 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.00048686699692031786, + "mmd2_mixture": 0.012755679331756475, + "mmd2_uniform": 0.09698314484068699, + "mmd2_single_source": [ + 0.07244562949164945, + 0.2568973175367495, + 0.39290896557558747, + 0.05287206442937192, + 0.2649845830320105, + 0.011934176639413208, + 0.33925886828153656, + 0.26844309711822195 + ], + "median_squared_distance": 0.1834397705913593, + "gammas": [ + 1.3628451409095685, + 2.725690281819137, + 5.451380563638274, + 10.902761127276548, + 21.805522254553097 + ], + "max_weight": 0.636412589592954, + "entropy": 1.2268812334836157, + "effective_sources": 3.4105761415276636, + "min_weight": 0.006877159468562143 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.049620160284728154, + 0.42663804362685775, + 0.8166050888075048, + 0.12600028070223912, + 0.36189803378480956, + 0.01001954120360149, + 0.42572157074448214, + 0.5321462950254623 + ], + "standardized": [ + -1.1548909896059252, + 0.32630755724364063, + 1.8583797643241364, + -0.8548147110811022, + 0.07196204012648774, + -1.3104708228948398, + 0.3227069898393355, + 0.740820172048267 + ], + "tau": 0.5, + "max_weight": 0.4363205819091585, + "entropy": 1.3076009419864412, + "effective_sources": 3.6972930433455575, + "min_weight": 0.0007715730533314212 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "seconds": 40.3 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.47571341080584045, + 0.6329400594773938, + 0.1419064245794933, + 0.4669233743357945, + 0.7079863797747427, + 0.8450584862058909, + 0.22902171788078127, + 0.11473171105409108 + ], + "ensemble_val_dice": { + "mixture": 0.8488016622672679, + "mixture@raw": 0.8468775415451569, + "mixture@logit": 0.8498181877900208, + "per_source": 0.8430718482924652, + "per_source@raw": 0.6192917168334541, + "per_source@logit": 0.8479512124278182, + "uniform": 0.7183086201637726, + "uniform@raw": 0.6771618878909939, + "uniform@logit": 0.8065091143369424 + }, + "val_patches": 768 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 3.547258138656616, + 4.074584007263184 + ], + "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "weight_stats": { + "max_weight": 0.636412589592954, + "entropy": 1.2268812334836157, + "effective_sources": 3.4105761415276636, + "min_weight": 0.006877159468562143 + }, + "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": 678.1, + "round0_val": { + "val_dice": 0.4067974008452868, + "val_vi": 0.15621365356541697, + "val_ece": 0.06549482541855627 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 29, + "selection_seconds": 107.5, + "round_seconds": 452.9, + "teacher_dice_labeled": [ + 0.29568020341690904, + 0.6193671974528233, + 0.19588368058977745, + 0.33035580668548686, + 0.5508702113809155, + 0.7308332381787659, + 0.15051414530012766, + 0.23402934928630323 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 237130, + "evidence_nonempty_patches": 23, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.7237450842918105, + "restart_teacher_dice": 0.7308332381787659, + "weights": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "ensemble_choice": "uniform", + "candidate_dice": { + "mixture": 0.7308488562528247, + "uniform": 0.7388948429036649, + "top1": 0.7308332381787659, + "top2": 0.7311335062716425, + "top3": 0.7306909902224555 + }, + "val_dice": 0.8265384832710779, + "val_vi": 0.16094128778513914, + "val_ece": 0.04887358321577289 + }, + { + "round": 2, + "n_labeled": 58, + "selection_seconds": 25.1, + "round_seconds": 97.3, + "teacher_dice_labeled": [ + 0.3171883581634619, + 0.46570797027165556, + 0.17555021224032277, + 0.2911379528704814, + 0.5171020431871733, + 0.6011370938587045, + 0.1856036164323916, + 0.30947977967970813 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 672023, + "evidence_nonempty_patches": 51, + "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": 0.6008390269849482, + "uniform": 0.6162660612905311, + "top1": 0.6011370938587045, + "top2": 0.6000244930568988, + "top3": 0.6001927615324153 + }, + "pseudo_student_dice": 0.613930989815908, + "pseudo_ensemble_dice": 0.5449893692757639, + "pseudo_source": "student", + "consistency_scale": 0.38, + "val_dice": 0.858755045097654, + "val_vi": 0.1513502459658489, + "val_ece": 0.012488309354615111 + }, + { + "round": 3, + "n_labeled": 87, + "selection_seconds": 24.9, + "round_seconds": 394.2, + "teacher_dice_labeled": [ + 0.3515194750513133, + 0.5081483731457119, + 0.1967517587863077, + 0.30607281205994263, + 0.4873790788992553, + 0.6494763538876254, + 0.27193648915742474, + 0.30833822986758297 + ], + "teacher_dice_pool": [ + 0.3713920260153615, + 0.6156583110479326, + 0.23892875624458382, + 0.3647158945021577, + 0.591553287347347, + 0.8728288107309928, + 0.2746848725997957, + 0.24697999986747887 + ], + "evidence_fg_pixels": 1597523, + "evidence_nonempty_patches": 80, + "first_correction": false, + "restart": "", + "weights": [ + 0.007602838779626683, + 0.010501498964006668, + 0.00011248570107448317, + 0.0072047729084134095, + 0.009461503682782231, + 0.962908280806494, + 0.000660211341494219, + 0.0015484078161083674 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.6498703208262191, + "uniform": 0.6380122833004724, + "top1": 0.6494763538876254, + "top2": 0.6498371076706292, + "top3": 0.6492208400303191 + }, + "pseudo_student_dice": 0.38323891424672846, + "pseudo_ensemble_dice": 0.2603137417028411, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.852483095049041, + "val_vi": 0.17500004822089713, + "val_ece": 0.010542839366299465 + }, + { + "round": 4, + "n_labeled": 116, + "selection_seconds": 24.6, + "round_seconds": 447.0, + "teacher_dice_labeled": [ + 0.37165118582176443, + 0.5498739512200651, + 0.19894122455555435, + 0.32796989413566147, + 0.49362131291149014, + 0.6299155752418457, + 0.29462911835367733, + 0.31849510251907787 + ], + "teacher_dice_pool": [ + 0.4432668421464532, + 0.6655030182527801, + 0.23613268382521105, + 0.3915591066403852, + 0.5820384495809867, + 0.7883949429794211, + 0.3380038795229039, + 0.3135432956883544 + ], + "evidence_fg_pixels": 1870770, + "evidence_nonempty_patches": 109, + "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": 0.6310020695428453, + "uniform": 0.6641508814931827, + "top1": 0.6299155752418457, + "top2": 0.6310825091972322, + "top3": 0.6297899101236573 + }, + "pseudo_student_dice": 0.7162450297722375, + "pseudo_ensemble_dice": 0.6766694366299267, + "pseudo_source": "student", + "consistency_scale": 0.721, + "val_dice": 0.8797632053364045, + "val_vi": 0.15423802631351424, + "val_ece": 0.007480832876547508 + }, + { + "round": 5, + "n_labeled": 145, + "selection_seconds": 24.3, + "round_seconds": 472.3, + "teacher_dice_labeled": [ + 0.3660794262070859, + 0.5490828422499876, + 0.20365563891426552, + 0.32856908333908263, + 0.508535981970067, + 0.670062722012975, + 0.2709322094932049, + 0.3038809822055101 + ], + "teacher_dice_pool": [ + 0.41543039333650833, + 0.6614358260242903, + 0.24373609609710786, + 0.384228244423019, + 0.603001507114574, + 0.8509662250083222, + 0.309811235614449, + 0.2822458495441181 + ], + "evidence_fg_pixels": 2186789, + "evidence_nonempty_patches": 138, + "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": 0.6707006667666381, + "uniform": 0.6809966041879847, + "top1": 0.670062722012975, + "top2": 0.6708633202970876, + "top3": 0.6700126434291589 + }, + "pseudo_student_dice": 0.7263052830088944, + "pseudo_ensemble_dice": 0.5965506068759152, + "pseudo_source": "student", + "consistency_scale": 0.754, + "val_dice": 0.875244862502354, + "val_vi": 0.15694088885158228, + "val_ece": 0.008585482330209854 + }, + { + "round": 6, + "n_labeled": 174, + "selection_seconds": 24.0, + "round_seconds": 499.4, + "teacher_dice_labeled": [ + 0.3681232181861549, + 0.5551674270712343, + 0.2021990584599616, + 0.33618850765162767, + 0.5048303375240749, + 0.6641611574975141, + 0.2725573674876931, + 0.30352943057597503 + ], + "teacher_dice_pool": [ + 0.43428106838072467, + 0.6680803435489528, + 0.23994257936146782, + 0.39529555441652675, + 0.5914807803652934, + 0.8240333760711983, + 0.3198122503561808, + 0.29933052019728335 + ], + "evidence_fg_pixels": 2478728, + "evidence_nonempty_patches": 165, + "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": 0.664855616294795, + "uniform": 0.6832583651582114, + "top1": 0.6641611574975141, + "top2": 0.6650440389558555, + "top3": 0.6640915000195716 + }, + "pseudo_student_dice": 0.7929019284967859, + "pseudo_ensemble_dice": 0.5956894105964962, + "pseudo_source": "student", + "consistency_scale": 0.976, + "val_dice": 0.8926858816531681, + "val_vi": 0.1439952713975571, + "val_ece": 0.007831732653372411 + }, + { + "round": 7, + "n_labeled": 174, + "selection_seconds": 0.0, + "round_seconds": 222.6, + "final_consolidation": true, + "val_dice": 0.8886851466384921, + "val_vi": 0.1466770562917574, + "val_ece": 0.007776746073254894 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.42973054159052243, + "auroc_top20_error": 0.01460704062277854 + }, + "al_seconds": 2616.4, + "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": 0.8644854064227285, + "patch_iou": 0.8015775339461436, + "patch_precision": 0.8736117259970388, + "patch_recall": 0.8942403794858601, + "patch_vi": 0.1769803664520788, + "n_patches": 768, + "cal_ece": 0.01005696782408574, + "cal_nll": 0.08044871377833289, + "cal_brier": 0.022497044457770853, + "cal_mean_confidence": 0.9738861268454938, + "cal_accuracy": 0.9699223836263021, + "cal_predicted_prevalence": 0.07838185628255208, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.084024411452732 + }, + "test_notta": { + "patch_dice": 0.8644854064227285, + "patch_iou": 0.8015775339461436, + "patch_precision": 0.8736117259970388, + "patch_recall": 0.8942403794858601, + "patch_vi": 0.1769803664520788, + "n_patches": 768, + "cal_ece": 0.01005696782408574, + "cal_nll": 0.08044871377833289, + "cal_brier": 0.022497044457770853, + "cal_mean_confidence": 0.9738861268454938, + "cal_accuracy": 0.9699223836263021, + "cal_predicted_prevalence": 0.07838185628255208, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.084024411452732 + }, + "test": { + "patch_dice": 0.8749669073071713, + "patch_iou": 0.8143105444897812, + "patch_precision": 0.8867455102594728, + "patch_recall": 0.8960972601232834, + "patch_vi": 0.17321693095203786, + "n_patches": 768, + "image_dice": 0.8094618761648809, + "image_iou": 0.6866971832851606, + "image_precision": 0.781044872770262, + "image_recall": 0.8462603726171656, + "image_vi": 0.19624323121594725, + "n_images": 6, + "cal_ece": 0.010036823548981465, + "cal_nll": 0.07660645530730685, + "cal_brier": 0.021545440108361295, + "cal_mean_confidence": 0.9716628545615914, + "cal_accuracy": 0.9708366394042969, + "cal_predicted_prevalence": 0.07863744099934895, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.08545583541733916 + }, + "eval_seconds": 40.5, + "total_seconds": 3456.9, + "resources": { + "phases": { + "setup": { + "seconds": 16.85, + "peak_cpu_rss_mb": 2160.3, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 98.94, + "peak_cpu_rss_mb": 4183.7, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 678.09, + "peak_cpu_rss_mb": 5202.8, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4598.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2616.44, + "peak_cpu_rss_mb": 5202.8, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4606.0, + "rounds": 6 + }, + "inference": { + "seconds": 40.52, + "peak_cpu_rss_mb": 5202.8, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4606.0, + "patches": 768, + "images": 6, + "patches_per_second": 19.0, + "seconds_per_image": 6.753 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5202.8, + "total_seconds": 3450.8 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..fe609b5f8159d7037845b7bb37cfbf7d649c1ee3 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-laplace/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 +1,29,107.5,452.9,"[0.29568020341690904, 0.6193671974528233, 0.19588368058977745, 0.33035580668548686, 0.5508702113809155, 0.7308332381787659, 0.15051414530012766, 0.23402934928630323]",,237130,23,True,kept,0.7237450842918105,0.7308332381787659,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.7308488562528247, 'uniform': 0.7388948429036649, 'top1': 0.7308332381787659, 'top2': 0.7311335062716425, 'top3': 0.7306909902224555}",0.8265384832710779,0.16094128778513914,0.04887358321577289,,,,, +2,58,25.1,97.3,"[0.3171883581634619, 0.46570797027165556, 0.17555021224032277, 0.2911379528704814, 0.5171020431871733, 0.6011370938587045, 0.1856036164323916, 0.30947977967970813]",,672023,51,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6008390269849482, 'uniform': 0.6162660612905311, 'top1': 0.6011370938587045, 'top2': 0.6000244930568988, 'top3': 0.6001927615324153}",0.858755045097654,0.1513502459658489,0.012488309354615111,0.613930989815908,0.5449893692757639,student,0.38, +3,87,24.9,394.2,"[0.3515194750513133, 0.5081483731457119, 0.1967517587863077, 0.30607281205994263, 0.4873790788992553, 0.6494763538876254, 0.27193648915742474, 0.30833822986758297]","[0.3713920260153615, 0.6156583110479326, 0.23892875624458382, 0.3647158945021577, 0.591553287347347, 0.8728288107309928, 0.2746848725997957, 0.24697999986747887]",1597523,80,False,,,,"[0.007602838779626683, 0.010501498964006668, 0.00011248570107448317, 0.0072047729084134095, 0.009461503682782231, 0.962908280806494, 0.000660211341494219, 0.0015484078161083674]",mixture,"{'mixture': 0.6498703208262191, 'uniform': 0.6380122833004724, 'top1': 0.6494763538876254, 'top2': 0.6498371076706292, 'top3': 0.6492208400303191}",0.852483095049041,0.17500004822089713,0.010542839366299465,0.38323891424672846,0.2603137417028411,student,0.0, +4,116,24.6,447.0,"[0.37165118582176443, 0.5498739512200651, 0.19894122455555435, 0.32796989413566147, 0.49362131291149014, 0.6299155752418457, 0.29462911835367733, 0.31849510251907787]","[0.4432668421464532, 0.6655030182527801, 0.23613268382521105, 0.3915591066403852, 0.5820384495809867, 0.7883949429794211, 0.3380038795229039, 0.3135432956883544]",1870770,109,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6310020695428453, 'uniform': 0.6641508814931827, 'top1': 0.6299155752418457, 'top2': 0.6310825091972322, 'top3': 0.6297899101236573}",0.8797632053364045,0.15423802631351424,0.007480832876547508,0.7162450297722375,0.6766694366299267,student,0.721, +5,145,24.3,472.3,"[0.3660794262070859, 0.5490828422499876, 0.20365563891426552, 0.32856908333908263, 0.508535981970067, 0.670062722012975, 0.2709322094932049, 0.3038809822055101]","[0.41543039333650833, 0.6614358260242903, 0.24373609609710786, 0.384228244423019, 0.603001507114574, 0.8509662250083222, 0.309811235614449, 0.2822458495441181]",2186789,138,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6707006667666381, 'uniform': 0.6809966041879847, 'top1': 0.670062722012975, 'top2': 0.6708633202970876, 'top3': 0.6700126434291589}",0.875244862502354,0.15694088885158228,0.008585482330209854,0.7263052830088944,0.5965506068759152,student,0.754, +6,174,24.0,499.4,"[0.3681232181861549, 0.5551674270712343, 0.2021990584599616, 0.33618850765162767, 0.5048303375240749, 0.6641611574975141, 0.2725573674876931, 0.30352943057597503]","[0.43428106838072467, 0.6680803435489528, 0.23994257936146782, 0.39529555441652675, 0.5914807803652934, 0.8240333760711983, 0.3198122503561808, 0.29933052019728335]",2478728,165,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.664855616294795, 'uniform': 0.6832583651582114, 'top1': 0.6641611574975141, 'top2': 0.6650440389558555, 'top3': 0.6640915000195716}",0.8926858816531681,0.1439952713975571,0.007831732653372411,0.7929019284967859,0.5956894105964962,student,0.976, +7,174,0.0,222.6,,,,,,,,,,,,0.8886851466384921,0.1466770562917574,0.007776746073254894,,,,,True diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/log.txt b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..cbcb4cbe608390a446091bf549d2c5dcb3186429 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/log.txt @@ -0,0 +1,108 @@ +[10:07:32][ 0s][START] target=octopus-vulgaris-vertical-lobe-sfltract variant=macs-v11-norestart-acq-mcdropout budget=5.0% seed=2 | {"task": "membrane", "target": "octopus-vulgaris-vertical-lobe-sfltract", "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": 8, "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} +[10:07:32][ 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. +[10:07:32][ 0s][DATA] octopus-vulgaris-vertical-lobe-sfltract: train=3584 val=768 test=768 patches +[10:07:34][ 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, snemi=3.55, mouse-brain=4.07 +[10:07:35][ 3s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[10:07:41][ 9s][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 +[10:07:41][ 9s][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 +[10:08:36][ 65s][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 +[10:08:36][ 65s][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 +[10:08:36][ 65s][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 +[10:08:36][ 65s][INIT] student initialised from teacher oct-glia (weight 0.636) +[10:08:40][ 68s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[10:08:40][ 68s][KD] distilling the mixture ensemble (prob space) into the student +[10:11:05][ 214s][KD] soft labels for 3584 patches from 8 teachers in 146s +[10:12:52][ 320s][KD] epoch 1/5 | patches=3584 | loss=0.5966 | 107s +[10:14:39][ 427s][KD] epoch 2/5 | patches=3584 | loss=0.4565 | 107s +[10:16:25][ 534s][KD] epoch 3/5 | patches=3584 | loss=0.4027 | 107s +[10:18:12][ 640s][KD] epoch 4/5 | patches=3584 | loss=0.3656 | 107s +[10:19:58][ 747s][KD] epoch 5/5 | patches=3584 | loss=0.3396 | 106s +[10:20:03][ 751s][ROUND0] before any target label: val_dice=0.8449 val_vi=0.1459 +[10:20:03][ 751s][AL] patches=3584 (pool 2338 after dropping blank; free blank negatives: True) | budget=179 (5.0%) | rounds=6 | per_round=29 | acquisition=mcdropout | first_round=stratified +[10:20:17][ 765s][AL] round-1 uncertainty quality on 2338 patches: {'spearman_uncertainty_error': 0.4329896382452679, 'auroc_top20_error': 0.6001805384158325} +[10:21:39][ 847s][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 +[10:21:48][ 857s][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 +[10:21:49][ 857s][WEIGHTS] round 1: student kept (labeled Dice 0.743 vs best teacher oct-glia 0.765); distilling the corrected ensemble into it +[10:21:49][ 857s][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 +[10:23:23][ 952s][KD] soft labels for 2309 patches from 8 teachers in 94s +[10:24:32][ 1021s][KD] epoch 1/2 | patches=2309 | loss=0.1381 | 69s +[10:25:41][ 1089s][KD] epoch 2/2 | patches=2309 | loss=0.1326 | 69s +[10:27:03][ 1171s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[10:27:08][ 1176s][FT] round 1/6 epoch 1/5 | patches=58 | loss=0.8672 +[10:27:12][ 1180s][FT] round 1/6 epoch 2/5 | patches=58 | loss=0.5693 +[10:27:17][ 1185s][FT] round 1/6 epoch 3/5 | patches=58 | loss=0.5764 +[10:27:21][ 1189s][FT] round 1/6 epoch 4/5 | patches=58 | loss=0.4051 +[10:27:25][ 1194s][FT] round 1/6 epoch 5/5 | patches=58 | loss=0.4886 +[10:27:30][ 1198s][AL] round 1/6 | labeled=29 | selection_seconds=96.1000 round_seconds=442.6000 restart_student_dice=0.7430 restart_teacher_dice=0.7648 val_dice=0.4271 val_vi=0.2099 val_ece=0.0166 +[10:27:44][ 1212s][AL] round 2: 288/2309 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[10:28:02][ 1230s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.362, uniform=0.381, top1=0.360, top2=0.360, top3=0.359 -> using uniform +[10:28:02][ 1230s][WEIGHTS] round 2: teacher Dice on 58 labeled patches + 58 blank (769883 fg px): c-elegans=0.226, cerebellum=0.317, flywire=0.125, berghia=0.184, microns=0.282, oct-glia=0.360, snemi=0.215, mouse-brain=0.253 | 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 +[10:28:19][ 1247s][CONS] round 2: on the 29 new patches student Dice=0.404 vs ensemble 0.189 -> pseudo-labels from the student | consistency x0.00 +[10:28:22][ 1251s][FT] round 2/6 epoch 1/5 | patches=116 | loss=1.2214 +[10:28:26][ 1255s][FT] round 2/6 epoch 2/5 | patches=116 | loss=0.9194 +[10:28:30][ 1259s][FT] round 2/6 epoch 3/5 | patches=116 | loss=1.1992 +[10:28:34][ 1262s][FT] round 2/6 epoch 4/5 | patches=116 | loss=1.0262 +[10:28:38][ 1266s][FT] round 2/6 epoch 5/5 | patches=116 | loss=0.7869 +[10:28:43][ 1271s][AL] round 2/6 | labeled=58 | selection_seconds=14.2000 round_seconds=68.2000 pseudo_student_dice=0.4044 pseudo_ensemble_dice=0.1885 consistency_scale=0.0000 val_dice=0.8393 val_vi=0.1872 val_ece=0.0451 +[10:28:56][ 1285s][AL] round 3: 558/2280 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[10:31:44][ 1453s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.403, cerebellum=0.620, flywire=0.226, berghia=0.366, microns=0.543, oct-glia=0.767, snemi=0.319, mouse-brain=0.296 | blended with labeled Dice at 0.50 +[10:34:23][ 1611s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.603, uniform=0.611, top1=0.603, top2=0.603, top3=0.603 -> using uniform +[10:34:23][ 1611s][WEIGHTS] round 3: teacher Dice on 87 labeled patches + 87 blank (1110404 fg px): c-elegans=0.335, cerebellum=0.500, flywire=0.183, berghia=0.301, microns=0.442, oct-glia=0.603, snemi=0.263, mouse-brain=0.282 | 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 +[10:34:39][ 1627s][CONS] round 3: on the 29 new patches student Dice=0.758 vs ensemble 0.613 -> pseudo-labels from the student | consistency x0.86 +[10:34:50][ 1638s][FT] round 3/6 epoch 1/5 | patches=174 | loss=0.9883 +[10:35:01][ 1649s][FT] round 3/6 epoch 2/5 | patches=174 | loss=0.7900 +[10:35:12][ 1660s][FT] round 3/6 epoch 3/5 | patches=174 | loss=0.6876 +[10:35:23][ 1672s][FT] round 3/6 epoch 4/5 | patches=174 | loss=0.6380 +[10:35:34][ 1683s][FT] round 3/6 epoch 5/5 | patches=174 | loss=0.7518 +[10:35:39][ 1687s][AL] round 3/6 | labeled=87 | selection_seconds=14.0000 round_seconds=411.9000 pseudo_student_dice=0.7578 pseudo_ensemble_dice=0.6134 consistency_scale=0.8590 val_dice=0.8606 val_vi=0.1726 val_ece=0.0064 +[10:35:53][ 1701s][AL] round 4: 563/2251 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[10:38:45][ 1873s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.409, cerebellum=0.642, flywire=0.233, berghia=0.378, microns=0.575, oct-glia=0.810, snemi=0.320, mouse-brain=0.287 | blended with labeled Dice at 0.50 +[10:41:27][ 2036s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.638, uniform=0.635, top1=0.637, top2=0.638, top3=0.637 -> using mixture +[10:41:27][ 2036s][WEIGHTS] round 4: teacher Dice on 116 labeled patches + 116 blank (1435427 fg px): c-elegans=0.338, cerebellum=0.505, flywire=0.188, berghia=0.309, microns=0.461, oct-glia=0.637, snemi=0.261, mouse-brain=0.273 | weights: c-elegans=0.008, cerebellum=0.012, flywire=0.000, berghia=0.008, microns=0.008, oct-glia=0.962, snemi=0.001, mouse-brain=0.001 +[10:41:44][ 2052s][CONS] round 4: on the 29 new patches student Dice=0.736 vs ensemble 0.545 -> pseudo-labels from the student | consistency x0.79 +[10:41:58][ 2066s][FT] round 4/6 epoch 1/5 | patches=232 | loss=0.6580 +[10:42:13][ 2081s][FT] round 4/6 epoch 2/5 | patches=232 | loss=0.6007 +[10:42:27][ 2095s][FT] round 4/6 epoch 3/5 | patches=232 | loss=0.4883 +[10:42:42][ 2110s][FT] round 4/6 epoch 4/5 | patches=232 | loss=0.4529 +[10:42:56][ 2124s][FT] round 4/6 epoch 5/5 | patches=232 | loss=0.4728 +[10:43:01][ 2129s][AL] round 4/6 | labeled=116 | selection_seconds=13.8000 round_seconds=437.2000 pseudo_student_dice=0.7360 pseudo_ensemble_dice=0.5446 consistency_scale=0.7870 val_dice=0.8463 val_vi=0.1834 val_ece=0.0116 +[10:43:14][ 2142s][AL] round 5: 554/2222 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[10:46:10][ 2319s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.429, cerebellum=0.644, flywire=0.232, berghia=0.392, microns=0.571, oct-glia=0.789, snemi=0.335, mouse-brain=0.307 | blended with labeled Dice at 0.50 +[10:48:57][ 2485s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.643, uniform=0.643, top1=0.643, top2=0.643, top3=0.642 -> using mixture +[10:48:57][ 2485s][WEIGHTS] round 5: teacher Dice on 145 labeled patches + 145 blank (1729181 fg px): c-elegans=0.348, cerebellum=0.509, flywire=0.190, berghia=0.317, microns=0.469, oct-glia=0.643, snemi=0.263, mouse-brain=0.283 | weights: c-elegans=0.008, cerebellum=0.011, flywire=0.000, berghia=0.009, microns=0.008, oct-glia=0.962, snemi=0.001, mouse-brain=0.001 +[10:49:13][ 2501s][CONS] round 5: on the 29 new patches student Dice=0.674 vs ensemble 0.635 -> pseudo-labels from the student | consistency x0.58 +[10:49:31][ 2519s][FT] round 5/6 epoch 1/5 | patches=290 | loss=0.5204 +[10:49:49][ 2537s][FT] round 5/6 epoch 2/5 | patches=290 | loss=0.4699 +[10:50:07][ 2555s][FT] round 5/6 epoch 3/5 | patches=290 | loss=0.4563 +[10:50:25][ 2573s][FT] round 5/6 epoch 4/5 | patches=290 | loss=0.3912 +[10:50:43][ 2591s][FT] round 5/6 epoch 5/5 | patches=290 | loss=0.3980 +[10:50:48][ 2596s][AL] round 5/6 | labeled=145 | selection_seconds=13.5000 round_seconds=462.7000 pseudo_student_dice=0.6738 pseudo_ensemble_dice=0.6353 consistency_scale=0.5790 val_dice=0.8727 val_vi=0.1596 val_ece=0.0109 +[10:51:01][ 2609s][AL] round 6: 570/2193 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[10:54:03][ 2791s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.402, cerebellum=0.642, flywire=0.237, berghia=0.377, microns=0.593, oct-glia=0.844, snemi=0.306, mouse-brain=0.275 | blended with labeled Dice at 0.50 +[10:56:54][ 2962s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.681, uniform=0.666, top1=0.680, top2=0.681, top3=0.680 -> using mixture +[10:56:54][ 2962s][WEIGHTS] round 6: teacher Dice on 174 labeled patches + 174 blank (2114988 fg px): c-elegans=0.339, cerebellum=0.523, flywire=0.197, berghia=0.317, microns=0.492, oct-glia=0.680, snemi=0.250, mouse-brain=0.271 | weights: c-elegans=0.005, cerebellum=0.009, flywire=0.000, berghia=0.006, microns=0.007, oct-glia=0.971, snemi=0.000, mouse-brain=0.001 +[10:57:10][ 2978s][CONS] round 6: on the 29 new patches student Dice=0.694 vs ensemble 0.595 -> pseudo-labels from the student | consistency x0.65 +[10:57:32][ 3000s][FT] round 6/6 epoch 1/5 | patches=348 | loss=0.4019 +[10:57:53][ 3021s][FT] round 6/6 epoch 2/5 | patches=348 | loss=0.3831 +[10:58:14][ 3043s][FT] round 6/6 epoch 3/5 | patches=348 | loss=0.3692 +[10:58:36][ 3064s][FT] round 6/6 epoch 4/5 | patches=348 | loss=0.3469 +[10:58:57][ 3085s][FT] round 6/6 epoch 5/5 | patches=348 | loss=0.3359 +[10:59:02][ 3090s][AL] round 6/6 | labeled=174 | selection_seconds=13.4000 round_seconds=489.5000 pseudo_student_dice=0.6941 pseudo_ensemble_dice=0.5954 consistency_scale=0.6470 val_dice=0.8811 val_vi=0.1517 val_ece=0.0088 +[10:59:12][ 3100s][CONS] final consolidation: pseudo-labels refreshed from the student +[10:59:33][ 3121s][FT] final 1/10 epoch 1/1 | patches=348 | loss=0.3833 +[10:59:54][ 3142s][FT] final 2/10 epoch 1/1 | patches=348 | loss=0.3832 +[11:00:16][ 3164s][FT] final 3/10 epoch 1/1 | patches=348 | loss=0.3428 +[11:00:37][ 3185s][FT] final 4/10 epoch 1/1 | patches=348 | loss=0.3186 +[11:00:58][ 3207s][FT] final 5/10 epoch 1/1 | patches=348 | loss=0.3337 +[11:01:20][ 3228s][FT] final 6/10 epoch 1/1 | patches=348 | loss=0.3198 +[11:01:41][ 3249s][FT] final 7/10 epoch 1/1 | patches=348 | loss=0.3078 +[11:02:02][ 3271s][FT] final 8/10 epoch 1/1 | patches=348 | loss=0.3041 +[11:02:24][ 3292s][FT] final 9/10 epoch 1/1 | patches=348 | loss=0.3015 +[11:02:45][ 3313s][FT] final 10/10 epoch 1/1 | patches=348 | loss=0.3106 +[11:02:50][ 3318s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=223.7000 val_dice=0.8877 val_vi=0.1480 val_ece=0.0073 +[11:02:50][ 3318s][WEIGHTS] final: c-elegans=0.005, cerebellum=0.009, flywire=0.000, berghia=0.006, microns=0.007, oct-glia=0.971, snemi=0.000, mouse-brain=0.001 +[11:03:26][ 3354s][TEST] raw student (no EMA, no TTA): patch_dice=0.8632 patch_vi=0.1811 +[11:03:30][ 3359s][TEST] final model without TTA: patch_dice=0.8632 patch_vi=0.1811 +[11:03:30][ 3359s][RESOURCES] {"phases": {"setup": {"seconds": 1.48, "peak_cpu_rss_mb": 2182.8, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 65.58, "peak_cpu_rss_mb": 4183.0, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 678.72, "peak_cpu_rss_mb": 5202.5, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4598.0, "kd_epochs": 5}, "active_learning": {"seconds": 2566.86, "peak_cpu_rss_mb": 5202.5, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4606.0, "rounds": 6}, "inference": {"seconds": 40.64, "peak_cpu_rss_mb": 5202.5, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4606.0, "patches": 768, "images": 6, "patches_per_second": 18.9, "seconds_per_image": 6.773}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5202.5, "total_seconds": 3353.3} +[11:03:30][ 3359s][TEST] patch_dice=0.8643 patch_iou=0.8032 patch_precision=0.8908 patch_recall=0.8768 patch_vi=0.1774 image_dice=0.7988 image_iou=0.6722 image_precision=0.7880 image_recall=0.8158 image_vi=0.2006 cal_ece=0.0069 cal_nll=0.0758 cal_brier=0.0221 cal_mean_confidence=0.9769 cal_accuracy=0.9700 cal_predicted_prevalence=0.0760 cal_true_prevalence=0.0718 cal_mean_probability=0.0768 +[11:03:31][ 3359s][DONE] octopus-vulgaris-vertical-lobe-sfltract macs-v11-norestart-acq-mcdropout b5 s2 | test patch Dice=0.8643 VI=0.1774 | image Dice=0.7988 VI=0.2006 | 3359s diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/result.json b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..a4f340a6b6aaa9d5fee36668f33be5b29f388262 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/result.json @@ -0,0 +1,785 @@ +{ + "target": "octopus-vulgaris-vertical-lobe-sfltract", + "variant": "macs-v11-norestart-acq-mcdropout", + "budget": 5.0, + "seed": 2, + "config": { + "task": "membrane", + "target": "octopus-vulgaris-vertical-lobe-sfltract", + "variant": "macs-v11-norestart-acq-mcdropout", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "octopus-vulgaris-vertical-lobe-sfltract", + "--variants", + "macs-v11-norestart-acq-mcdropout", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.09888716021456875, + 0.02852231397957182, + 0.006877159468562143, + 0.147623932500446, + 0.03162958721157525, + 0.636412589592954, + 0.019031438719961654, + 0.031015818312360402 + ], + "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": [ + 0.09888716021456875, + 0.02852231397957182, + 0.006877159468562143, + 0.147623932500446, + 0.03162958721157525, + 0.636412589592954, + 0.019031438719961654, + 0.031015818312360402 + ], + "per_source": [ + 0.3196471036584027, + 0.016524109438155077, + 0.0007715730533314212, + 0.17539928949217498, + 0.027481458110301667, + 0.4363205819091585, + 0.016643531248023022, + 0.007212353090452595 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.00048686699692031786, + "mmd2_mixture": 0.012755679331756475, + "mmd2_uniform": 0.09698314484068699, + "mmd2_single_source": [ + 0.07244562949164945, + 0.2568973175367495, + 0.39290896557558747, + 0.05287206442937192, + 0.2649845830320105, + 0.011934176639413208, + 0.33925886828153656, + 0.26844309711822195 + ], + "median_squared_distance": 0.1834397705913593, + "gammas": [ + 1.3628451409095685, + 2.725690281819137, + 5.451380563638274, + 10.902761127276548, + 21.805522254553097 + ], + "max_weight": 0.636412589592954, + "entropy": 1.2268812334836157, + "effective_sources": 3.4105761415276636, + "min_weight": 0.006877159468562143 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.049620160284728154, + 0.42663804362685775, + 0.8166050888075048, + 0.12600028070223912, + 0.36189803378480956, + 0.01001954120360149, + 0.42572157074448214, + 0.5321462950254623 + ], + "standardized": [ + -1.1548909896059252, + 0.32630755724364063, + 1.8583797643241364, + -0.8548147110811022, + 0.07196204012648774, + -1.3104708228948398, + 0.3227069898393355, + 0.740820172048267 + ], + "tau": 0.5, + "max_weight": 0.4363205819091585, + "entropy": 1.3076009419864412, + "effective_sources": 3.6972930433455575, + "min_weight": 0.0007715730533314212 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "seconds": 6.6 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.47571341080584045, + 0.6329400594773938, + 0.1419064245794933, + 0.4669233743357945, + 0.7079863797747427, + 0.8450584862058909, + 0.22902171788078127, + 0.11473171105409108 + ], + "ensemble_val_dice": { + "mixture": 0.8488016622672679, + "mixture@raw": 0.8468775415451569, + "mixture@logit": 0.8498181877900208, + "per_source": 0.8430718482924652, + "per_source@raw": 0.6192917168334541, + "per_source@logit": 0.8479512124278182, + "uniform": 0.7183086201637726, + "uniform@raw": 0.6771618878909939, + "uniform@logit": 0.8065091143369424 + }, + "val_patches": 768 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 3.547258138656616, + 4.074584007263184 + ], + "top_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "weight_stats": { + "max_weight": 0.636412589592954, + "entropy": 1.2268812334836157, + "effective_sources": 3.4105761415276636, + "min_weight": 0.006877159468562143 + }, + "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": 678.7, + "round0_val": { + "val_dice": 0.844884651592599, + "val_vi": 0.14593690508010287, + "val_ece": 0.06297539540092424 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 29, + "selection_seconds": 96.1, + "round_seconds": 442.6, + "teacher_dice_labeled": [ + 0.25717887499142283, + 0.5690769623050445, + 0.1686301357941244, + 0.27579610687630757, + 0.5811094797160338, + 0.7648037373241484, + 0.10985651675729007, + 0.20342648065706517 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 179039, + "evidence_nonempty_patches": 21, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.7429788813082652, + "restart_teacher_dice": 0.7648037373241484, + "weights": [ + 0.0009536818496459393, + 0.006223068059955061, + 2.7359642534432497e-05, + 0.0017150389707529289, + 0.0077834094558282, + 0.9830806325167469, + 4.206514211543207e-05, + 0.0001747443624209932 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.7647068027764391, + "uniform": 0.7405350046482169, + "top1": 0.7648037373241484, + "top2": 0.7643927747284575, + "top3": 0.7646270894165438 + }, + "val_dice": 0.4270724097964022, + "val_vi": 0.20986386428805637, + "val_ece": 0.016562829469566765 + }, + { + "round": 2, + "n_labeled": 58, + "selection_seconds": 14.2, + "round_seconds": 68.2, + "teacher_dice_labeled": [ + 0.22587745940791537, + 0.31698070371161113, + 0.12462178899546054, + 0.18431241169575352, + 0.2822639835725904, + 0.35970436518487914, + 0.21512327894146918, + 0.25303882275789313 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 769883, + "evidence_nonempty_patches": 47, + "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": 0.3621730033525568, + "uniform": 0.38051786207576516, + "top1": 0.35970436518487914, + "top2": 0.36039438680228353, + "top3": 0.3593144880664838 + }, + "pseudo_student_dice": 0.404397978293183, + "pseudo_ensemble_dice": 0.1885413347061179, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.839313244845056, + "val_vi": 0.1871619919775931, + "val_ece": 0.04511803920160227 + }, + { + "round": 3, + "n_labeled": 87, + "selection_seconds": 14.0, + "round_seconds": 411.9, + "teacher_dice_labeled": [ + 0.3347328518610615, + 0.5001404973411617, + 0.18257684052307566, + 0.301204699930828, + 0.4424734262397891, + 0.6025802942919263, + 0.2628953407838804, + 0.2819626314895449 + ], + "teacher_dice_pool": [ + 0.4027438950161543, + 0.6200607333924688, + 0.22590870742348548, + 0.366050843102077, + 0.5429376205791411, + 0.7671385769259249, + 0.31892228647137716, + 0.2960467280637949 + ], + "evidence_fg_pixels": 1110404, + "evidence_nonempty_patches": 74, + "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": 0.6034687502246806, + "uniform": 0.6108538918617112, + "top1": 0.6025802942919263, + "top2": 0.6033994926595407, + "top3": 0.6025226457729906 + }, + "pseudo_student_dice": 0.7577983328352536, + "pseudo_ensemble_dice": 0.6134312812346644, + "pseudo_source": "student", + "consistency_scale": 0.859, + "val_dice": 0.8606391248002924, + "val_vi": 0.17258987468099388, + "val_ece": 0.006442501714751189 + }, + { + "round": 4, + "n_labeled": 116, + "selection_seconds": 13.8, + "round_seconds": 437.2, + "teacher_dice_labeled": [ + 0.3381786272232266, + 0.5052704073606991, + 0.1878705984148606, + 0.3089617272275621, + 0.46140194760202413, + 0.6371013035682522, + 0.26058193018573017, + 0.27253356202443413 + ], + "teacher_dice_pool": [ + 0.4089333685603796, + 0.6416396488347919, + 0.23295701412877842, + 0.3780127247463931, + 0.5753647964100921, + 0.8100108079944658, + 0.3195180987768935, + 0.2872763217692721 + ], + "evidence_fg_pixels": 1435427, + "evidence_nonempty_patches": 100, + "first_correction": false, + "restart": "", + "weights": [ + 0.007524908071981864, + 0.01154028224680597, + 0.00011641016732088002, + 0.008387464194102436, + 0.008252905997431267, + 0.9622872783077605, + 0.0006665530822541776, + 0.0012241979323426642 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.637705710703107, + "uniform": 0.6351334535951196, + "top1": 0.6371013035682522, + "top2": 0.6377429880954093, + "top3": 0.636979409689884 + }, + "pseudo_student_dice": 0.7359883623178871, + "pseudo_ensemble_dice": 0.5445880183283505, + "pseudo_source": "student", + "consistency_scale": 0.787, + "val_dice": 0.8462507575573692, + "val_vi": 0.18344553848250167, + "val_ece": 0.011602489143117903 + }, + { + "round": 5, + "n_labeled": 145, + "selection_seconds": 13.5, + "round_seconds": 462.7, + "teacher_dice_labeled": [ + 0.34783195263869726, + 0.5085847810456393, + 0.18989485557335906, + 0.31691661905716506, + 0.46906773359921405, + 0.6426591476342762, + 0.2632133787604556, + 0.2833962106620097 + ], + "teacher_dice_pool": [ + 0.4291246304900841, + 0.6440484423766484, + 0.23188019058719783, + 0.39202517928015146, + 0.5713678412064529, + 0.7885259140627896, + 0.33496846094096766, + 0.306770538899655 + ], + "evidence_fg_pixels": 1729181, + "evidence_nonempty_patches": 128, + "first_correction": false, + "restart": "", + "weights": [ + 0.007835715882556655, + 0.011278833739633114, + 0.00011231448567228763, + 0.008586812058593643, + 0.008424664331596915, + 0.9618243695229529, + 0.0006470183518167428, + 0.0012902716271775594 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.6431960380116553, + "uniform": 0.6427129289897711, + "top1": 0.6426591476342762, + "top2": 0.64328273965317, + "top3": 0.6424447067725132 + }, + "pseudo_student_dice": 0.6737726910002576, + "pseudo_ensemble_dice": 0.6352852470158447, + "pseudo_source": "student", + "consistency_scale": 0.579, + "val_dice": 0.872681189635275, + "val_vi": 0.15964571873942823, + "val_ece": 0.010866263393352124 + }, + { + "round": 6, + "n_labeled": 174, + "selection_seconds": 13.4, + "round_seconds": 489.5, + "teacher_dice_labeled": [ + 0.3394639924416737, + 0.5227582892078185, + 0.197064979552099, + 0.3167971266642039, + 0.4920812887171, + 0.6802133425724319, + 0.25012603080841567, + 0.2710178116401448 + ], + "teacher_dice_pool": [ + 0.4019549577525064, + 0.6424874197113001, + 0.23732978669094343, + 0.37713044373411325, + 0.5927859949030386, + 0.8444809846026134, + 0.3055219464803523, + 0.2746842105078218 + ], + "evidence_fg_pixels": 2114988, + "evidence_nonempty_patches": 157, + "first_correction": false, + "restart": "", + "weights": [ + 0.004999738768501313, + 0.00901631717958687, + 8.371156243438058e-05, + 0.00595008583835789, + 0.007357146913473611, + 0.9714082673176162, + 0.00039381222300683295, + 0.0007909201970230076 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.6805360799138476, + "uniform": 0.6664409579587651, + "top1": 0.6802133425724319, + "top2": 0.6806826202966263, + "top3": 0.6800856327938514 + }, + "pseudo_student_dice": 0.6941139408873892, + "pseudo_ensemble_dice": 0.5953877530550494, + "pseudo_source": "student", + "consistency_scale": 0.647, + "val_dice": 0.8810563017493284, + "val_vi": 0.1516974844445877, + "val_ece": 0.008790176335206875 + }, + { + "round": 7, + "n_labeled": 174, + "selection_seconds": 0.0, + "round_seconds": 223.7, + "final_consolidation": true, + "val_dice": 0.8877374793949646, + "val_vi": 0.1480138674777313, + "val_ece": 0.00734695886139486 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.4329896382452679, + "auroc_top20_error": 0.6001805384158325 + }, + "al_seconds": 2566.9, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.004999738768501313, + 0.00901631717958687, + 8.371156243438058e-05, + 0.00595008583835789, + 0.007357146913473611, + 0.9714082673176162, + 0.00039381222300683295, + 0.0007909201970230076 + ], + "weight_stats_final": { + "max_weight": 0.9714082673176162, + "entropy": 0.17327628466338119, + "effective_sources": 1.1891946159348459, + "min_weight": 8.371156243438058e-05 + }, + "top_source_final": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.8631586445206421, + "patch_iou": 0.799814331641088, + "patch_precision": 0.8821052042572802, + "patch_recall": 0.8785268631377646, + "patch_vi": 0.18109389859865985, + "n_patches": 768, + "cal_ece": 0.008786372724529175, + "cal_nll": 0.08220366447098529, + "cal_brier": 0.022817539634909845, + "cal_mean_confidence": 0.9779814207388521, + "cal_accuracy": 0.9691950480143229, + "cal_predicted_prevalence": 0.07668304443359375, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.0772628882504308 + }, + "test_notta": { + "patch_dice": 0.8631586445206421, + "patch_iou": 0.799814331641088, + "patch_precision": 0.8821052042572802, + "patch_recall": 0.8785268631377646, + "patch_vi": 0.18109389859865985, + "n_patches": 768, + "cal_ece": 0.008786372724529175, + "cal_nll": 0.08220366447098529, + "cal_brier": 0.022817539634909845, + "cal_mean_confidence": 0.9779814207388521, + "cal_accuracy": 0.9691950480143229, + "cal_predicted_prevalence": 0.07668304443359375, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.0772628882504308 + }, + "test": { + "patch_dice": 0.8642838710090276, + "patch_iou": 0.8032092821758569, + "patch_precision": 0.890816581492523, + "patch_recall": 0.8768043999013256, + "patch_vi": 0.17736875597624022, + "n_patches": 768, + "image_dice": 0.7988423190301718, + "image_iou": 0.672154603933092, + "image_precision": 0.7880281390563937, + "image_recall": 0.8158464700761794, + "image_vi": 0.2006298433573774, + "n_images": 6, + "cal_ece": 0.006921095013166664, + "cal_nll": 0.07579974673544221, + "cal_brier": 0.02208113620039101, + "cal_mean_confidence": 0.9768672505378339, + "cal_accuracy": 0.9700101216634115, + "cal_predicted_prevalence": 0.07602818806966145, + "cal_true_prevalence": 0.07175954182942708, + "cal_mean_probability": 0.07675483929619768 + }, + "eval_seconds": 40.6, + "total_seconds": 3358.6, + "resources": { + "phases": { + "setup": { + "seconds": 1.48, + "peak_cpu_rss_mb": 2182.8, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 65.58, + "peak_cpu_rss_mb": 4183.0, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 678.72, + "peak_cpu_rss_mb": 5202.5, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4598.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2566.86, + "peak_cpu_rss_mb": 5202.5, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4606.0, + "rounds": 6 + }, + "inference": { + "seconds": 40.64, + "peak_cpu_rss_mb": 5202.5, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4606.0, + "patches": 768, + "images": 6, + "patches_per_second": 18.9, + "seconds_per_image": 6.773 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5202.5, + "total_seconds": 3353.3 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/macs-v11-norestart-acq-mcdropout/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..f8c09fbbaf828d4a337f84a8d370a872b97bac10 --- /dev/null +++ b/macs_v2_r26/runs/membrane/octopus-vulgaris-vertical-lobe-sfltract/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 +1,29,96.1,442.6,"[0.25717887499142283, 0.5690769623050445, 0.1686301357941244, 0.27579610687630757, 0.5811094797160338, 0.7648037373241484, 0.10985651675729007, 0.20342648065706517]",,179039,21,True,kept,0.7429788813082652,0.7648037373241484,"[0.0009536818496459393, 0.006223068059955061, 2.7359642534432497e-05, 0.0017150389707529289, 0.0077834094558282, 0.9830806325167469, 4.206514211543207e-05, 0.0001747443624209932]",mixture,"{'mixture': 0.7647068027764391, 'uniform': 0.7405350046482169, 'top1': 0.7648037373241484, 'top2': 0.7643927747284575, 'top3': 0.7646270894165438}",0.4270724097964022,0.20986386428805637,0.016562829469566765,,,,, +2,58,14.2,68.2,"[0.22587745940791537, 0.31698070371161113, 0.12462178899546054, 0.18431241169575352, 0.2822639835725904, 0.35970436518487914, 0.21512327894146918, 0.25303882275789313]",,769883,47,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.3621730033525568, 'uniform': 0.38051786207576516, 'top1': 0.35970436518487914, 'top2': 0.36039438680228353, 'top3': 0.3593144880664838}",0.839313244845056,0.1871619919775931,0.04511803920160227,0.404397978293183,0.1885413347061179,student,0.0, +3,87,14.0,411.9,"[0.3347328518610615, 0.5001404973411617, 0.18257684052307566, 0.301204699930828, 0.4424734262397891, 0.6025802942919263, 0.2628953407838804, 0.2819626314895449]","[0.4027438950161543, 0.6200607333924688, 0.22590870742348548, 0.366050843102077, 0.5429376205791411, 0.7671385769259249, 0.31892228647137716, 0.2960467280637949]",1110404,74,False,,,,"[0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125, 0.125]",uniform,"{'mixture': 0.6034687502246806, 'uniform': 0.6108538918617112, 'top1': 0.6025802942919263, 'top2': 0.6033994926595407, 'top3': 0.6025226457729906}",0.8606391248002924,0.17258987468099388,0.006442501714751189,0.7577983328352536,0.6134312812346644,student,0.859, +4,116,13.8,437.2,"[0.3381786272232266, 0.5052704073606991, 0.1878705984148606, 0.3089617272275621, 0.46140194760202413, 0.6371013035682522, 0.26058193018573017, 0.27253356202443413]","[0.4089333685603796, 0.6416396488347919, 0.23295701412877842, 0.3780127247463931, 0.5753647964100921, 0.8100108079944658, 0.3195180987768935, 0.2872763217692721]",1435427,100,False,,,,"[0.007524908071981864, 0.01154028224680597, 0.00011641016732088002, 0.008387464194102436, 0.008252905997431267, 0.9622872783077605, 0.0006665530822541776, 0.0012241979323426642]",mixture,"{'mixture': 0.637705710703107, 'uniform': 0.6351334535951196, 'top1': 0.6371013035682522, 'top2': 0.6377429880954093, 'top3': 0.636979409689884}",0.8462507575573692,0.18344553848250167,0.011602489143117903,0.7359883623178871,0.5445880183283505,student,0.787, +5,145,13.5,462.7,"[0.34783195263869726, 0.5085847810456393, 0.18989485557335906, 0.31691661905716506, 0.46906773359921405, 0.6426591476342762, 0.2632133787604556, 0.2833962106620097]","[0.4291246304900841, 0.6440484423766484, 0.23188019058719783, 0.39202517928015146, 0.5713678412064529, 0.7885259140627896, 0.33496846094096766, 0.306770538899655]",1729181,128,False,,,,"[0.007835715882556655, 0.011278833739633114, 0.00011231448567228763, 0.008586812058593643, 0.008424664331596915, 0.9618243695229529, 0.0006470183518167428, 0.0012902716271775594]",mixture,"{'mixture': 0.6431960380116553, 'uniform': 0.6427129289897711, 'top1': 0.6426591476342762, 'top2': 0.64328273965317, 'top3': 0.6424447067725132}",0.872681189635275,0.15964571873942823,0.010866263393352124,0.6737726910002576,0.6352852470158447,student,0.579, +6,174,13.4,489.5,"[0.3394639924416737, 0.5227582892078185, 0.197064979552099, 0.3167971266642039, 0.4920812887171, 0.6802133425724319, 0.25012603080841567, 0.2710178116401448]","[0.4019549577525064, 0.6424874197113001, 0.23732978669094343, 0.37713044373411325, 0.5927859949030386, 0.8444809846026134, 0.3055219464803523, 0.2746842105078218]",2114988,157,False,,,,"[0.004999738768501313, 0.00901631717958687, 8.371156243438058e-05, 0.00595008583835789, 0.007357146913473611, 0.9714082673176162, 0.00039381222300683295, 0.0007909201970230076]",mixture,"{'mixture': 0.6805360799138476, 'uniform': 0.6664409579587651, 'top1': 0.6802133425724319, 'top2': 0.6806826202966263, 'top3': 0.6800856327938514}",0.8810563017493284,0.1516974844445877,0.008790176335206875,0.6941139408873892,0.5953877530550494,student,0.647, +7,174,0.0,223.7,,,,,,,,,,,,0.8877374793949646,0.1480138674777313,0.00734695886139486,,,,,True diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..565a95f06d1bb8225c268c5c8fb0467371306851 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt @@ -0,0 +1,104 @@ +[09:14:20][ 0s][START] target=snemi3d variant=macs-v11-norestart-acq-laplace budget=5.0% seed=2 | {"task": "membrane", "target": "snemi3d", "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": 8, "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} +[09:14:20][ 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. +[09:14:20][ 0s][DATA] snemi3d: train=3430 val=735 test=735 patches +[09:14:22][ 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=4.07 +[09:14:22][ 3s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[09:14:29][ 9s][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 +[09:14:29][ 9s][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 +[09:15:22][ 62s][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 +[09:15:22][ 62s][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 +[09:15:22][ 62s][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 +[09:15:22][ 62s][INIT] student initialised from teacher microns (weight 0.477) +[09:15:25][ 66s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[09:15:25][ 66s][KD] distilling the mixture ensemble (prob space) into the student +[09:17:45][ 206s][KD] soft labels for 3430 patches from 8 teachers in 140s +[09:19:27][ 308s][KD] epoch 1/5 | patches=3430 | loss=0.4414 | 102s +[09:21:09][ 410s][KD] epoch 2/5 | patches=3430 | loss=0.3718 | 102s +[09:22:51][ 512s][KD] epoch 3/5 | patches=3430 | loss=0.3565 | 102s +[09:24:33][ 614s][KD] epoch 4/5 | patches=3430 | loss=0.3515 | 102s +[09:26:15][ 715s][KD] epoch 5/5 | patches=3430 | loss=0.3483 | 102s +[09:26:19][ 720s][ROUND0] before any target label: val_dice=0.3835 val_vi=0.6426 +[09:26:19][ 720s][AL] patches=3430 (pool 3430 after dropping blank; free blank negatives: True) | budget=171 (5.0%) | rounds=6 | per_round=28 | acquisition=laplace | first_round=stratified +[09:26:57][ 757s][AL] round-1 uncertainty quality on 3430 patches: {'spearman_uncertainty_error': 0.5072424045387947, 'auroc_top20_error': 0.8403184472456204} +[09:28:19][ 839s][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 +[09:28:24][ 844s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.521, uniform=0.380, top1=0.522, top2=0.522, top3=0.522 -> using mixture +[09:28:24][ 845s][WEIGHTS] round 1: student kept (labeled Dice 0.412 vs best teacher microns 0.522); distilling the corrected ensemble into it +[09:28:24][ 845s][WEIGHTS] round 1: teacher Dice on 28 labeled patches + 0 blank (232196 fg px): c-elegans=0.309, cerebellum=0.401, flywire=0.213, berghia=0.349, microns=0.522, oct-glia=0.163, oct-sfl=0.203, mouse-brain=0.230 | weights: c-elegans=0.001, cerebellum=0.033, flywire=0.035, berghia=0.009, microns=0.918, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.002 +[09:30:43][ 984s][KD] soft labels for 3402 patches from 8 teachers in 139s +[09:32:24][ 1085s][KD] epoch 1/2 | patches=3402 | loss=0.2505 | 101s +[09:34:05][ 1186s][KD] epoch 2/2 | patches=3402 | loss=0.2471 | 101s +[09:35:27][ 1268s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[09:35:30][ 1270s][FT] round 1/6 epoch 1/5 | patches=28 | loss=1.0468 +[09:35:32][ 1272s][FT] round 1/6 epoch 2/5 | patches=28 | loss=0.9720 +[09:35:35][ 1275s][FT] round 1/6 epoch 3/5 | patches=28 | loss=0.9385 +[09:35:37][ 1277s][FT] round 1/6 epoch 4/5 | patches=28 | loss=0.9014 +[09:35:40][ 1280s][FT] round 1/6 epoch 5/5 | patches=28 | loss=0.8928 +[09:35:44][ 1284s][AL] round 1/6 | labeled=28 | selection_seconds=119.3000 round_seconds=560.4000 restart_student_dice=0.4118 restart_teacher_dice=0.5217 val_dice=0.7175 val_vi=0.5862 val_ece=0.0104 +[09:36:21][ 1321s][AL] round 2: 3/3402 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[09:36:30][ 1331s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.482, uniform=0.311, top1=0.486, top2=0.487, top3=0.487 -> using top3 +[09:36:30][ 1331s][WEIGHTS] round 2: teacher Dice on 56 labeled patches + 0 blank (551979 fg px): c-elegans=0.330, cerebellum=0.390, flywire=0.196, berghia=0.293, microns=0.486, oct-glia=0.129, oct-sfl=0.220, mouse-brain=0.271 | weights: c-elegans=0.002, cerebellum=0.044, flywire=0.000, berghia=0.000, microns=0.954, oct-glia=0.000, oct-sfl=0.000, mouse-brain=0.000 +[09:36:44][ 1344s][CONS] round 2: on the 28 new patches student Dice=0.609 vs ensemble 0.454 -> pseudo-labels from the student | consistency x0.36 +[09:36:48][ 1348s][FT] round 2/6 epoch 1/5 | patches=56 | loss=0.8503 +[09:36:52][ 1352s][FT] round 2/6 epoch 2/5 | patches=56 | loss=0.8056 +[09:36:56][ 1356s][FT] round 2/6 epoch 3/5 | patches=56 | loss=0.7764 +[09:37:00][ 1360s][FT] round 2/6 epoch 4/5 | patches=56 | loss=0.7538 +[09:37:04][ 1364s][FT] round 2/6 epoch 5/5 | patches=56 | loss=0.7283 +[09:37:08][ 1369s][AL] round 2/6 | labeled=56 | selection_seconds=37.0000 round_seconds=80.0000 pseudo_student_dice=0.6088 pseudo_ensemble_dice=0.4538 consistency_scale=0.3630 val_dice=0.7391 val_vi=0.5686 val_ece=0.0112 +[09:40:27][ 1567s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.390, cerebellum=0.476, flywire=0.239, berghia=0.389, microns=0.668, oct-glia=0.198, oct-sfl=0.276, mouse-brain=0.309 | blended with labeled Dice at 0.50 +[09:42:59][ 1719s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.575, uniform=0.366, top1=0.576, top2=0.578, top3=0.578 -> using mixture +[09:42:59][ 1719s][WEIGHTS] round 3: teacher Dice on 84 labeled patches + 0 blank (913408 fg px): c-elegans=0.371, cerebellum=0.445, flywire=0.207, berghia=0.335, microns=0.576, oct-glia=0.155, oct-sfl=0.237, mouse-brain=0.298 | weights: c-elegans=0.001, cerebellum=0.031, flywire=0.019, berghia=0.005, microns=0.940, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[09:43:13][ 1733s][CONS] round 3: on the 28 new patches student Dice=0.636 vs ensemble 0.480 -> pseudo-labels from the student | consistency x0.45 +[09:43:18][ 1739s][FT] round 3/6 epoch 1/5 | patches=84 | loss=0.7969 +[09:43:24][ 1745s][FT] round 3/6 epoch 2/5 | patches=84 | loss=0.7781 +[09:43:30][ 1750s][FT] round 3/6 epoch 3/5 | patches=84 | loss=0.7675 +[09:43:36][ 1756s][FT] round 3/6 epoch 4/5 | patches=84 | loss=0.7291 +[09:43:42][ 1762s][FT] round 3/6 epoch 5/5 | patches=84 | loss=0.7268 +[09:43:46][ 1766s][AL] round 3/6 | labeled=84 | selection_seconds=36.7000 round_seconds=393.4000 pseudo_student_dice=0.6362 pseudo_ensemble_dice=0.4801 consistency_scale=0.4540 val_dice=0.7496 val_vi=0.5489 val_ece=0.0118 +[09:47:06][ 1966s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.400, cerebellum=0.482, flywire=0.235, berghia=0.379, microns=0.649, oct-glia=0.197, oct-sfl=0.269, mouse-brain=0.319 | blended with labeled Dice at 0.50 +[09:49:40][ 2120s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.565, uniform=0.360, top1=0.566, top2=0.567, top3=0.567 -> using mixture +[09:49:40][ 2120s][WEIGHTS] round 4: teacher Dice on 112 labeled patches + 0 blank (1179147 fg px): c-elegans=0.373, cerebellum=0.443, flywire=0.204, berghia=0.328, microns=0.566, oct-glia=0.155, oct-sfl=0.243, mouse-brain=0.300 | weights: c-elegans=0.002, cerebellum=0.033, flywire=0.021, berghia=0.005, microns=0.936, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[09:49:54][ 2134s][CONS] round 4: on the 28 new patches student Dice=0.665 vs ensemble 0.476 -> pseudo-labels from the student | consistency x0.55 +[09:50:01][ 2142s][FT] round 4/6 epoch 1/5 | patches=112 | loss=0.7404 +[09:50:08][ 2149s][FT] round 4/6 epoch 2/5 | patches=112 | loss=0.7197 +[09:50:16][ 2156s][FT] round 4/6 epoch 3/5 | patches=112 | loss=0.7162 +[09:50:23][ 2163s][FT] round 4/6 epoch 4/5 | patches=112 | loss=0.6985 +[09:50:30][ 2171s][FT] round 4/6 epoch 5/5 | patches=112 | loss=0.6919 +[09:50:35][ 2175s][AL] round 4/6 | labeled=112 | selection_seconds=36.4000 round_seconds=404.4000 pseudo_student_dice=0.6648 pseudo_ensemble_dice=0.4758 consistency_scale=0.5490 val_dice=0.7550 val_vi=0.5303 val_ece=0.0146 +[09:53:57][ 2377s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.379, cerebellum=0.471, flywire=0.242, berghia=0.378, microns=0.642, oct-glia=0.196, oct-sfl=0.279, mouse-brain=0.302 | blended with labeled Dice at 0.50 +[09:56:33][ 2533s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.558, uniform=0.358, top1=0.559, top2=0.560, top3=0.560 -> using mixture +[09:56:33][ 2533s][WEIGHTS] round 5: teacher Dice on 140 labeled patches + 0 blank (1474014 fg px): c-elegans=0.363, cerebellum=0.436, flywire=0.203, berghia=0.330, microns=0.559, oct-glia=0.156, oct-sfl=0.240, mouse-brain=0.291 | weights: c-elegans=0.001, cerebellum=0.033, flywire=0.022, berghia=0.005, microns=0.934, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[09:56:47][ 2547s][CONS] round 5: on the 28 new patches student Dice=0.644 vs ensemble 0.449 -> pseudo-labels from the student | consistency x0.48 +[09:56:56][ 2556s][FT] round 5/6 epoch 1/5 | patches=140 | loss=0.7162 +[09:57:05][ 2565s][FT] round 5/6 epoch 2/5 | patches=140 | loss=0.6915 +[09:57:14][ 2574s][FT] round 5/6 epoch 3/5 | patches=140 | loss=0.7144 +[09:57:23][ 2583s][FT] round 5/6 epoch 4/5 | patches=140 | loss=0.7015 +[09:57:32][ 2593s][FT] round 5/6 epoch 5/5 | patches=140 | loss=0.6746 +[09:57:36][ 2597s][AL] round 5/6 | labeled=140 | selection_seconds=36.2000 round_seconds=417.5000 pseudo_student_dice=0.6435 pseudo_ensemble_dice=0.4493 consistency_scale=0.4780 val_dice=0.7550 val_vi=0.5302 val_ece=0.0132 +[10:01:00][ 2801s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.384, cerebellum=0.468, flywire=0.235, berghia=0.383, microns=0.635, oct-glia=0.192, oct-sfl=0.273, mouse-brain=0.307 | blended with labeled Dice at 0.50 +[10:03:38][ 2959s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.557, uniform=0.359, top1=0.558, top2=0.560, top3=0.560 -> using mixture +[10:03:38][ 2959s][WEIGHTS] round 6: teacher Dice on 168 labeled patches + 0 blank (1744619 fg px): c-elegans=0.364, cerebellum=0.435, flywire=0.201, berghia=0.333, microns=0.558, oct-glia=0.157, oct-sfl=0.239, mouse-brain=0.293 | weights: c-elegans=0.001, cerebellum=0.033, flywire=0.022, berghia=0.006, microns=0.934, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[10:03:52][ 2972s][CONS] round 6: on the 28 new patches student Dice=0.682 vs ensemble 0.509 -> pseudo-labels from the student | consistency x0.61 +[10:04:02][ 2983s][FT] round 6/6 epoch 1/5 | patches=168 | loss=0.6951 +[10:04:13][ 2993s][FT] round 6/6 epoch 2/5 | patches=168 | loss=0.6870 +[10:04:24][ 3004s][FT] round 6/6 epoch 3/5 | patches=168 | loss=0.6925 +[10:04:34][ 3014s][FT] round 6/6 epoch 4/5 | patches=168 | loss=0.6733 +[10:04:45][ 3025s][FT] round 6/6 epoch 5/5 | patches=168 | loss=0.6769 +[10:04:49][ 3030s][AL] round 6/6 | labeled=168 | selection_seconds=35.9000 round_seconds=428.2000 pseudo_student_dice=0.6815 pseudo_ensemble_dice=0.5090 consistency_scale=0.6050 val_dice=0.7581 val_vi=0.5190 val_ece=0.0157 +[10:04:59][ 3040s][CONS] final consolidation: pseudo-labels refreshed from the student +[10:05:10][ 3050s][FT] final 1/10 epoch 1/1 | patches=168 | loss=0.6609 +[10:05:20][ 3061s][FT] final 2/10 epoch 1/1 | patches=168 | loss=0.6543 +[10:05:31][ 3071s][FT] final 3/10 epoch 1/1 | patches=168 | loss=0.6464 +[10:05:42][ 3082s][FT] final 4/10 epoch 1/1 | patches=168 | loss=0.6473 +[10:05:52][ 3093s][FT] final 5/10 epoch 1/1 | patches=168 | loss=0.6407 +[10:06:03][ 3103s][FT] final 6/10 epoch 1/1 | patches=168 | loss=0.6366 +[10:06:13][ 3114s][FT] final 7/10 epoch 1/1 | patches=168 | loss=0.6301 +[10:06:24][ 3124s][FT] final 8/10 epoch 1/1 | patches=168 | loss=0.6261 +[10:06:34][ 3135s][FT] final 9/10 epoch 1/1 | patches=168 | loss=0.6188 +[10:06:45][ 3145s][FT] final 10/10 epoch 1/1 | patches=168 | loss=0.6215 +[10:06:49][ 3150s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=115.9000 val_dice=0.7586 val_vi=0.5180 val_ece=0.0218 +[10:06:49][ 3150s][WEIGHTS] final: c-elegans=0.001, cerebellum=0.033, flywire=0.022, berghia=0.006, microns=0.934, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[10:07:24][ 3184s][TEST] raw student (no EMA, no TTA): patch_dice=0.7503 patch_vi=0.5264 +[10:07:28][ 3189s][TEST] final model without TTA: patch_dice=0.7503 patch_vi=0.5264 +[10:07:28][ 3189s][RESOURCES] {"phases": {"setup": {"seconds": 1.48, "peak_cpu_rss_mb": 2191.8, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 63.32, "peak_cpu_rss_mb": 4181.8, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 649.83, "peak_cpu_rss_mb": 5156.0, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 2430.01, "peak_cpu_rss_mb": 5156.0, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 38.92, "peak_cpu_rss_mb": 5156.0, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 735, "images": 15, "patches_per_second": 18.9, "seconds_per_image": 2.595}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5156.0, "total_seconds": 3183.6} +[10:07:28][ 3189s][TEST] patch_dice=0.7592 patch_iou=0.6215 patch_precision=0.7384 patch_recall=0.7841 patch_vi=0.5102 image_dice=0.8083 image_iou=0.6787 image_precision=0.7898 image_recall=0.8282 image_vi=0.5167 cal_ece=0.0124 cal_nll=0.1814 cal_brier=0.0572 cal_mean_confidence=0.9298 cal_accuracy=0.9174 cal_predicted_prevalence=0.2178 cal_true_prevalence=0.2059 cal_mean_probability=0.2169 +[10:07:29][ 3189s][DONE] snemi3d macs-v11-norestart-acq-laplace b5 s2 | test patch Dice=0.7592 VI=0.5102 | image Dice=0.8083 VI=0.5167 | 3189s diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..04692c8f969fb94babf9496a7066c0db5d27b7b9 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json @@ -0,0 +1,785 @@ +{ + "target": "snemi3d", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 2, + "config": { + "task": "membrane", + "target": "snemi3d", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "snemi3d", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.005265163985947525, + 0.05775356661687405, + 0.3937117170688725, + 0.02742108315870487, + 0.4770168284742763, + 0.006835511403011292, + 0.011200087973044965, + 0.020796041319268533 + ], + "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": [ + 0.005265163985947525, + 0.05775356661687405, + 0.3937117170688725, + 0.02742108315870487, + 0.4770168284742763, + 0.006835511403011292, + 0.011200087973044965, + 0.020796041319268533 + ], + "per_source": [ + 0.0030099407416697968, + 0.30826237672231105, + 0.030874093815323564, + 0.005310674836551046, + 0.6243418715179362, + 0.02131301441017302, + 0.0048036204646279375, + 0.002084407491407262 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.048363233996627016, + "mmd2_mixture": 0.060122682064390254, + "mmd2_uniform": 0.15285688303099165, + "mmd2_single_source": [ + 0.4522663898122707, + 0.14701298584193184, + 0.12241687119492284, + 0.2726521753211397, + 0.08995170742180114, + 0.3744254300178685, + 0.33957846092746485, + 0.23688038103280662 + ], + "median_squared_distance": 0.18662990120119327, + "gammas": [ + 1.3395495490858758, + 2.6790990981717515, + 5.358198196343503, + 10.716396392687006, + 21.432792785374012 + ], + "max_weight": 0.4770168284742763, + "entropy": 1.1759448067763238, + "effective_sources": 3.24120380896937, + "min_weight": 0.005265163985947525 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.4624401356050084, + 0.12211205990556462, + 0.29128492533416295, + 0.4206953767024899, + 0.07022529260621624, + 0.3185315424880186, + 0.42807307387367177, + 0.489454149021297 + ], + "standardized": [ + 0.9323134642607491, + -1.3822019864377946, + -0.23168506223542185, + 0.6484142024470035, + -1.735075384714815, + -0.04638528568228213, + 0.6985887114356643, + 1.1160313409268967 + ], + "tau": 0.5, + "max_weight": 0.6243418715179362, + "entropy": 0.9300659298676801, + "effective_sources": 2.534676282981126, + "min_weight": 0.002084407491407262 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "microns_basil", + "seconds": 6.7 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.41247125221353, + 0.47321264983034456, + 0.19639403256049165, + 0.31013587127331976, + 0.5894646206275136, + 0.13744302537529526, + 0.3054434345563225, + 0.35012697230796425 + ], + "ensemble_val_dice": { + "mixture": 0.4205061683646676, + "mixture@raw": 0.462918877571213, + "mixture@logit": 0.3665782427199844, + "per_source": 0.5892228765147073, + "per_source@raw": 0.5905348345941427, + "per_source@logit": 0.5781053080039771, + "uniform": 0.41124230184319877, + "uniform@raw": 0.4669572710990906, + "uniform@logit": 0.22782834015245554 + }, + "val_patches": 735 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 2.642308473587036, + 4.074584007263184 + ], + "top_source": "microns_basil", + "weight_stats": { + "max_weight": 0.4770168284742763, + "entropy": 1.1759448067763238, + "effective_sources": 3.24120380896937, + "min_weight": 0.005265163985947525 + }, + "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": 649.8, + "round0_val": { + "val_dice": 0.38345661880027987, + "val_vi": 0.6425668875430072, + "val_ece": 0.07046950570811931 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 28, + "selection_seconds": 119.3, + "round_seconds": 560.4, + "teacher_dice_labeled": [ + 0.30923963036625823, + 0.40141865169099034, + 0.21298253591750957, + 0.34862235687346954, + 0.5216517007952947, + 0.16310498252910807, + 0.20307595354559912, + 0.2304380164604082 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 232196, + "evidence_nonempty_patches": 28, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.4117972424112869, + "restart_teacher_dice": 0.5216517007952947, + "weights": [ + 0.0012112506663100886, + 0.03339868283529253, + 0.0345908486511191, + 0.009352845080749347, + 0.9180148960941229, + 0.0003647022371434577, + 0.000891211066458353, + 0.0021755633688042347 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5209970401005555, + "uniform": 0.3803190293127032, + "top1": 0.5216517007952947, + "top2": 0.5224545359489741, + "top3": 0.5224050337353666 + }, + "val_dice": 0.7174647731793253, + "val_vi": 0.5862075152354411, + "val_ece": 0.01040988329108115 + }, + { + "round": 2, + "n_labeled": 56, + "selection_seconds": 37.0, + "round_seconds": 80.0, + "teacher_dice_labeled": [ + 0.32968286739321734, + 0.3899514887912725, + 0.1956341759576112, + 0.29252135560018844, + 0.4860608454924249, + 0.12947791408419423, + 0.21962028025532235, + 0.2707968487830147 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 551979, + "evidence_nonempty_patches": 56, + "first_correction": false, + "restart": "", + "weights": [ + 0.0022035355669607175, + 0.044160122494152035, + 0.0, + 0.0, + 0.9536363419388872, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.4822414733525932, + "uniform": 0.31097768281455124, + "top1": 0.4860608454924249, + "top2": 0.4872193497639249, + "top3": 0.4872618953534774 + }, + "pseudo_student_dice": 0.6088462828961946, + "pseudo_ensemble_dice": 0.45378946490070243, + "pseudo_source": "student", + "consistency_scale": 0.363, + "val_dice": 0.7390804504089595, + "val_vi": 0.5685614483112967, + "val_ece": 0.011165779298286696 + }, + { + "round": 3, + "n_labeled": 84, + "selection_seconds": 36.7, + "round_seconds": 393.4, + "teacher_dice_labeled": [ + 0.37086031388801677, + 0.4453667316802009, + 0.20723350644101518, + 0.33534482250576847, + 0.5762288381413482, + 0.15528764228810432, + 0.23703538953587494, + 0.2977320132507578 + ], + "teacher_dice_pool": [ + 0.3898740464376021, + 0.4755429928578754, + 0.23910686384852367, + 0.3886655657749072, + 0.668424476613459, + 0.19800193940952124, + 0.2762308341946166, + 0.30910087439506295 + ], + "evidence_fg_pixels": 913408, + "evidence_nonempty_patches": 84, + "first_correction": false, + "restart": "", + "weights": [ + 0.0013310497510279384, + 0.03075660939664637, + 0.019379319848684722, + 0.004859875828552802, + 0.9402000187080747, + 0.00020013967092648373, + 0.0007426933897470523, + 0.002530293406339934 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5754038750517916, + "uniform": 0.36575840911571345, + "top1": 0.5762288381413482, + "top2": 0.5777510078628051, + "top3": 0.5778078801028542 + }, + "pseudo_student_dice": 0.6361807161340302, + "pseudo_ensemble_dice": 0.480113758300175, + "pseudo_source": "student", + "consistency_scale": 0.454, + "val_dice": 0.7495523539931895, + "val_vi": 0.5489239394094233, + "val_ece": 0.011783477941583785 + }, + { + "round": 4, + "n_labeled": 112, + "selection_seconds": 36.4, + "round_seconds": 404.4, + "teacher_dice_labeled": [ + 0.37321174857677786, + 0.44278635107161046, + 0.20425591711599464, + 0.327772126479294, + 0.5657931122865841, + 0.1549070230922441, + 0.24313737842154926, + 0.29954958369556517 + ], + "teacher_dice_pool": [ + 0.4001746244878322, + 0.48220145867073366, + 0.23499525796249365, + 0.3786014216751024, + 0.6490690617814941, + 0.19743975469322234, + 0.26903872285765806, + 0.31888259615125375 + ], + "evidence_fg_pixels": 1179147, + "evidence_nonempty_patches": 112, + "first_correction": false, + "restart": "", + "weights": [ + 0.0015053565082109753, + 0.033110438433428105, + 0.020779732351032795, + 0.004977030440218871, + 0.9356887305472527, + 0.00022024857037611335, + 0.0008720534096111227, + 0.0028464097398693314 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5647967946865813, + "uniform": 0.360234486572068, + "top1": 0.5657931122865841, + "top2": 0.5673055047680104, + "top3": 0.5673725014471637 + }, + "pseudo_student_dice": 0.6648006350571386, + "pseudo_ensemble_dice": 0.4757541945192298, + "pseudo_source": "student", + "consistency_scale": 0.549, + "val_dice": 0.7550029429579439, + "val_vi": 0.5302964007298193, + "val_ece": 0.014572060405993651 + }, + { + "round": 5, + "n_labeled": 140, + "selection_seconds": 36.2, + "round_seconds": 417.5, + "teacher_dice_labeled": [ + 0.3625410938591591, + 0.4355784897218964, + 0.20264899618503762, + 0.3295045734408726, + 0.5588364198985812, + 0.15605754418445766, + 0.2401934049537906, + 0.2913797382341989 + ], + "teacher_dice_pool": [ + 0.3786420858421469, + 0.47081271613508735, + 0.24180065149346994, + 0.3775880634035507, + 0.6415367429953798, + 0.19634054315397, + 0.2792974265186966, + 0.30249360274341297 + ], + "evidence_fg_pixels": 1474014, + "evidence_nonempty_patches": 140, + "first_correction": false, + "restart": "", + "weights": [ + 0.0014483361718081902, + 0.03297871007616184, + 0.021889369773742826, + 0.005420836341058693, + 0.9343098877686563, + 0.0002384963984002584, + 0.0009064203503226087, + 0.002807943119849266 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.557823971764361, + "uniform": 0.3578505232668598, + "top1": 0.5588364198985812, + "top2": 0.5602451647369182, + "top3": 0.5603049948006693 + }, + "pseudo_student_dice": 0.643506272736545, + "pseudo_ensemble_dice": 0.44930986227259595, + "pseudo_source": "student", + "consistency_scale": 0.478, + "val_dice": 0.7549671911288235, + "val_vi": 0.5302006794911862, + "val_ece": 0.013241020018444667 + }, + { + "round": 6, + "n_labeled": 168, + "selection_seconds": 35.9, + "round_seconds": 428.2, + "teacher_dice_labeled": [ + 0.36353511631327795, + 0.43497302604973287, + 0.20074485294527286, + 0.3327935590239828, + 0.5582466822186675, + 0.1567819919136381, + 0.23909282753480066, + 0.2925195195546896 + ], + "teacher_dice_pool": [ + 0.3839850647413449, + 0.46831841356244397, + 0.23537658668363787, + 0.3826950837371619, + 0.6349460257289151, + 0.19188588782082197, + 0.27303537783430637, + 0.30658306254260204 + ], + "evidence_fg_pixels": 1744619, + "evidence_nonempty_patches": 168, + "first_correction": false, + "restart": "", + "weights": [ + 0.0014714657544328564, + 0.03297372159029093, + 0.02160366658468838, + 0.005635260229233188, + 0.9343154793074689, + 0.00024165281514750797, + 0.0009018071986931946, + 0.0028569465200449728 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.557289318573895, + "uniform": 0.3591946978714461, + "top1": 0.5582466822186675, + "top2": 0.5595871272722308, + "top3": 0.5596502336834291 + }, + "pseudo_student_dice": 0.6815412134626851, + "pseudo_ensemble_dice": 0.5090477033045854, + "pseudo_source": "student", + "consistency_scale": 0.605, + "val_dice": 0.7581036709465326, + "val_vi": 0.5190172698185712, + "val_ece": 0.01572609371733461 + }, + { + "round": 7, + "n_labeled": 168, + "selection_seconds": 0.0, + "round_seconds": 115.9, + "final_consolidation": true, + "val_dice": 0.7585554059155132, + "val_vi": 0.5180210703137048, + "val_ece": 0.0218110641494762 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.5072424045387947, + "auroc_top20_error": 0.8403184472456204 + }, + "al_seconds": 2430.0, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.0014714657544328564, + 0.03297372159029093, + 0.02160366658468838, + 0.005635260229233188, + 0.9343154793074689, + 0.00024165281514750797, + 0.0009018071986931946, + 0.0028569465200449728 + ], + "weight_stats_final": { + "max_weight": 0.9343154793074689, + "entropy": 0.322684680800098, + "effective_sources": 1.3808298800800942, + "min_weight": 0.00024165281514750797 + }, + "top_source_final": "microns_basil", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.7502993844021068, + "patch_iou": 0.610265425863319, + "patch_precision": 0.726259786758238, + "patch_recall": 0.779129192487549, + "patch_vi": 0.5263768898094677, + "n_patches": 735, + "cal_ece": 0.020970448656017486, + "cal_nll": 0.19405148446924358, + "cal_brier": 0.060640539511800845, + "cal_mean_confidence": 0.934816377652616, + "cal_accuracy": 0.9138459289965987, + "cal_predicted_prevalence": 0.2200547406462585, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.21815898913369627 + }, + "test_notta": { + "patch_dice": 0.7502993844021068, + "patch_iou": 0.610265425863319, + "patch_precision": 0.726259786758238, + "patch_recall": 0.779129192487549, + "patch_vi": 0.5263768898094677, + "n_patches": 735, + "cal_ece": 0.020970448656017486, + "cal_nll": 0.19405148446924358, + "cal_brier": 0.060640539511800845, + "cal_mean_confidence": 0.934816377652616, + "cal_accuracy": 0.9138459289965987, + "cal_predicted_prevalence": 0.2200547406462585, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.21815898913369627 + }, + "test": { + "patch_dice": 0.7591598943652398, + "patch_iou": 0.6215398307814628, + "patch_precision": 0.7383584817490859, + "patch_recall": 0.7840826714217108, + "patch_vi": 0.5102328361065348, + "n_patches": 735, + "image_dice": 0.8083227144183348, + "image_iou": 0.6787329371214111, + "image_precision": 0.7897767967423324, + "image_recall": 0.828165408514625, + "image_vi": 0.5166640790459083, + "n_images": 15, + "cal_ece": 0.012392638944076592, + "cal_nll": 0.1813798769940655, + "cal_brier": 0.05722207631886344, + "cal_mean_confidence": 0.9298206257638044, + "cal_accuracy": 0.9174279868197279, + "cal_predicted_prevalence": 0.21780665391156462, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.21693074734329712 + }, + "eval_seconds": 38.9, + "total_seconds": 3188.7, + "resources": { + "phases": { + "setup": { + "seconds": 1.48, + "peak_cpu_rss_mb": 2191.8, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 63.32, + "peak_cpu_rss_mb": 4181.8, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 649.83, + "peak_cpu_rss_mb": 5156.0, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4638.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2430.01, + "peak_cpu_rss_mb": 5156.0, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4646.0, + "rounds": 6 + }, + "inference": { + "seconds": 38.92, + "peak_cpu_rss_mb": 5156.0, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4646.0, + "patches": 735, + "images": 15, + "patches_per_second": 18.9, + "seconds_per_image": 2.595 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5156.0, + "total_seconds": 3183.6 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/snemi3d/macs-v11-norestart-acq-laplace_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..5ff93f36cd81ea77dab0e05892fc682a61cf2807 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/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 +1,28,119.3,560.4,"[0.30923963036625823, 0.40141865169099034, 0.21298253591750957, 0.34862235687346954, 0.5216517007952947, 0.16310498252910807, 0.20307595354559912, 0.2304380164604082]",,232196,28,True,kept,0.4117972424112869,0.5216517007952947,"[0.0012112506663100886, 0.03339868283529253, 0.0345908486511191, 0.009352845080749347, 0.9180148960941229, 0.0003647022371434577, 0.000891211066458353, 0.0021755633688042347]",mixture,"{'mixture': 0.5209970401005555, 'uniform': 0.3803190293127032, 'top1': 0.5216517007952947, 'top2': 0.5224545359489741, 'top3': 0.5224050337353666}",0.7174647731793253,0.5862075152354411,0.01040988329108115,,,,, +2,56,37.0,80.0,"[0.32968286739321734, 0.3899514887912725, 0.1956341759576112, 0.29252135560018844, 0.4860608454924249, 0.12947791408419423, 0.21962028025532235, 0.2707968487830147]",,551979,56,False,,,,"[0.0022035355669607175, 0.044160122494152035, 0.0, 0.0, 0.9536363419388872, 0.0, 0.0, 0.0]",top3,"{'mixture': 0.4822414733525932, 'uniform': 0.31097768281455124, 'top1': 0.4860608454924249, 'top2': 0.4872193497639249, 'top3': 0.4872618953534774}",0.7390804504089595,0.5685614483112967,0.011165779298286696,0.6088462828961946,0.45378946490070243,student,0.363, +3,84,36.7,393.4,"[0.37086031388801677, 0.4453667316802009, 0.20723350644101518, 0.33534482250576847, 0.5762288381413482, 0.15528764228810432, 0.23703538953587494, 0.2977320132507578]","[0.3898740464376021, 0.4755429928578754, 0.23910686384852367, 0.3886655657749072, 0.668424476613459, 0.19800193940952124, 0.2762308341946166, 0.30910087439506295]",913408,84,False,,,,"[0.0013310497510279384, 0.03075660939664637, 0.019379319848684722, 0.004859875828552802, 0.9402000187080747, 0.00020013967092648373, 0.0007426933897470523, 0.002530293406339934]",mixture,"{'mixture': 0.5754038750517916, 'uniform': 0.36575840911571345, 'top1': 0.5762288381413482, 'top2': 0.5777510078628051, 'top3': 0.5778078801028542}",0.7495523539931895,0.5489239394094233,0.011783477941583785,0.6361807161340302,0.480113758300175,student,0.454, +4,112,36.4,404.4,"[0.37321174857677786, 0.44278635107161046, 0.20425591711599464, 0.327772126479294, 0.5657931122865841, 0.1549070230922441, 0.24313737842154926, 0.29954958369556517]","[0.4001746244878322, 0.48220145867073366, 0.23499525796249365, 0.3786014216751024, 0.6490690617814941, 0.19743975469322234, 0.26903872285765806, 0.31888259615125375]",1179147,112,False,,,,"[0.0015053565082109753, 0.033110438433428105, 0.020779732351032795, 0.004977030440218871, 0.9356887305472527, 0.00022024857037611335, 0.0008720534096111227, 0.0028464097398693314]",mixture,"{'mixture': 0.5647967946865813, 'uniform': 0.360234486572068, 'top1': 0.5657931122865841, 'top2': 0.5673055047680104, 'top3': 0.5673725014471637}",0.7550029429579439,0.5302964007298193,0.014572060405993651,0.6648006350571386,0.4757541945192298,student,0.549, +5,140,36.2,417.5,"[0.3625410938591591, 0.4355784897218964, 0.20264899618503762, 0.3295045734408726, 0.5588364198985812, 0.15605754418445766, 0.2401934049537906, 0.2913797382341989]","[0.3786420858421469, 0.47081271613508735, 0.24180065149346994, 0.3775880634035507, 0.6415367429953798, 0.19634054315397, 0.2792974265186966, 0.30249360274341297]",1474014,140,False,,,,"[0.0014483361718081902, 0.03297871007616184, 0.021889369773742826, 0.005420836341058693, 0.9343098877686563, 0.0002384963984002584, 0.0009064203503226087, 0.002807943119849266]",mixture,"{'mixture': 0.557823971764361, 'uniform': 0.3578505232668598, 'top1': 0.5588364198985812, 'top2': 0.5602451647369182, 'top3': 0.5603049948006693}",0.7549671911288235,0.5302006794911862,0.013241020018444667,0.643506272736545,0.44930986227259595,student,0.478, +6,168,35.9,428.2,"[0.36353511631327795, 0.43497302604973287, 0.20074485294527286, 0.3327935590239828, 0.5582466822186675, 0.1567819919136381, 0.23909282753480066, 0.2925195195546896]","[0.3839850647413449, 0.46831841356244397, 0.23537658668363787, 0.3826950837371619, 0.6349460257289151, 0.19188588782082197, 0.27303537783430637, 0.30658306254260204]",1744619,168,False,,,,"[0.0014714657544328564, 0.03297372159029093, 0.02160366658468838, 0.005635260229233188, 0.9343154793074689, 0.00024165281514750797, 0.0009018071986931946, 0.0028569465200449728]",mixture,"{'mixture': 0.557289318573895, 'uniform': 0.3591946978714461, 'top1': 0.5582466822186675, 'top2': 0.5595871272722308, 'top3': 0.5596502336834291}",0.7581036709465326,0.5190172698185712,0.01572609371733461,0.6815412134626851,0.5090477033045854,student,0.605, +7,168,0.0,115.9,,,,,,,,,,,,0.7585554059155132,0.5180210703137048,0.0218110641494762,,,,,True diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/log.txt b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..d38f87f585ba447ddc79967691f86b9d7a19a86f --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/log.txt @@ -0,0 +1,103 @@ +[06:00:54][ 0s][START] target=snemi3d variant=macs-v11-norestart-acq-margin budget=5.0% seed=1 | {"task": "membrane", "target": "snemi3d", "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": 8, "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} +[06:00:54][ 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. +[06:00:54][ 0s][DATA] snemi3d: train=3430 val=735 test=735 patches +[06:00:56][ 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=4.07 +[06:00:57][ 3s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[06:01:03][ 9s][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 +[06:01:03][ 9s][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 +[06:01:56][ 62s][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 +[06:01:56][ 62s][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 +[06:01:56][ 62s][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 +[06:01:56][ 62s][INIT] student initialised from teacher microns (weight 0.477) +[06:01:59][ 66s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[06:01:59][ 66s][KD] distilling the mixture ensemble (prob space) into the student +[06:04:19][ 205s][KD] soft labels for 3430 patches from 8 teachers in 140s +[06:06:01][ 307s][KD] epoch 1/5 | patches=3430 | loss=0.4461 | 102s +[06:07:43][ 409s][KD] epoch 2/5 | patches=3430 | loss=0.3728 | 102s +[06:09:25][ 511s][KD] epoch 3/5 | patches=3430 | loss=0.3563 | 102s +[06:11:07][ 613s][KD] epoch 4/5 | patches=3430 | loss=0.3506 | 102s +[06:12:49][ 715s][KD] epoch 5/5 | patches=3430 | loss=0.3475 | 102s +[06:12:53][ 719s][ROUND0] before any target label: val_dice=0.3968 val_vi=0.6469 +[06:12:53][ 719s][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 +[06:13:11][ 737s][AL] round-1 uncertainty quality on 3430 patches: {'spearman_uncertainty_error': 0.3088075353000357, 'auroc_top20_error': 0.7218027777541671} +[06:14:33][ 819s][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 +[06:14:38][ 824s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.543, uniform=0.427, top1=0.543, top2=0.545, top3=0.545 -> using mixture +[06:14:39][ 825s][WEIGHTS] round 1: student kept (labeled Dice 0.433 vs best teacher microns 0.543); distilling the corrected ensemble into it +[06:14:39][ 825s][WEIGHTS] round 1: teacher Dice on 28 labeled patches + 0 blank (279577 fg px): c-elegans=0.358, cerebellum=0.442, flywire=0.213, berghia=0.348, microns=0.543, oct-glia=0.243, oct-sfl=0.282, mouse-brain=0.275 | weights: c-elegans=0.002, cerebellum=0.040, flywire=0.028, berghia=0.008, microns=0.917, oct-glia=0.001, oct-sfl=0.002, mouse-brain=0.003 +[06:16:57][ 964s][KD] soft labels for 3402 patches from 8 teachers in 139s +[06:18:38][ 1065s][KD] epoch 1/2 | patches=3402 | loss=0.2525 | 101s +[06:20:19][ 1166s][KD] epoch 2/2 | patches=3402 | loss=0.2504 | 101s +[06:21:41][ 1247s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[06:21:44][ 1250s][FT] round 1/6 epoch 1/5 | patches=28 | loss=1.0453 +[06:21:46][ 1252s][FT] round 1/6 epoch 2/5 | patches=28 | loss=0.9823 +[06:21:49][ 1255s][FT] round 1/6 epoch 3/5 | patches=28 | loss=0.8996 +[06:21:51][ 1257s][FT] round 1/6 epoch 4/5 | patches=28 | loss=0.8659 +[06:21:54][ 1260s][FT] round 1/6 epoch 5/5 | patches=28 | loss=0.8378 +[06:21:58][ 1264s][AL] round 1/6 | labeled=28 | selection_seconds=100.1000 round_seconds=540.4000 restart_student_dice=0.4330 restart_teacher_dice=0.5428 val_dice=0.7295 val_vi=0.5884 val_ece=0.0197 +[06:22:25][ 1291s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.413, uniform=0.307, top1=0.416, top2=0.422, top3=0.422 -> using top3 +[06:22:25][ 1291s][WEIGHTS] round 2: teacher Dice on 56 labeled patches + 0 blank (566444 fg px): c-elegans=0.321, cerebellum=0.413, flywire=0.204, berghia=0.296, microns=0.416, oct-glia=0.179, oct-sfl=0.203, mouse-brain=0.273 | weights: c-elegans=0.004, cerebellum=0.104, flywire=0.000, berghia=0.000, microns=0.892, oct-glia=0.000, oct-sfl=0.000, mouse-brain=0.000 +[06:22:39][ 1305s][CONS] round 2: on the 28 new patches student Dice=0.611 vs ensemble 0.235 -> pseudo-labels from the student | consistency x0.37 +[06:22:43][ 1309s][FT] round 2/6 epoch 1/5 | patches=56 | loss=0.8161 +[06:22:47][ 1313s][FT] round 2/6 epoch 2/5 | patches=56 | loss=0.7693 +[06:22:50][ 1317s][FT] round 2/6 epoch 3/5 | patches=56 | loss=0.7324 +[06:22:54][ 1321s][FT] round 2/6 epoch 4/5 | patches=56 | loss=0.7172 +[06:22:58][ 1325s][FT] round 2/6 epoch 5/5 | patches=56 | loss=0.7192 +[06:23:03][ 1329s][AL] round 2/6 | labeled=56 | selection_seconds=18.1000 round_seconds=60.7000 pseudo_student_dice=0.6111 pseudo_ensemble_dice=0.2345 consistency_scale=0.3700 val_dice=0.7433 val_vi=0.5467 val_ece=0.0149 +[06:26:02][ 1508s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.390, cerebellum=0.476, flywire=0.236, berghia=0.386, microns=0.659, oct-glia=0.187, oct-sfl=0.263, mouse-brain=0.311 | blended with labeled Dice at 0.50 +[06:28:33][ 1660s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.539, uniform=0.366, top1=0.540, top2=0.542, top3=0.542 -> using mixture +[06:28:33][ 1660s][WEIGHTS] round 3: teacher Dice on 84 labeled patches + 0 blank (865230 fg px): c-elegans=0.363, cerebellum=0.437, flywire=0.211, berghia=0.339, microns=0.540, oct-glia=0.176, oct-sfl=0.225, mouse-brain=0.293 | weights: c-elegans=0.002, cerebellum=0.040, flywire=0.028, berghia=0.007, microns=0.919, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[06:28:47][ 1673s][CONS] round 3: on the 28 new patches student Dice=0.625 vs ensemble 0.429 -> pseudo-labels from the student | consistency x0.42 +[06:28:53][ 1679s][FT] round 3/6 epoch 1/5 | patches=84 | loss=0.7737 +[06:28:59][ 1685s][FT] round 3/6 epoch 2/5 | patches=84 | loss=0.7402 +[06:29:04][ 1690s][FT] round 3/6 epoch 3/5 | patches=84 | loss=0.7220 +[06:29:11][ 1697s][FT] round 3/6 epoch 4/5 | patches=84 | loss=0.7066 +[06:29:16][ 1702s][FT] round 3/6 epoch 5/5 | patches=84 | loss=0.6914 +[06:29:21][ 1707s][AL] round 3/6 | labeled=84 | selection_seconds=18.0000 round_seconds=373.7000 pseudo_student_dice=0.6252 pseudo_ensemble_dice=0.4286 consistency_scale=0.4170 val_dice=0.7527 val_vi=0.5420 val_ece=0.0168 +[06:32:22][ 1888s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.408, cerebellum=0.489, flywire=0.238, berghia=0.386, microns=0.657, oct-glia=0.193, oct-sfl=0.272, mouse-brain=0.322 | blended with labeled Dice at 0.50 +[06:34:55][ 2041s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.549, uniform=0.372, top1=0.550, top2=0.551, top3=0.551 -> using mixture +[06:34:55][ 2041s][WEIGHTS] round 4: teacher Dice on 112 labeled patches + 0 blank (1141567 fg px): c-elegans=0.375, cerebellum=0.441, flywire=0.210, berghia=0.339, microns=0.550, oct-glia=0.175, oct-sfl=0.235, mouse-brain=0.299 | weights: c-elegans=0.002, cerebellum=0.038, flywire=0.025, berghia=0.006, microns=0.924, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[06:35:09][ 2055s][CONS] round 4: on the 28 new patches student Dice=0.663 vs ensemble 0.499 -> pseudo-labels from the student | consistency x0.54 +[06:35:16][ 2062s][FT] round 4/6 epoch 1/5 | patches=112 | loss=0.7368 +[06:35:23][ 2069s][FT] round 4/6 epoch 2/5 | patches=112 | loss=0.7219 +[06:35:31][ 2077s][FT] round 4/6 epoch 3/5 | patches=112 | loss=0.7088 +[06:35:38][ 2084s][FT] round 4/6 epoch 4/5 | patches=112 | loss=0.6966 +[06:35:45][ 2091s][FT] round 4/6 epoch 5/5 | patches=112 | loss=0.6849 +[06:35:49][ 2096s][AL] round 4/6 | labeled=112 | selection_seconds=17.8000 round_seconds=384.6000 pseudo_student_dice=0.6629 pseudo_ensemble_dice=0.4986 consistency_scale=0.5430 val_dice=0.7510 val_vi=0.5316 val_ece=0.0170 +[06:38:52][ 2279s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.382, cerebellum=0.468, flywire=0.238, berghia=0.375, microns=0.650, oct-glia=0.198, oct-sfl=0.290, mouse-brain=0.304 | blended with labeled Dice at 0.50 +[06:41:28][ 2434s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.553, uniform=0.381, top1=0.554, top2=0.555, top3=0.555 -> using mixture +[06:41:28][ 2434s][WEIGHTS] round 5: teacher Dice on 140 labeled patches + 0 blank (1422513 fg px): c-elegans=0.364, cerebellum=0.434, flywire=0.211, berghia=0.335, microns=0.554, oct-glia=0.178, oct-sfl=0.249, mouse-brain=0.291 | weights: c-elegans=0.002, cerebellum=0.034, flywire=0.025, berghia=0.006, microns=0.929, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[06:41:42][ 2448s][CONS] round 5: on the 28 new patches student Dice=0.661 vs ensemble 0.511 -> pseudo-labels from the student | consistency x0.54 +[06:41:51][ 2457s][FT] round 5/6 epoch 1/5 | patches=140 | loss=0.6910 +[06:42:00][ 2466s][FT] round 5/6 epoch 2/5 | patches=140 | loss=0.6891 +[06:42:09][ 2475s][FT] round 5/6 epoch 3/5 | patches=140 | loss=0.6933 +[06:42:18][ 2484s][FT] round 5/6 epoch 4/5 | patches=140 | loss=0.6768 +[06:42:27][ 2493s][FT] round 5/6 epoch 5/5 | patches=140 | loss=0.6743 +[06:42:31][ 2497s][AL] round 5/6 | labeled=140 | selection_seconds=17.7000 round_seconds=397.3000 pseudo_student_dice=0.6609 pseudo_ensemble_dice=0.5111 consistency_scale=0.5360 val_dice=0.7545 val_vi=0.5338 val_ece=0.0179 +[06:45:36][ 2682s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.394, cerebellum=0.475, flywire=0.242, berghia=0.381, microns=0.646, oct-glia=0.202, oct-sfl=0.277, mouse-brain=0.310 | blended with labeled Dice at 0.50 +[06:48:13][ 2839s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.552, uniform=0.374, top1=0.553, top2=0.554, top3=0.554 -> using mixture +[06:48:13][ 2839s][WEIGHTS] round 6: teacher Dice on 168 labeled patches + 0 blank (1683796 fg px): c-elegans=0.368, cerebellum=0.434, flywire=0.209, berghia=0.336, microns=0.553, oct-glia=0.177, oct-sfl=0.237, mouse-brain=0.292 | weights: c-elegans=0.002, cerebellum=0.034, flywire=0.025, berghia=0.006, microns=0.930, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[06:48:26][ 2853s][CONS] round 6: on the 28 new patches student Dice=0.643 vs ensemble 0.475 -> pseudo-labels from the student | consistency x0.48 +[06:48:37][ 2863s][FT] round 6/6 epoch 1/5 | patches=168 | loss=0.6883 +[06:48:47][ 2874s][FT] round 6/6 epoch 2/5 | patches=168 | loss=0.6804 +[06:48:58][ 2884s][FT] round 6/6 epoch 3/5 | patches=168 | loss=0.6762 +[06:49:08][ 2895s][FT] round 6/6 epoch 4/5 | patches=168 | loss=0.6747 +[06:49:19][ 2905s][FT] round 6/6 epoch 5/5 | patches=168 | loss=0.6631 +[06:49:23][ 2909s][AL] round 6/6 | labeled=168 | selection_seconds=17.5000 round_seconds=408.0000 pseudo_student_dice=0.6425 pseudo_ensemble_dice=0.4749 consistency_scale=0.4750 val_dice=0.7557 val_vi=0.5299 val_ece=0.0180 +[06:49:33][ 2920s][CONS] final consolidation: pseudo-labels refreshed from the student +[06:49:44][ 2930s][FT] final 1/10 epoch 1/1 | patches=168 | loss=0.6598 +[06:49:54][ 2940s][FT] final 2/10 epoch 1/1 | patches=168 | loss=0.6540 +[06:50:05][ 2951s][FT] final 3/10 epoch 1/1 | patches=168 | loss=0.6513 +[06:50:15][ 2962s][FT] final 4/10 epoch 1/1 | patches=168 | loss=0.6429 +[06:50:26][ 2972s][FT] final 5/10 epoch 1/1 | patches=168 | loss=0.6299 +[06:50:37][ 2983s][FT] final 6/10 epoch 1/1 | patches=168 | loss=0.6251 +[06:50:47][ 2993s][FT] final 7/10 epoch 1/1 | patches=168 | loss=0.6197 +[06:50:58][ 3004s][FT] final 8/10 epoch 1/1 | patches=168 | loss=0.6161 +[06:51:08][ 3014s][FT] final 9/10 epoch 1/1 | patches=168 | loss=0.6184 +[06:51:19][ 3025s][FT] final 10/10 epoch 1/1 | patches=168 | loss=0.6177 +[06:51:23][ 3029s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=115.3000 val_dice=0.7554 val_vi=0.5248 val_ece=0.0231 +[06:51:23][ 3029s][WEIGHTS] final: c-elegans=0.002, cerebellum=0.034, flywire=0.025, berghia=0.006, microns=0.930, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[06:51:57][ 3064s][TEST] raw student (no EMA, no TTA): patch_dice=0.7491 patch_vi=0.5299 +[06:52:02][ 3068s][TEST] final model without TTA: patch_dice=0.7491 patch_vi=0.5299 +[06:52:02][ 3068s][RESOURCES] {"phases": {"setup": {"seconds": 1.48, "peak_cpu_rss_mb": 2192.2, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 63.24, "peak_cpu_rss_mb": 4183.8, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 649.47, "peak_cpu_rss_mb": 5161.3, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 2309.76, "peak_cpu_rss_mb": 5161.3, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 38.77, "peak_cpu_rss_mb": 5161.3, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 735, "images": 15, "patches_per_second": 19.0, "seconds_per_image": 2.585}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5161.3, "total_seconds": 3062.7} +[06:52:02][ 3068s][TEST] patch_dice=0.7606 patch_iou=0.6231 patch_precision=0.7390 patch_recall=0.7867 patch_vi=0.5100 image_dice=0.8081 image_iou=0.6785 image_precision=0.7903 image_recall=0.8273 image_vi=0.5165 cal_ece=0.0104 cal_nll=0.1794 cal_brier=0.0567 cal_mean_confidence=0.9284 cal_accuracy=0.9180 cal_predicted_prevalence=0.2157 cal_true_prevalence=0.2059 cal_mean_probability=0.2157 +[06:52:02][ 3068s][DONE] snemi3d macs-v11-norestart-acq-margin b5 s1 | test patch Dice=0.7606 VI=0.5100 | image Dice=0.8081 VI=0.5165 | 3068s diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/result.json b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..f636873a24437ccf6b9da23033463c30f4483745 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/result.json @@ -0,0 +1,785 @@ +{ + "target": "snemi3d", + "variant": "macs-v11-norestart-acq-margin", + "budget": 5.0, + "seed": 1, + "config": { + "task": "membrane", + "target": "snemi3d", + "variant": "macs-v11-norestart-acq-margin", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "snemi3d", + "--variants", + "macs-v11-norestart-acq-margin", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.005265163985947525, + 0.05775356661687405, + 0.3937117170688725, + 0.02742108315870487, + 0.4770168284742763, + 0.006835511403011292, + 0.011200087973044965, + 0.020796041319268533 + ], + "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": [ + 0.005265163985947525, + 0.05775356661687405, + 0.3937117170688725, + 0.02742108315870487, + 0.4770168284742763, + 0.006835511403011292, + 0.011200087973044965, + 0.020796041319268533 + ], + "per_source": [ + 0.0030099407416697968, + 0.30826237672231105, + 0.030874093815323564, + 0.005310674836551046, + 0.6243418715179362, + 0.02131301441017302, + 0.0048036204646279375, + 0.002084407491407262 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.048363233996627016, + "mmd2_mixture": 0.060122682064390254, + "mmd2_uniform": 0.15285688303099165, + "mmd2_single_source": [ + 0.4522663898122707, + 0.14701298584193184, + 0.12241687119492284, + 0.2726521753211397, + 0.08995170742180114, + 0.3744254300178685, + 0.33957846092746485, + 0.23688038103280662 + ], + "median_squared_distance": 0.18662990120119327, + "gammas": [ + 1.3395495490858758, + 2.6790990981717515, + 5.358198196343503, + 10.716396392687006, + 21.432792785374012 + ], + "max_weight": 0.4770168284742763, + "entropy": 1.1759448067763238, + "effective_sources": 3.24120380896937, + "min_weight": 0.005265163985947525 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.4624401356050084, + 0.12211205990556462, + 0.29128492533416295, + 0.4206953767024899, + 0.07022529260621624, + 0.3185315424880186, + 0.42807307387367177, + 0.489454149021297 + ], + "standardized": [ + 0.9323134642607491, + -1.3822019864377946, + -0.23168506223542185, + 0.6484142024470035, + -1.735075384714815, + -0.04638528568228213, + 0.6985887114356643, + 1.1160313409268967 + ], + "tau": 0.5, + "max_weight": 0.6243418715179362, + "entropy": 0.9300659298676801, + "effective_sources": 2.534676282981126, + "min_weight": 0.002084407491407262 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "microns_basil", + "seconds": 6.7 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.41247125221353, + 0.47321264983034456, + 0.19639403256049165, + 0.31013587127331976, + 0.5894646206275136, + 0.13744302537529526, + 0.3054434345563225, + 0.35012697230796425 + ], + "ensemble_val_dice": { + "mixture": 0.4205061683646676, + "mixture@raw": 0.462918877571213, + "mixture@logit": 0.3665782427199844, + "per_source": 0.5892228765147073, + "per_source@raw": 0.5905348345941427, + "per_source@logit": 0.5781053080039771, + "uniform": 0.41124230184319877, + "uniform@raw": 0.4669572710990906, + "uniform@logit": 0.22782834015245554 + }, + "val_patches": 735 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 2.642308473587036, + 4.074584007263184 + ], + "top_source": "microns_basil", + "weight_stats": { + "max_weight": 0.4770168284742763, + "entropy": 1.1759448067763238, + "effective_sources": 3.24120380896937, + "min_weight": 0.005265163985947525 + }, + "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": 649.5, + "round0_val": { + "val_dice": 0.39676821516355265, + "val_vi": 0.6468693296003585, + "val_ece": 0.07077332120435081 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 28, + "selection_seconds": 100.1, + "round_seconds": 540.4, + "teacher_dice_labeled": [ + 0.35791764681874827, + 0.4417052610943664, + 0.21306314427094966, + 0.3478407216013448, + 0.5427570406835327, + 0.2427000034404966, + 0.2815004375801934, + 0.2748319195941523 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 279577, + "evidence_nonempty_patches": 28, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.4329726063476451, + "restart_teacher_dice": 0.5427570406835327, + "weights": [ + 0.0015948525493022832, + 0.04043639495708322, + 0.0280150797666998, + 0.007509823162345696, + 0.9174652491168103, + 0.0006541777664380059, + 0.001579990027762249, + 0.002744432653558335 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.543427983950943, + "uniform": 0.42744369108351205, + "top1": 0.5427570406835327, + "top2": 0.5446170820822631, + "top3": 0.5446748258576916 + }, + "val_dice": 0.7295256186222652, + "val_vi": 0.5884044823108097, + "val_ece": 0.019701315203793958 + }, + { + "round": 2, + "n_labeled": 56, + "selection_seconds": 18.1, + "round_seconds": 60.7, + "teacher_dice_labeled": [ + 0.3208247601267692, + 0.41290094315652537, + 0.2040200567361768, + 0.2958021771680537, + 0.41646418759010906, + 0.1794556776225178, + 0.20312701542316977, + 0.2733950553226702 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 566444, + "evidence_nonempty_patches": 56, + "first_correction": false, + "restart": "", + "weights": [ + 0.003783436799557813, + 0.10421618553102502, + 0.0, + 0.0, + 0.8920003776694172, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.4129965036621742, + "uniform": 0.30746055861849897, + "top1": 0.41646418759010906, + "top2": 0.4222413401292082, + "top3": 0.4224928778934572 + }, + "pseudo_student_dice": 0.6110748083498765, + "pseudo_ensemble_dice": 0.23453104616733697, + "pseudo_source": "student", + "consistency_scale": 0.37, + "val_dice": 0.7432631786091862, + "val_vi": 0.5466669819950293, + "val_ece": 0.01494093341048987 + }, + { + "round": 3, + "n_labeled": 84, + "selection_seconds": 18.0, + "round_seconds": 373.7, + "teacher_dice_labeled": [ + 0.36273121427086064, + 0.4372980514434087, + 0.21105590198097918, + 0.3393942140665256, + 0.540192314721962, + 0.17578747616669918, + 0.22484020938853538, + 0.29315816733536 + ], + "teacher_dice_pool": [ + 0.39034661919349994, + 0.4761150428010431, + 0.23573876902414095, + 0.38585503964914447, + 0.659109861832459, + 0.18712604425893262, + 0.2631012063047301, + 0.3108627311160762 + ], + "evidence_fg_pixels": 865230, + "evidence_nonempty_patches": 84, + "first_correction": false, + "restart": "", + "weights": [ + 0.0017190749968959006, + 0.03974672870367697, + 0.02820614601955311, + 0.007089510065337752, + 0.9185871992481591, + 0.0003441657073025338, + 0.0009209823989749294, + 0.0033861928600996084 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5393467185906318, + "uniform": 0.3655772808350024, + "top1": 0.540192314721962, + "top2": 0.541732206975406, + "top3": 0.541822389111206 + }, + "pseudo_student_dice": 0.6251532581237091, + "pseudo_ensemble_dice": 0.4286494839036211, + "pseudo_source": "student", + "consistency_scale": 0.417, + "val_dice": 0.7527323068124526, + "val_vi": 0.541985946096611, + "val_ece": 0.01677360838441982 + }, + { + "round": 4, + "n_labeled": 112, + "selection_seconds": 17.8, + "round_seconds": 384.6, + "teacher_dice_labeled": [ + 0.3749811196798546, + 0.44113104576281453, + 0.20960975570483995, + 0.3387416433056273, + 0.5495413048885877, + 0.17535012944458866, + 0.23520326785288623, + 0.2988056113727459 + ], + "teacher_dice_pool": [ + 0.40815563458449466, + 0.4890113030690118, + 0.23786569305882985, + 0.3857946049443244, + 0.6569114664585771, + 0.1934436301469982, + 0.2718589618428649, + 0.3221797503419022 + ], + "evidence_fg_pixels": 1141567, + "evidence_nonempty_patches": 112, + "first_correction": false, + "restart": "", + "weights": [ + 0.001779979044505536, + 0.037832694697906344, + 0.02546728326344037, + 0.006452106793159957, + 0.9239361080593134, + 0.00031389733227876727, + 0.0009357868195762976, + 0.003282143989819337 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5488637066017804, + "uniform": 0.37210821529489313, + "top1": 0.5495413048885877, + "top2": 0.5509644238645611, + "top3": 0.5510551274892082 + }, + "pseudo_student_dice": 0.6628713841310659, + "pseudo_ensemble_dice": 0.49864015261866945, + "pseudo_source": "student", + "consistency_scale": 0.543, + "val_dice": 0.7509571821136031, + "val_vi": 0.5315801650431149, + "val_ece": 0.01703855312358479 + }, + { + "round": 5, + "n_labeled": 140, + "selection_seconds": 17.7, + "round_seconds": 397.3, + "teacher_dice_labeled": [ + 0.3642759867907599, + 0.43358117185195766, + 0.21068374327687786, + 0.33467724413782185, + 0.5535326066210269, + 0.17798854274162795, + 0.24864286910367464, + 0.2906701335016776 + ], + "teacher_dice_pool": [ + 0.38218506328160307, + 0.4682752407941175, + 0.23816118972250283, + 0.3750151946437794, + 0.6498135108202829, + 0.19764850889095936, + 0.2903312054949055, + 0.30404861986044174 + ], + "evidence_fg_pixels": 1422513, + "evidence_nonempty_patches": 140, + "first_correction": false, + "restart": "", + "weights": [ + 0.001545796995137486, + 0.03390845176712642, + 0.02488148279719892, + 0.005987967802871813, + 0.9294056619302986, + 0.0003115125004738827, + 0.0010346030187799481, + 0.002924523188112882 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5528844985001343, + "uniform": 0.3808982550980504, + "top1": 0.5535326066210269, + "top2": 0.5547303978453834, + "top3": 0.554809582158575 + }, + "pseudo_student_dice": 0.6609389649034607, + "pseudo_ensemble_dice": 0.5110580295086186, + "pseudo_source": "student", + "consistency_scale": 0.536, + "val_dice": 0.7545000122039486, + "val_vi": 0.533773990628012, + "val_ece": 0.017922561350632537 + }, + { + "round": 6, + "n_labeled": 168, + "selection_seconds": 17.5, + "round_seconds": 408.0, + "teacher_dice_labeled": [ + 0.3679188297764999, + 0.4335570192451118, + 0.20895861407370306, + 0.33643354064960024, + 0.5531799194234047, + 0.17692680619249224, + 0.2365003677000348, + 0.2917085664928225 + ], + "teacher_dice_pool": [ + 0.394125303581231, + 0.47494787946458666, + 0.24170386806015567, + 0.3806315415340604, + 0.6459687009021247, + 0.20181318421776986, + 0.2769592359329296, + 0.3104454867226209 + ], + "evidence_fg_pixels": 1683796, + "evidence_nonempty_patches": 168, + "first_correction": false, + "restart": "", + "weights": [ + 0.0016089924285309985, + 0.034023888217738066, + 0.02454511057925963, + 0.00611628664709167, + 0.9295109032521134, + 0.0003093465602107464, + 0.0009196454009503332, + 0.0029658269141052424 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5524438317045043, + "uniform": 0.37409291675858214, + "top1": 0.5531799194234047, + "top2": 0.5542681111280099, + "top3": 0.5543604600633237 + }, + "pseudo_student_dice": 0.642503799192999, + "pseudo_ensemble_dice": 0.47485483653762367, + "pseudo_source": "student", + "consistency_scale": 0.475, + "val_dice": 0.7556781192564479, + "val_vi": 0.5299056030507209, + "val_ece": 0.01800360725531013 + }, + { + "round": 7, + "n_labeled": 168, + "selection_seconds": 0.0, + "round_seconds": 115.3, + "final_consolidation": true, + "val_dice": 0.7553752308778334, + "val_vi": 0.5248065360487096, + "val_ece": 0.023144835746207494 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.3088075353000357, + "auroc_top20_error": 0.7218027777541671 + }, + "al_seconds": 2309.8, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.0016089924285309985, + 0.034023888217738066, + 0.02454511057925963, + 0.00611628664709167, + 0.9295109032521134, + 0.0003093465602107464, + 0.0009196454009503332, + 0.0029658269141052424 + ], + "weight_stats_final": { + "max_weight": 0.9295109032521134, + "entropy": 0.3416784214280678, + "effective_sources": 1.4073076647504048, + "min_weight": 0.0003093465602107464 + }, + "top_source_final": "microns_basil", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.7490979386466647, + "patch_iou": 0.6085798192529274, + "patch_precision": 0.7248434080358238, + "patch_recall": 0.7785118795775513, + "patch_vi": 0.5298728975152815, + "n_patches": 735, + "cal_ece": 0.020920834522142353, + "cal_nll": 0.1941572906083464, + "cal_brier": 0.060966210654075155, + "cal_mean_confidence": 0.9338938362228226, + "cal_accuracy": 0.9129730017006803, + "cal_predicted_prevalence": 0.2180125956632653, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.21629047139314697 + }, + "test_notta": { + "patch_dice": 0.7490979386466647, + "patch_iou": 0.6085798192529274, + "patch_precision": 0.7248434080358238, + "patch_recall": 0.7785118795775513, + "patch_vi": 0.5298728975152815, + "n_patches": 735, + "cal_ece": 0.020920834522142353, + "cal_nll": 0.1941572906083464, + "cal_brier": 0.060966210654075155, + "cal_mean_confidence": 0.9338938362228226, + "cal_accuracy": 0.9129730017006803, + "cal_predicted_prevalence": 0.2180125956632653, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.21629047139314697 + }, + "test": { + "patch_dice": 0.7605704968659796, + "patch_iou": 0.6231444504129582, + "patch_precision": 0.7389570238203815, + "patch_recall": 0.7866511074140025, + "patch_vi": 0.5100490645122637, + "n_patches": 735, + "image_dice": 0.8081044889984601, + "image_iou": 0.6784548862789703, + "image_precision": 0.790270861952727, + "image_recall": 0.8272684535945679, + "image_vi": 0.516524529058358, + "n_images": 15, + "cal_ece": 0.010442009413797379, + "cal_nll": 0.17936422828495757, + "cal_brier": 0.05674318918072776, + "cal_mean_confidence": 0.9284094307573348, + "cal_accuracy": 0.9179674213435374, + "cal_predicted_prevalence": 0.21568877551020407, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.21565880639981094 + }, + "eval_seconds": 38.8, + "total_seconds": 3067.8, + "resources": { + "phases": { + "setup": { + "seconds": 1.48, + "peak_cpu_rss_mb": 2192.2, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 63.24, + "peak_cpu_rss_mb": 4183.8, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 649.47, + "peak_cpu_rss_mb": 5161.3, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4638.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2309.76, + "peak_cpu_rss_mb": 5161.3, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4646.0, + "rounds": 6 + }, + "inference": { + "seconds": 38.77, + "peak_cpu_rss_mb": 5161.3, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4646.0, + "patches": 735, + "images": 15, + "patches_per_second": 19.0, + "seconds_per_image": 2.585 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5161.3, + "total_seconds": 3062.7 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/snemi3d/macs-v11-norestart-acq-margin_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-margin/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..a84eecc4caf275e8c060afd0ce75aaa7be235677 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/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 +1,28,100.1,540.4,"[0.35791764681874827, 0.4417052610943664, 0.21306314427094966, 0.3478407216013448, 0.5427570406835327, 0.2427000034404966, 0.2815004375801934, 0.2748319195941523]",,279577,28,True,kept,0.4329726063476451,0.5427570406835327,"[0.0015948525493022832, 0.04043639495708322, 0.0280150797666998, 0.007509823162345696, 0.9174652491168103, 0.0006541777664380059, 0.001579990027762249, 0.002744432653558335]",mixture,"{'mixture': 0.543427983950943, 'uniform': 0.42744369108351205, 'top1': 0.5427570406835327, 'top2': 0.5446170820822631, 'top3': 0.5446748258576916}",0.7295256186222652,0.5884044823108097,0.019701315203793958,,,,, +2,56,18.1,60.7,"[0.3208247601267692, 0.41290094315652537, 0.2040200567361768, 0.2958021771680537, 0.41646418759010906, 0.1794556776225178, 0.20312701542316977, 0.2733950553226702]",,566444,56,False,,,,"[0.003783436799557813, 0.10421618553102502, 0.0, 0.0, 0.8920003776694172, 0.0, 0.0, 0.0]",top3,"{'mixture': 0.4129965036621742, 'uniform': 0.30746055861849897, 'top1': 0.41646418759010906, 'top2': 0.4222413401292082, 'top3': 0.4224928778934572}",0.7432631786091862,0.5466669819950293,0.01494093341048987,0.6110748083498765,0.23453104616733697,student,0.37, +3,84,18.0,373.7,"[0.36273121427086064, 0.4372980514434087, 0.21105590198097918, 0.3393942140665256, 0.540192314721962, 0.17578747616669918, 0.22484020938853538, 0.29315816733536]","[0.39034661919349994, 0.4761150428010431, 0.23573876902414095, 0.38585503964914447, 0.659109861832459, 0.18712604425893262, 0.2631012063047301, 0.3108627311160762]",865230,84,False,,,,"[0.0017190749968959006, 0.03974672870367697, 0.02820614601955311, 0.007089510065337752, 0.9185871992481591, 0.0003441657073025338, 0.0009209823989749294, 0.0033861928600996084]",mixture,"{'mixture': 0.5393467185906318, 'uniform': 0.3655772808350024, 'top1': 0.540192314721962, 'top2': 0.541732206975406, 'top3': 0.541822389111206}",0.7527323068124526,0.541985946096611,0.01677360838441982,0.6251532581237091,0.4286494839036211,student,0.417, +4,112,17.8,384.6,"[0.3749811196798546, 0.44113104576281453, 0.20960975570483995, 0.3387416433056273, 0.5495413048885877, 0.17535012944458866, 0.23520326785288623, 0.2988056113727459]","[0.40815563458449466, 0.4890113030690118, 0.23786569305882985, 0.3857946049443244, 0.6569114664585771, 0.1934436301469982, 0.2718589618428649, 0.3221797503419022]",1141567,112,False,,,,"[0.001779979044505536, 0.037832694697906344, 0.02546728326344037, 0.006452106793159957, 0.9239361080593134, 0.00031389733227876727, 0.0009357868195762976, 0.003282143989819337]",mixture,"{'mixture': 0.5488637066017804, 'uniform': 0.37210821529489313, 'top1': 0.5495413048885877, 'top2': 0.5509644238645611, 'top3': 0.5510551274892082}",0.7509571821136031,0.5315801650431149,0.01703855312358479,0.6628713841310659,0.49864015261866945,student,0.543, +5,140,17.7,397.3,"[0.3642759867907599, 0.43358117185195766, 0.21068374327687786, 0.33467724413782185, 0.5535326066210269, 0.17798854274162795, 0.24864286910367464, 0.2906701335016776]","[0.38218506328160307, 0.4682752407941175, 0.23816118972250283, 0.3750151946437794, 0.6498135108202829, 0.19764850889095936, 0.2903312054949055, 0.30404861986044174]",1422513,140,False,,,,"[0.001545796995137486, 0.03390845176712642, 0.02488148279719892, 0.005987967802871813, 0.9294056619302986, 0.0003115125004738827, 0.0010346030187799481, 0.002924523188112882]",mixture,"{'mixture': 0.5528844985001343, 'uniform': 0.3808982550980504, 'top1': 0.5535326066210269, 'top2': 0.5547303978453834, 'top3': 0.554809582158575}",0.7545000122039486,0.533773990628012,0.017922561350632537,0.6609389649034607,0.5110580295086186,student,0.536, +6,168,17.5,408.0,"[0.3679188297764999, 0.4335570192451118, 0.20895861407370306, 0.33643354064960024, 0.5531799194234047, 0.17692680619249224, 0.2365003677000348, 0.2917085664928225]","[0.394125303581231, 0.47494787946458666, 0.24170386806015567, 0.3806315415340604, 0.6459687009021247, 0.20181318421776986, 0.2769592359329296, 0.3104454867226209]",1683796,168,False,,,,"[0.0016089924285309985, 0.034023888217738066, 0.02454511057925963, 0.00611628664709167, 0.9295109032521134, 0.0003093465602107464, 0.0009196454009503332, 0.0029658269141052424]",mixture,"{'mixture': 0.5524438317045043, 'uniform': 0.37409291675858214, 'top1': 0.5531799194234047, 'top2': 0.5542681111280099, 'top3': 0.5543604600633237}",0.7556781192564479,0.5299056030507209,0.01800360725531013,0.642503799192999,0.47485483653762367,student,0.475, +7,168,0.0,115.3,,,,,,,,,,,,0.7553752308778334,0.5248065360487096,0.023144835746207494,,,,,True diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/log.txt b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..f924eeef6615f30723b65267b9de1db76821b6fd --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/log.txt @@ -0,0 +1,103 @@ +[01:53:44][ 0s][START] target=snemi3d variant=macs-v11-norestart-acq-mcdropout budget=5.0% seed=1 | {"task": "membrane", "target": "snemi3d", "variant": "macs-v11-norestart-acq-mcdropout", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[01:53:44][ 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. +[01:53:44][ 1s][DATA] snemi3d: train=3430 val=735 test=735 patches +[01:53:50][ 6s][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=4.07 +[01:53:51][ 7s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 1s +[01:53:57][ 13s][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 +[01:53:57][ 13s][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 +[01:54:51][ 67s][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 +[01:54:51][ 67s][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 +[01:54:51][ 67s][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 +[01:54:51][ 67s][INIT] student initialised from teacher microns (weight 0.477) +[01:54:56][ 73s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[01:54:56][ 73s][KD] distilling the mixture ensemble (prob space) into the student +[01:57:16][ 213s][KD] soft labels for 3430 patches from 8 teachers in 140s +[01:58:59][ 315s][KD] epoch 1/5 | patches=3430 | loss=0.4461 | 102s +[02:00:41][ 417s][KD] epoch 2/5 | patches=3430 | loss=0.3728 | 102s +[02:02:23][ 519s][KD] epoch 3/5 | patches=3430 | loss=0.3563 | 102s +[02:04:05][ 621s][KD] epoch 4/5 | patches=3430 | loss=0.3563 | 102s +[02:05:46][ 723s][KD] epoch 5/5 | patches=3430 | loss=0.3474 | 102s +[02:05:51][ 727s][ROUND0] before any target label: val_dice=0.3756 val_vi=0.6391 +[02:05:51][ 727s][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 +[02:06:11][ 748s][AL] round-1 uncertainty quality on 3430 patches: {'spearman_uncertainty_error': 0.4788310711035088, 'auroc_top20_error': 0.8198074888014347} +[02:07:33][ 830s][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 +[02:07:38][ 835s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.543, uniform=0.427, top1=0.543, top2=0.545, top3=0.545 -> using mixture +[02:07:39][ 835s][WEIGHTS] round 1: student kept (labeled Dice 0.418 vs best teacher microns 0.543); distilling the corrected ensemble into it +[02:07:39][ 835s][WEIGHTS] round 1: teacher Dice on 28 labeled patches + 0 blank (279577 fg px): c-elegans=0.358, cerebellum=0.442, flywire=0.213, berghia=0.348, microns=0.543, oct-glia=0.243, oct-sfl=0.282, mouse-brain=0.275 | weights: c-elegans=0.002, cerebellum=0.040, flywire=0.028, berghia=0.008, microns=0.917, oct-glia=0.001, oct-sfl=0.002, mouse-brain=0.003 +[02:09:58][ 974s][KD] soft labels for 3402 patches from 8 teachers in 139s +[02:11:39][ 1076s][KD] epoch 1/2 | patches=3402 | loss=0.2532 | 101s +[02:13:21][ 1177s][KD] epoch 2/2 | patches=3402 | loss=0.2504 | 101s +[02:14:43][ 1259s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 82s +[02:14:45][ 1261s][FT] round 1/6 epoch 1/5 | patches=28 | loss=1.0414 +[02:14:48][ 1264s][FT] round 1/6 epoch 2/5 | patches=28 | loss=0.9800 +[02:14:50][ 1266s][FT] round 1/6 epoch 3/5 | patches=28 | loss=0.9003 +[02:14:52][ 1269s][FT] round 1/6 epoch 4/5 | patches=28 | loss=0.8646 +[02:14:55][ 1271s][FT] round 1/6 epoch 5/5 | patches=28 | loss=0.8398 +[02:14:59][ 1275s][AL] round 1/6 | labeled=28 | selection_seconds=102.6000 round_seconds=544.0000 restart_student_dice=0.4183 restart_teacher_dice=0.5428 val_dice=0.7278 val_vi=0.5919 val_ece=0.0210 +[02:15:29][ 1305s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.422, uniform=0.320, top1=0.424, top2=0.430, top3=0.430 -> using top3 +[02:15:29][ 1305s][WEIGHTS] round 2: teacher Dice on 56 labeled patches + 0 blank (542568 fg px): c-elegans=0.312, cerebellum=0.412, flywire=0.206, berghia=0.270, microns=0.424, oct-glia=0.186, oct-sfl=0.216, mouse-brain=0.263 | weights: c-elegans=0.003, cerebellum=0.097, flywire=0.000, berghia=0.000, microns=0.900, oct-glia=0.000, oct-sfl=0.000, mouse-brain=0.000 +[02:15:42][ 1318s][CONS] round 2: on the 28 new patches student Dice=0.595 vs ensemble 0.236 -> pseudo-labels from the student | consistency x0.32 +[02:15:46][ 1322s][FT] round 2/6 epoch 1/5 | patches=56 | loss=0.8159 +[02:15:50][ 1326s][FT] round 2/6 epoch 2/5 | patches=56 | loss=0.7590 +[02:15:54][ 1330s][FT] round 2/6 epoch 3/5 | patches=56 | loss=0.7268 +[02:15:58][ 1334s][FT] round 2/6 epoch 4/5 | patches=56 | loss=0.7150 +[02:16:01][ 1338s][FT] round 2/6 epoch 5/5 | patches=56 | loss=0.7119 +[02:16:06][ 1342s][AL] round 2/6 | labeled=56 | selection_seconds=20.5000 round_seconds=62.5000 pseudo_student_dice=0.5949 pseudo_ensemble_dice=0.2362 consistency_scale=0.3160 val_dice=0.7417 val_vi=0.5611 val_ece=0.0185 +[02:19:08][ 1524s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.412, cerebellum=0.490, flywire=0.240, berghia=0.396, microns=0.664, oct-glia=0.193, oct-sfl=0.265, mouse-brain=0.328 | blended with labeled Dice at 0.50 +[02:21:40][ 1676s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.534, uniform=0.357, top1=0.536, top2=0.537, top3=0.537 -> using mixture +[02:21:40][ 1676s][WEIGHTS] round 3: teacher Dice on 84 labeled patches + 0 blank (821274 fg px): c-elegans=0.361, cerebellum=0.441, flywire=0.214, berghia=0.334, microns=0.536, oct-glia=0.174, oct-sfl=0.224, mouse-brain=0.294 | weights: c-elegans=0.002, cerebellum=0.043, flywire=0.030, berghia=0.007, microns=0.913, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.004 +[02:21:53][ 1689s][CONS] round 3: on the 28 new patches student Dice=0.647 vs ensemble 0.369 -> pseudo-labels from the student | consistency x0.49 +[02:21:59][ 1695s][FT] round 3/6 epoch 1/5 | patches=84 | loss=0.7596 +[02:22:05][ 1701s][FT] round 3/6 epoch 2/5 | patches=84 | loss=0.7431 +[02:22:10][ 1706s][FT] round 3/6 epoch 3/5 | patches=84 | loss=0.7259 +[02:22:16][ 1712s][FT] round 3/6 epoch 4/5 | patches=84 | loss=0.6955 +[02:22:22][ 1718s][FT] round 3/6 epoch 5/5 | patches=84 | loss=0.6926 +[02:22:26][ 1722s][AL] round 3/6 | labeled=84 | selection_seconds=20.2000 round_seconds=375.8000 pseudo_student_dice=0.6475 pseudo_ensemble_dice=0.3694 consistency_scale=0.4920 val_dice=0.7496 val_vi=0.5421 val_ece=0.0175 +[02:25:30][ 1906s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.391, cerebellum=0.475, flywire=0.235, berghia=0.386, microns=0.652, oct-glia=0.185, oct-sfl=0.267, mouse-brain=0.310 | blended with labeled Dice at 0.50 +[02:28:03][ 2060s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.543, uniform=0.359, top1=0.544, top2=0.545, top3=0.545 -> using mixture +[02:28:03][ 2060s][WEIGHTS] round 4: teacher Dice on 112 labeled patches + 0 blank (1131999 fg px): c-elegans=0.358, cerebellum=0.430, flywire=0.206, berghia=0.329, microns=0.544, oct-glia=0.165, oct-sfl=0.231, mouse-brain=0.290 | weights: c-elegans=0.002, cerebellum=0.036, flywire=0.026, berghia=0.006, microns=0.926, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[02:28:17][ 2073s][CONS] round 4: on the 28 new patches student Dice=0.661 vs ensemble 0.499 -> pseudo-labels from the student | consistency x0.54 +[02:28:24][ 2080s][FT] round 4/6 epoch 1/5 | patches=112 | loss=0.7447 +[02:28:31][ 2088s][FT] round 4/6 epoch 2/5 | patches=112 | loss=0.7579 +[02:28:39][ 2095s][FT] round 4/6 epoch 3/5 | patches=112 | loss=0.7504 +[02:28:46][ 2102s][FT] round 4/6 epoch 4/5 | patches=112 | loss=0.7336 +[02:28:53][ 2109s][FT] round 4/6 epoch 5/5 | patches=112 | loss=0.7056 +[02:28:57][ 2113s][AL] round 4/6 | labeled=112 | selection_seconds=20.2000 round_seconds=387.2000 pseudo_student_dice=0.6605 pseudo_ensemble_dice=0.4986 consistency_scale=0.5350 val_dice=0.7432 val_vi=0.5390 val_ece=0.0128 +[02:32:03][ 2299s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.367, cerebellum=0.452, flywire=0.235, berghia=0.373, microns=0.640, oct-glia=0.192, oct-sfl=0.282, mouse-brain=0.291 | blended with labeled Dice at 0.50 +[02:34:39][ 2455s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.549, uniform=0.365, top1=0.550, top2=0.551, top3=0.551 -> using mixture +[02:34:39][ 2455s][WEIGHTS] round 5: teacher Dice on 140 labeled patches + 0 blank (1433291 fg px): c-elegans=0.352, cerebellum=0.422, flywire=0.208, berghia=0.323, microns=0.550, oct-glia=0.166, oct-sfl=0.241, mouse-brain=0.283 | weights: c-elegans=0.001, cerebellum=0.031, flywire=0.025, berghia=0.006, microns=0.933, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[02:34:52][ 2468s][CONS] round 5: on the 28 new patches student Dice=0.692 vs ensemble 0.540 -> pseudo-labels from the student | consistency x0.64 +[02:35:01][ 2477s][FT] round 5/6 epoch 1/5 | patches=140 | loss=0.7161 +[02:35:10][ 2486s][FT] round 5/6 epoch 2/5 | patches=140 | loss=0.7050 +[02:35:19][ 2495s][FT] round 5/6 epoch 3/5 | patches=140 | loss=0.6953 +[02:35:28][ 2504s][FT] round 5/6 epoch 4/5 | patches=140 | loss=0.6930 +[02:35:37][ 2513s][FT] round 5/6 epoch 5/5 | patches=140 | loss=0.6856 +[02:35:41][ 2517s][AL] round 5/6 | labeled=140 | selection_seconds=19.9000 round_seconds=399.7000 pseudo_student_dice=0.6920 pseudo_ensemble_dice=0.5397 consistency_scale=0.6400 val_dice=0.7580 val_vi=0.5310 val_ece=0.0156 +[02:38:49][ 2705s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.395, cerebellum=0.477, flywire=0.236, berghia=0.376, microns=0.639, oct-glia=0.197, oct-sfl=0.282, mouse-brain=0.316 | blended with labeled Dice at 0.50 +[02:41:27][ 2863s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.558, uniform=0.364, top1=0.559, top2=0.560, top3=0.560 -> using mixture +[02:41:27][ 2863s][WEIGHTS] round 6: teacher Dice on 168 labeled patches + 0 blank (1730353 fg px): c-elegans=0.369, cerebellum=0.435, flywire=0.207, berghia=0.325, microns=0.559, oct-glia=0.162, oct-sfl=0.245, mouse-brain=0.297 | weights: c-elegans=0.002, cerebellum=0.033, flywire=0.023, berghia=0.005, microns=0.933, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[02:41:40][ 2876s][CONS] round 6: on the 28 new patches student Dice=0.684 vs ensemble 0.552 -> pseudo-labels from the student | consistency x0.61 +[02:41:51][ 2887s][FT] round 6/6 epoch 1/5 | patches=168 | loss=0.6902 +[02:42:01][ 2897s][FT] round 6/6 epoch 2/5 | patches=168 | loss=0.6889 +[02:42:12][ 2908s][FT] round 6/6 epoch 3/5 | patches=168 | loss=0.6774 +[02:42:22][ 2918s][FT] round 6/6 epoch 4/5 | patches=168 | loss=0.6739 +[02:42:32][ 2929s][FT] round 6/6 epoch 5/5 | patches=168 | loss=0.6641 +[02:42:37][ 2933s][AL] round 6/6 | labeled=168 | selection_seconds=19.9000 round_seconds=411.2000 pseudo_student_dice=0.6845 pseudo_ensemble_dice=0.5515 consistency_scale=0.6150 val_dice=0.7619 val_vi=0.5293 val_ece=0.0177 +[02:42:47][ 2943s][CONS] final consolidation: pseudo-labels refreshed from the student +[02:42:57][ 2953s][FT] final 1/10 epoch 1/1 | patches=168 | loss=0.6622 +[02:43:08][ 2964s][FT] final 2/10 epoch 1/1 | patches=168 | loss=0.6623 +[02:43:18][ 2974s][FT] final 3/10 epoch 1/1 | patches=168 | loss=0.6467 +[02:43:29][ 2985s][FT] final 4/10 epoch 1/1 | patches=168 | loss=0.6459 +[02:43:39][ 2995s][FT] final 5/10 epoch 1/1 | patches=168 | loss=0.6379 +[02:43:50][ 3006s][FT] final 6/10 epoch 1/1 | patches=168 | loss=0.6291 +[02:44:00][ 3016s][FT] final 7/10 epoch 1/1 | patches=168 | loss=0.6209 +[02:44:11][ 3027s][FT] final 8/10 epoch 1/1 | patches=168 | loss=0.6170 +[02:44:21][ 3037s][FT] final 9/10 epoch 1/1 | patches=168 | loss=0.6202 +[02:44:32][ 3048s][FT] final 10/10 epoch 1/1 | patches=168 | loss=0.6203 +[02:44:36][ 3052s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=114.9000 val_dice=0.7586 val_vi=0.5257 val_ece=0.0234 +[02:44:36][ 3052s][WEIGHTS] final: c-elegans=0.002, cerebellum=0.033, flywire=0.023, berghia=0.005, microns=0.933, oct-glia=0.000, oct-sfl=0.001, mouse-brain=0.003 +[02:45:11][ 3087s][TEST] raw student (no EMA, no TTA): patch_dice=0.7514 patch_vi=0.5301 +[02:45:15][ 3092s][TEST] final model without TTA: patch_dice=0.7514 patch_vi=0.5301 +[02:45:15][ 3092s][RESOURCES] {"phases": {"setup": {"seconds": 4.3, "peak_cpu_rss_mb": 2193.5, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 66.88, "peak_cpu_rss_mb": 4186.1, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 650.11, "peak_cpu_rss_mb": 5163.3, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4638.0, "kd_epochs": 5}, "active_learning": {"seconds": 2325.13, "peak_cpu_rss_mb": 5163.3, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4646.0, "rounds": 6}, "inference": {"seconds": 39.44, "peak_cpu_rss_mb": 5163.3, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4646.0, "patches": 735, "images": 15, "patches_per_second": 18.6, "seconds_per_image": 2.629}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5163.3, "total_seconds": 3085.9} +[02:45:15][ 3092s][TEST] patch_dice=0.7617 patch_iou=0.6247 patch_precision=0.7320 patch_recall=0.7970 patch_vi=0.5112 image_dice=0.8086 image_iou=0.6791 image_precision=0.7826 image_recall=0.8367 image_vi=0.5191 cal_ece=0.0115 cal_nll=0.1807 cal_brier=0.0572 cal_mean_confidence=0.9288 cal_accuracy=0.9173 cal_predicted_prevalence=0.2214 cal_true_prevalence=0.2059 cal_mean_probability=0.2207 +[02:45:16][ 3092s][DONE] snemi3d macs-v11-norestart-acq-mcdropout b5 s1 | test patch Dice=0.7617 VI=0.5112 | image Dice=0.8086 VI=0.5191 | 3092s diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/result.json b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..18009942002a675507b8bc6641d9263ece85b9d4 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/result.json @@ -0,0 +1,785 @@ +{ + "target": "snemi3d", + "variant": "macs-v11-norestart-acq-mcdropout", + "budget": 5.0, + "seed": 1, + "config": { + "task": "membrane", + "target": "snemi3d", + "variant": "macs-v11-norestart-acq-mcdropout", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "snemi3d", + "--variants", + "macs-v11-norestart-acq-mcdropout", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.005265163985947525, + 0.05775356661687405, + 0.3937117170688725, + 0.02742108315870487, + 0.4770168284742763, + 0.006835511403011292, + 0.011200087973044965, + 0.020796041319268533 + ], + "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": [ + 0.005265163985947525, + 0.05775356661687405, + 0.3937117170688725, + 0.02742108315870487, + 0.4770168284742763, + 0.006835511403011292, + 0.011200087973044965, + 0.020796041319268533 + ], + "per_source": [ + 0.0030099407416697968, + 0.30826237672231105, + 0.030874093815323564, + 0.005310674836551046, + 0.6243418715179362, + 0.02131301441017302, + 0.0048036204646279375, + 0.002084407491407262 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.048363233996627016, + "mmd2_mixture": 0.060122682064390254, + "mmd2_uniform": 0.15285688303099165, + "mmd2_single_source": [ + 0.4522663898122707, + 0.14701298584193184, + 0.12241687119492284, + 0.2726521753211397, + 0.08995170742180114, + 0.3744254300178685, + 0.33957846092746485, + 0.23688038103280662 + ], + "median_squared_distance": 0.18662990120119327, + "gammas": [ + 1.3395495490858758, + 2.6790990981717515, + 5.358198196343503, + 10.716396392687006, + 21.432792785374012 + ], + "max_weight": 0.4770168284742763, + "entropy": 1.1759448067763238, + "effective_sources": 3.24120380896937, + "min_weight": 0.005265163985947525 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.4624401356050084, + 0.12211205990556462, + 0.29128492533416295, + 0.4206953767024899, + 0.07022529260621624, + 0.3185315424880186, + 0.42807307387367177, + 0.489454149021297 + ], + "standardized": [ + 0.9323134642607491, + -1.3822019864377946, + -0.23168506223542185, + 0.6484142024470035, + -1.735075384714815, + -0.04638528568228213, + 0.6985887114356643, + 1.1160313409268967 + ], + "tau": 0.5, + "max_weight": 0.6243418715179362, + "entropy": 0.9300659298676801, + "effective_sources": 2.534676282981126, + "min_weight": 0.002084407491407262 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "microns_basil", + "seconds": 7.4 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.41247125221353, + 0.47321264983034456, + 0.19639403256049165, + 0.31013587127331976, + 0.5894646206275136, + 0.13744302537529526, + 0.3054434345563225, + 0.35012697230796425 + ], + "ensemble_val_dice": { + "mixture": 0.4205061683646676, + "mixture@raw": 0.462918877571213, + "mixture@logit": 0.3665782427199844, + "per_source": 0.5892228765147073, + "per_source@raw": 0.5905348345941427, + "per_source@logit": 0.5781053080039771, + "uniform": 0.41124230184319877, + "uniform@raw": 0.4669572710990906, + "uniform@logit": 0.22782834015245554 + }, + "val_patches": 735 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 2.642308473587036, + 4.074584007263184 + ], + "top_source": "microns_basil", + "weight_stats": { + "max_weight": 0.4770168284742763, + "entropy": 1.1759448067763238, + "effective_sources": 3.24120380896937, + "min_weight": 0.005265163985947525 + }, + "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": 650.1, + "round0_val": { + "val_dice": 0.3755735946634418, + "val_vi": 0.6390912588283757, + "val_ece": 0.07344130447917224 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 28, + "selection_seconds": 102.6, + "round_seconds": 544.0, + "teacher_dice_labeled": [ + 0.35791764681874827, + 0.4417052610943664, + 0.21306314427094966, + 0.3478407216013448, + 0.5427570406835327, + 0.2427000034404966, + 0.2815004375801934, + 0.2748319195941523 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 279577, + "evidence_nonempty_patches": 28, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.41833134967864793, + "restart_teacher_dice": 0.5427570406835327, + "weights": [ + 0.0015948525493022832, + 0.04043639495708322, + 0.0280150797666998, + 0.007509823162345696, + 0.9174652491168103, + 0.0006541777664380059, + 0.001579990027762249, + 0.002744432653558335 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.543427983950943, + "uniform": 0.42744369108351205, + "top1": 0.5427570406835327, + "top2": 0.5446170820822631, + "top3": 0.5446748258576916 + }, + "val_dice": 0.7278339915577575, + "val_vi": 0.5919187624963894, + "val_ece": 0.020966535856338883 + }, + { + "round": 2, + "n_labeled": 56, + "selection_seconds": 20.5, + "round_seconds": 62.5, + "teacher_dice_labeled": [ + 0.31181172477305286, + 0.4123490463356484, + 0.20608301371020069, + 0.27008440705749637, + 0.4244959608987849, + 0.18641832734596342, + 0.21613912931956966, + 0.26276946418977837 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 542568, + "evidence_nonempty_patches": 56, + "first_correction": false, + "restart": "", + "weights": [ + 0.003220035189592252, + 0.09652850169918203, + 0.0, + 0.0, + 0.9002514631112257, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.4215898139097271, + "uniform": 0.3200273826153171, + "top1": 0.4244959608987849, + "top2": 0.43002979418608317, + "top3": 0.4302272878082905 + }, + "pseudo_student_dice": 0.5949284088746447, + "pseudo_ensemble_dice": 0.23617475681690378, + "pseudo_source": "student", + "consistency_scale": 0.316, + "val_dice": 0.7416816901534429, + "val_vi": 0.5611395139148813, + "val_ece": 0.018502803413247136 + }, + { + "round": 3, + "n_labeled": 84, + "selection_seconds": 20.2, + "round_seconds": 375.8, + "teacher_dice_labeled": [ + 0.36133371836338635, + 0.44085686194906737, + 0.2136680721683852, + 0.3340398695594378, + 0.535725225724645, + 0.17392592764506856, + 0.2235620921091399, + 0.2938632487285832 + ], + "teacher_dice_pool": [ + 0.41194979078777727, + 0.49042473385394364, + 0.240436043260484, + 0.3960085279038956, + 0.6642957966312557, + 0.19332109643781695, + 0.2651361815473013, + 0.3280756406570733 + ], + "evidence_fg_pixels": 821274, + "evidence_nonempty_patches": 84, + "first_correction": false, + "restart": "", + "weights": [ + 0.0017627896179064318, + 0.04282841635643022, + 0.030106700948834032, + 0.0069877524086444905, + 0.913471445141928, + 0.0003512836206721712, + 0.0009455302888582306, + 0.003546081616726441 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5344386028800578, + "uniform": 0.35706953542833364, + "top1": 0.535725225724645, + "top2": 0.5373175326220334, + "top3": 0.5374325155195703 + }, + "pseudo_student_dice": 0.647461004157173, + "pseudo_ensemble_dice": 0.3694070799234942, + "pseudo_source": "student", + "consistency_scale": 0.492, + "val_dice": 0.749626075653042, + "val_vi": 0.5421390132818664, + "val_ece": 0.017509178777303697 + }, + { + "round": 4, + "n_labeled": 112, + "selection_seconds": 20.2, + "round_seconds": 387.2, + "teacher_dice_labeled": [ + 0.3576869817133024, + 0.4298228509106173, + 0.20622688258575914, + 0.3286969851255738, + 0.5436391225695805, + 0.16475371617754514, + 0.23146357713103805, + 0.2901076096672102 + ], + "teacher_dice_pool": [ + 0.39138772527845306, + 0.475016594588615, + 0.23501316215553472, + 0.38595151149100176, + 0.6518267028414817, + 0.1851465791692955, + 0.266669253417049, + 0.31029983774319086 + ], + "evidence_fg_pixels": 1131999, + "evidence_nonempty_patches": 112, + "first_correction": false, + "restart": "", + "weights": [ + 0.0015913184560729483, + 0.03590923115993884, + 0.02616620169627435, + 0.006201944108146959, + 0.9256755334459202, + 0.00030006661828817027, + 0.0009580437690906246, + 0.0031976607462679196 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.5425422974902394, + "uniform": 0.35911726422428564, + "top1": 0.5436391225695805, + "top2": 0.5449977421327727, + "top3": 0.5450780283194936 + }, + "pseudo_student_dice": 0.6605078248292389, + "pseudo_ensemble_dice": 0.4986002301881694, + "pseudo_source": "student", + "consistency_scale": 0.535, + "val_dice": 0.743164012252046, + "val_vi": 0.5390013919900811, + "val_ece": 0.012813062188756805 + }, + { + "round": 5, + "n_labeled": 140, + "selection_seconds": 19.9, + "round_seconds": 399.7, + "teacher_dice_labeled": [ + 0.3518329705798253, + 0.4218277901383832, + 0.2075243638051535, + 0.3233491742376384, + 0.5503626085755353, + 0.1655991382860516, + 0.2414640614809438, + 0.2833941818452663 + ], + "teacher_dice_pool": [ + 0.36696379474936547, + 0.45218328641072336, + 0.23535295439566672, + 0.3729089636121604, + 0.6402919292608433, + 0.1915812146114833, + 0.28248284990582206, + 0.29116450464463495 + ], + "evidence_fg_pixels": 1433291, + "evidence_nonempty_patches": 140, + "first_correction": false, + "restart": "", + "weights": [ + 0.0014139596753337835, + 0.03123112082644807, + 0.024973484663534776, + 0.005538678696212871, + 0.9327431865927268, + 0.0002850971507548484, + 0.0009975173861996782, + 0.0028169550087891832 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.549249434136135, + "uniform": 0.3652362373751127, + "top1": 0.5503626085755353, + "top2": 0.5514334084307191, + "top3": 0.5514877157094626 + }, + "pseudo_student_dice": 0.6920095778498158, + "pseudo_ensemble_dice": 0.5397081235622629, + "pseudo_source": "student", + "consistency_scale": 0.64, + "val_dice": 0.7580254046092908, + "val_vi": 0.5310221427433142, + "val_ece": 0.015614619822543528 + }, + { + "round": 6, + "n_labeled": 168, + "selection_seconds": 19.9, + "round_seconds": 411.2, + "teacher_dice_labeled": [ + 0.368547271203964, + 0.43520792181378254, + 0.20708955846823568, + 0.3250520172516318, + 0.5587925243587022, + 0.1624849748096609, + 0.24510524644786275, + 0.2974032896168033 + ], + "teacher_dice_pool": [ + 0.39540179798971437, + 0.476869730677141, + 0.2360398325311521, + 0.3762109930294652, + 0.6388893327362402, + 0.19723726041841333, + 0.28174786281265013, + 0.31583038741705616 + ], + "evidence_fg_pixels": 1730353, + "evidence_nonempty_patches": 168, + "first_correction": false, + "restart": "", + "weights": [ + 0.0015371438732314031, + 0.032838624156987614, + 0.02287067693233035, + 0.005181908431598925, + 0.9333852928032684, + 0.0002541888205423604, + 0.0009515292617998961, + 0.0029806357202409647 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.557801610985773, + "uniform": 0.3639750335187959, + "top1": 0.5587925243587022, + "top2": 0.5598847509339682, + "top3": 0.559953710155138 + }, + "pseudo_student_dice": 0.6844702929054793, + "pseudo_ensemble_dice": 0.5515304120552174, + "pseudo_source": "student", + "consistency_scale": 0.615, + "val_dice": 0.7619157981028477, + "val_vi": 0.5293165228256188, + "val_ece": 0.0177300918492753 + }, + { + "round": 7, + "n_labeled": 168, + "selection_seconds": 0.0, + "round_seconds": 114.9, + "final_consolidation": true, + "val_dice": 0.7586223312419698, + "val_vi": 0.5257261950571073, + "val_ece": 0.02341865030215247 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.4788310711035088, + "auroc_top20_error": 0.8198074888014347 + }, + "al_seconds": 2325.1, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.0015371438732314031, + 0.032838624156987614, + 0.02287067693233035, + 0.005181908431598925, + 0.9333852928032684, + 0.0002541888205423604, + 0.0009515292617998961, + 0.0029806357202409647 + ], + "weight_stats_final": { + "max_weight": 0.9333852928032684, + "entropy": 0.3262158082697485, + "effective_sources": 1.3857143852303648, + "min_weight": 0.0002541888205423604 + }, + "top_source_final": "microns_basil", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.7513668328802934, + "patch_iou": 0.611391144333655, + "patch_precision": 0.7196353193099454, + "patch_recall": 0.7894107777903655, + "patch_vi": 0.5301334119821381, + "n_patches": 735, + "cal_ece": 0.02146211620065379, + "cal_nll": 0.1954418500392805, + "cal_brier": 0.06123293992316221, + "cal_mean_confidence": 0.9340989678162999, + "cal_accuracy": 0.9126368516156462, + "cal_predicted_prevalence": 0.22335246598639455, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.22105475044292994 + }, + "test_notta": { + "patch_dice": 0.7513668328802934, + "patch_iou": 0.611391144333655, + "patch_precision": 0.7196353193099454, + "patch_recall": 0.7894107777903655, + "patch_vi": 0.5301334119821381, + "n_patches": 735, + "cal_ece": 0.02146211620065379, + "cal_nll": 0.1954418500392805, + "cal_brier": 0.06123293992316221, + "cal_mean_confidence": 0.9340989678162999, + "cal_accuracy": 0.9126368516156462, + "cal_predicted_prevalence": 0.22335246598639455, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.22105475044292994 + }, + "test": { + "patch_dice": 0.7617087743524753, + "patch_iou": 0.6246963651828682, + "patch_precision": 0.7320438250707431, + "patch_recall": 0.7970178665759515, + "patch_vi": 0.5112018400107708, + "n_patches": 735, + "image_dice": 0.8085869204312826, + "image_iou": 0.67914010219418, + "image_precision": 0.7826366296591518, + "image_recall": 0.8366991090311863, + "image_vi": 0.5190660044759824, + "n_images": 15, + "cal_ece": 0.011528958416880475, + "cal_nll": 0.18065925569353553, + "cal_brier": 0.057158908950022694, + "cal_mean_confidence": 0.9287868772093976, + "cal_accuracy": 0.9172579187925171, + "cal_predicted_prevalence": 0.22142591411564627, + "cal_true_prevalence": 0.20588727678571428, + "cal_mean_probability": 0.22065802468645127 + }, + "eval_seconds": 39.4, + "total_seconds": 3091.5, + "resources": { + "phases": { + "setup": { + "seconds": 4.3, + "peak_cpu_rss_mb": 2193.5, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 66.88, + "peak_cpu_rss_mb": 4186.1, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 650.11, + "peak_cpu_rss_mb": 5163.3, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4638.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2325.13, + "peak_cpu_rss_mb": 5163.3, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4646.0, + "rounds": 6 + }, + "inference": { + "seconds": 39.44, + "peak_cpu_rss_mb": 5163.3, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4646.0, + "patches": 735, + "images": 15, + "patches_per_second": 18.6, + "seconds_per_image": 2.629 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5163.3, + "total_seconds": 3085.9 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/snemi3d/macs-v11-norestart-acq-mcdropout_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..b8fc81c8e2f9f50d4ea7cb48a6619b420f8e5eb9 --- /dev/null +++ b/macs_v2_r26/runs/membrane/snemi3d/macs-v11-norestart-acq-mcdropout/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 +1,28,102.6,544.0,"[0.35791764681874827, 0.4417052610943664, 0.21306314427094966, 0.3478407216013448, 0.5427570406835327, 0.2427000034404966, 0.2815004375801934, 0.2748319195941523]",,279577,28,True,kept,0.41833134967864793,0.5427570406835327,"[0.0015948525493022832, 0.04043639495708322, 0.0280150797666998, 0.007509823162345696, 0.9174652491168103, 0.0006541777664380059, 0.001579990027762249, 0.002744432653558335]",mixture,"{'mixture': 0.543427983950943, 'uniform': 0.42744369108351205, 'top1': 0.5427570406835327, 'top2': 0.5446170820822631, 'top3': 0.5446748258576916}",0.7278339915577575,0.5919187624963894,0.020966535856338883,,,,, +2,56,20.5,62.5,"[0.31181172477305286, 0.4123490463356484, 0.20608301371020069, 0.27008440705749637, 0.4244959608987849, 0.18641832734596342, 0.21613912931956966, 0.26276946418977837]",,542568,56,False,,,,"[0.003220035189592252, 0.09652850169918203, 0.0, 0.0, 0.9002514631112257, 0.0, 0.0, 0.0]",top3,"{'mixture': 0.4215898139097271, 'uniform': 0.3200273826153171, 'top1': 0.4244959608987849, 'top2': 0.43002979418608317, 'top3': 0.4302272878082905}",0.7416816901534429,0.5611395139148813,0.018502803413247136,0.5949284088746447,0.23617475681690378,student,0.316, +3,84,20.2,375.8,"[0.36133371836338635, 0.44085686194906737, 0.2136680721683852, 0.3340398695594378, 0.535725225724645, 0.17392592764506856, 0.2235620921091399, 0.2938632487285832]","[0.41194979078777727, 0.49042473385394364, 0.240436043260484, 0.3960085279038956, 0.6642957966312557, 0.19332109643781695, 0.2651361815473013, 0.3280756406570733]",821274,84,False,,,,"[0.0017627896179064318, 0.04282841635643022, 0.030106700948834032, 0.0069877524086444905, 0.913471445141928, 0.0003512836206721712, 0.0009455302888582306, 0.003546081616726441]",mixture,"{'mixture': 0.5344386028800578, 'uniform': 0.35706953542833364, 'top1': 0.535725225724645, 'top2': 0.5373175326220334, 'top3': 0.5374325155195703}",0.749626075653042,0.5421390132818664,0.017509178777303697,0.647461004157173,0.3694070799234942,student,0.492, +4,112,20.2,387.2,"[0.3576869817133024, 0.4298228509106173, 0.20622688258575914, 0.3286969851255738, 0.5436391225695805, 0.16475371617754514, 0.23146357713103805, 0.2901076096672102]","[0.39138772527845306, 0.475016594588615, 0.23501316215553472, 0.38595151149100176, 0.6518267028414817, 0.1851465791692955, 0.266669253417049, 0.31029983774319086]",1131999,112,False,,,,"[0.0015913184560729483, 0.03590923115993884, 0.02616620169627435, 0.006201944108146959, 0.9256755334459202, 0.00030006661828817027, 0.0009580437690906246, 0.0031976607462679196]",mixture,"{'mixture': 0.5425422974902394, 'uniform': 0.35911726422428564, 'top1': 0.5436391225695805, 'top2': 0.5449977421327727, 'top3': 0.5450780283194936}",0.743164012252046,0.5390013919900811,0.012813062188756805,0.6605078248292389,0.4986002301881694,student,0.535, +5,140,19.9,399.7,"[0.3518329705798253, 0.4218277901383832, 0.2075243638051535, 0.3233491742376384, 0.5503626085755353, 0.1655991382860516, 0.2414640614809438, 0.2833941818452663]","[0.36696379474936547, 0.45218328641072336, 0.23535295439566672, 0.3729089636121604, 0.6402919292608433, 0.1915812146114833, 0.28248284990582206, 0.29116450464463495]",1433291,140,False,,,,"[0.0014139596753337835, 0.03123112082644807, 0.024973484663534776, 0.005538678696212871, 0.9327431865927268, 0.0002850971507548484, 0.0009975173861996782, 0.0028169550087891832]",mixture,"{'mixture': 0.549249434136135, 'uniform': 0.3652362373751127, 'top1': 0.5503626085755353, 'top2': 0.5514334084307191, 'top3': 0.5514877157094626}",0.7580254046092908,0.5310221427433142,0.015614619822543528,0.6920095778498158,0.5397081235622629,student,0.64, +6,168,19.9,411.2,"[0.368547271203964, 0.43520792181378254, 0.20708955846823568, 0.3250520172516318, 0.5587925243587022, 0.1624849748096609, 0.24510524644786275, 0.2974032896168033]","[0.39540179798971437, 0.476869730677141, 0.2360398325311521, 0.3762109930294652, 0.6388893327362402, 0.19723726041841333, 0.28174786281265013, 0.31583038741705616]",1730353,168,False,,,,"[0.0015371438732314031, 0.032838624156987614, 0.02287067693233035, 0.005181908431598925, 0.9333852928032684, 0.0002541888205423604, 0.0009515292617998961, 0.0029806357202409647]",mixture,"{'mixture': 0.557801610985773, 'uniform': 0.3639750335187959, 'top1': 0.5587925243587022, 'top2': 0.5598847509339682, 'top3': 0.559953710155138}",0.7619157981028477,0.5293165228256188,0.0177300918492753,0.6844702929054793,0.5515304120552174,student,0.615, +7,168,0.0,114.9,,,,,,,,,,,,0.7586223312419698,0.5257261950571073,0.02341865030215247,,,,,True diff --git a/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/log.txt b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..c5f8cba69b565e105a0eb04b29c2000ff3e63b1a --- /dev/null +++ b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/log.txt @@ -0,0 +1,108 @@ +[11:03:34][ 0s][START] target=whole-mouse-brain variant=macs-v11-norestart-acq-entropy budget=5.0% seed=2 | {"task": "membrane", "target": "whole-mouse-brain", "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": 8, "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} +[11:03:34][ 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. +[11:03:34][ 0s][DATA] whole-mouse-brain: train=5925 val=1250 test=1300 patches +[11:03:36][ 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, snemi=3.55 +[11:03:37][ 3s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[11:03:43][ 9s][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 +[11:03:43][ 9s][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 +[11:04:57][ 83s][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 +[11:04:57][ 83s][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 +[11:04:57][ 83s][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 +[11:04:57][ 83s][INIT] student initialised from teacher microns (weight 0.376) +[11:05:00][ 86s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[11:05:00][ 86s][KD] distilling the mixture ensemble (prob space) into the student +[11:09:02][ 328s][KD] soft labels for 5925 patches from 8 teachers in 242s +[11:11:59][ 505s][KD] epoch 1/5 | patches=5925 | loss=0.6655 | 176s +[11:14:55][ 681s][KD] epoch 2/5 | patches=5925 | loss=0.4946 | 176s +[11:17:51][ 857s][KD] epoch 3/5 | patches=5925 | loss=0.4649 | 176s +[11:20:47][ 1033s][KD] epoch 4/5 | patches=5925 | loss=0.4584 | 176s +[11:23:43][ 1209s][KD] epoch 5/5 | patches=5925 | loss=0.4562 | 176s +[11:23:49][ 1215s][ROUND0] before any target label: val_dice=0.3932 val_vi=0.5071 +[11:23:49][ 1215s][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 +[11:24:21][ 1247s][AL] round-1 uncertainty quality on 5925 patches: {'spearman_uncertainty_error': -0.2505878322752116, 'auroc_top20_error': 0.3924166090497315} +[11:25:43][ 1329s][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 +[11:25:51][ 1337s][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 +[11:25:52][ 1338s][WEIGHTS] round 1: student kept (labeled Dice 0.292 vs best teacher oct-sfl 0.453); distilling the corrected ensemble into it +[11:25:52][ 1338s][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 +[11:26:52][ 1398s][KD] soft labels for 5876 patches from 8 teachers in 61s +[11:29:47][ 1573s][KD] epoch 1/2 | patches=5876 | loss=0.2658 | 175s +[11:32:43][ 1749s][KD] epoch 2/2 | patches=5876 | loss=0.2580 | 175s +[11:33:04][ 1770s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 21s +[11:33:07][ 1773s][FT] round 1/6 epoch 1/5 | patches=49 | loss=1.4617 +[11:33:11][ 1777s][FT] round 1/6 epoch 2/5 | patches=49 | loss=1.2396 +[11:33:15][ 1781s][FT] round 1/6 epoch 3/5 | patches=49 | loss=1.1040 +[11:33:19][ 1785s][FT] round 1/6 epoch 4/5 | patches=49 | loss=0.9489 +[11:33:23][ 1789s][FT] round 1/6 epoch 5/5 | patches=49 | loss=1.1202 +[11:33:28][ 1794s][AL] round 1/6 | labeled=49 | selection_seconds=113.6000 round_seconds=573.4000 restart_student_dice=0.2920 restart_teacher_dice=0.4531 val_dice=0.6300 val_vi=0.5107 val_ece=0.0330 +[11:33:59][ 1825s][AL] round 2: 474/5876 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[11:34:15][ 1841s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.302, uniform=0.301, top1=0.358, top2=0.360, top3=0.325 -> using top2 +[11:34:15][ 1841s][WEIGHTS] round 2: teacher Dice on 98 labeled patches + 0 blank (1155835 fg px): c-elegans=0.325, cerebellum=0.262, flywire=0.116, berghia=0.268, microns=0.190, oct-glia=0.316, oct-sfl=0.358, snemi=0.269 | weights: c-elegans=0.054, cerebellum=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.946, snemi=0.000 +[11:34:30][ 1856s][CONS] round 2: on the 49 new patches student Dice=0.338 vs ensemble 0.125 -> pseudo-labels from the student | consistency x0.00 +[11:34:33][ 1859s][FT] round 2/6 epoch 1/5 | patches=98 | loss=1.0522 +[11:34:37][ 1863s][FT] round 2/6 epoch 2/5 | patches=98 | loss=1.0992 +[11:34:40][ 1866s][FT] round 2/6 epoch 3/5 | patches=98 | loss=0.9450 +[11:34:43][ 1869s][FT] round 2/6 epoch 4/5 | patches=98 | loss=0.9605 +[11:34:47][ 1873s][FT] round 2/6 epoch 5/5 | patches=98 | loss=0.9465 +[11:34:53][ 1879s][AL] round 2/6 | labeled=98 | selection_seconds=31.1000 round_seconds=78.3000 pseudo_student_dice=0.3380 pseudo_ensemble_dice=0.1250 consistency_scale=0.0000 val_dice=0.6671 val_vi=0.4213 val_ece=0.0160 +[11:35:23][ 1909s][AL] round 3: 768/5827 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[11:38:11][ 2077s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.383, cerebellum=0.374, flywire=0.154, berghia=0.264, microns=0.332, oct-glia=0.446, oct-sfl=0.555, snemi=0.293 | blended with labeled Dice at 0.50 +[11:40:48][ 2234s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.365, uniform=0.342, top1=0.415, top2=0.381, top3=0.383 -> using top1 +[11:40:48][ 2234s][WEIGHTS] round 3: teacher Dice on 147 labeled patches + 0 blank (1901791 fg px): c-elegans=0.336, cerebellum=0.330, flywire=0.116, berghia=0.257, microns=0.231, oct-glia=0.343, oct-sfl=0.415, snemi=0.299 | 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 +[11:41:04][ 2250s][CONS] round 3: on the 49 new patches student Dice=0.490 vs ensemble 0.116 -> pseudo-labels from the student | consistency x0.00 +[11:41:08][ 2254s][FT] round 3/6 epoch 1/5 | patches=147 | loss=1.1947 +[11:41:13][ 2259s][FT] round 3/6 epoch 2/5 | patches=147 | loss=1.0835 +[11:41:18][ 2264s][FT] round 3/6 epoch 3/5 | patches=147 | loss=1.1484 +[11:41:23][ 2269s][FT] round 3/6 epoch 4/5 | patches=147 | loss=1.0334 +[11:41:27][ 2273s][FT] round 3/6 epoch 5/5 | patches=147 | loss=0.9493 +[11:41:33][ 2279s][AL] round 3/6 | labeled=147 | selection_seconds=30.8000 round_seconds=394.7000 pseudo_student_dice=0.4902 pseudo_ensemble_dice=0.1158 consistency_scale=0.0000 val_dice=0.6500 val_vi=0.4666 val_ece=0.0266 +[11:42:04][ 2310s][AL] round 4: 544/5778 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[11:44:54][ 2480s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.412, cerebellum=0.380, flywire=0.153, berghia=0.304, microns=0.302, oct-glia=0.433, oct-sfl=0.529, snemi=0.329 | blended with labeled Dice at 0.50 +[11:47:34][ 2640s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.364, uniform=0.322, top1=0.395, top2=0.389, top3=0.386 -> using top1 +[11:47:34][ 2640s][WEIGHTS] round 4: teacher Dice on 196 labeled patches + 0 blank (2568324 fg px): c-elegans=0.352, cerebellum=0.355, flywire=0.110, berghia=0.253, microns=0.219, oct-glia=0.324, oct-sfl=0.395, snemi=0.321 | 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 +[11:47:49][ 2655s][CONS] round 4: on the 49 new patches student Dice=0.536 vs ensemble 0.226 -> pseudo-labels from the student | consistency x0.12 +[11:48:02][ 2668s][FT] round 4/6 epoch 1/5 | patches=196 | loss=1.1507 +[11:48:14][ 2680s][FT] round 4/6 epoch 2/5 | patches=196 | loss=1.0430 +[11:48:27][ 2693s][FT] round 4/6 epoch 3/5 | patches=196 | loss=1.0931 +[11:48:39][ 2705s][FT] round 4/6 epoch 4/5 | patches=196 | loss=1.0568 +[11:48:51][ 2717s][FT] round 4/6 epoch 5/5 | patches=196 | loss=0.9960 +[11:48:57][ 2723s][AL] round 4/6 | labeled=196 | selection_seconds=30.6000 round_seconds=438.1000 pseudo_student_dice=0.5360 pseudo_ensemble_dice=0.2257 consistency_scale=0.1200 val_dice=0.6867 val_vi=0.4335 val_ece=0.0067 +[11:49:27][ 2753s][AL] round 5: 766/5729 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[11:52:21][ 2927s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.388, cerebellum=0.400, flywire=0.151, berghia=0.267, microns=0.312, oct-glia=0.430, oct-sfl=0.528, snemi=0.320 | blended with labeled Dice at 0.50 +[11:55:05][ 3091s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.391, uniform=0.319, top1=0.394, top2=0.406, top3=0.405 -> using top3 +[11:55:05][ 3091s][WEIGHTS] round 5: teacher Dice on 245 labeled patches + 0 blank (3582198 fg px): c-elegans=0.358, cerebellum=0.381, flywire=0.111, berghia=0.244, microns=0.218, oct-glia=0.316, oct-sfl=0.394, snemi=0.332 | weights: c-elegans=0.016, cerebellum=0.689, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.294, snemi=0.000 +[11:55:20][ 3106s][CONS] round 5: on the 49 new patches student Dice=0.611 vs ensemble 0.448 -> pseudo-labels from the student | consistency x0.37 +[11:55:35][ 3121s][FT] round 5/6 epoch 1/5 | patches=245 | loss=1.1906 +[11:55:50][ 3136s][FT] round 5/6 epoch 2/5 | patches=245 | loss=1.1607 +[11:56:05][ 3151s][FT] round 5/6 epoch 3/5 | patches=245 | loss=1.1352 +[11:56:21][ 3167s][FT] round 5/6 epoch 4/5 | patches=245 | loss=1.0207 +[11:56:36][ 3182s][FT] round 5/6 epoch 5/5 | patches=245 | loss=1.0371 +[11:56:42][ 3188s][AL] round 5/6 | labeled=245 | selection_seconds=30.3000 round_seconds=458.7000 pseudo_student_dice=0.6111 pseudo_ensemble_dice=0.4482 consistency_scale=0.3700 val_dice=0.7010 val_vi=0.4148 val_ece=0.0096 +[11:57:12][ 3218s][AL] round 6: 701/5680 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:00:09][ 3395s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.400, cerebellum=0.392, flywire=0.148, berghia=0.280, microns=0.298, oct-glia=0.433, oct-sfl=0.526, snemi=0.334 | blended with labeled Dice at 0.50 +[12:02:57][ 3563s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.409, uniform=0.308, top1=0.393, top2=0.411, top3=0.411 -> using mixture +[12:02:57][ 3563s][WEIGHTS] round 6: teacher Dice on 294 labeled patches + 0 blank (4504610 fg px): c-elegans=0.378, cerebellum=0.393, flywire=0.108, berghia=0.247, microns=0.210, oct-glia=0.315, oct-sfl=0.385, snemi=0.347 | weights: c-elegans=0.013, cerebellum=0.509, flywire=0.004, berghia=0.035, microns=0.120, oct-glia=0.123, oct-sfl=0.176, snemi=0.020 +[12:03:13][ 3579s][CONS] round 6: on the 49 new patches student Dice=0.650 vs ensemble 0.298 -> pseudo-labels from the student | consistency x0.50 +[12:03:30][ 3596s][FT] round 6/6 epoch 1/5 | patches=294 | loss=1.1132 +[12:03:48][ 3614s][FT] round 6/6 epoch 2/5 | patches=294 | loss=1.0773 +[12:04:07][ 3633s][FT] round 6/6 epoch 3/5 | patches=294 | loss=1.1183 +[12:04:25][ 3651s][FT] round 6/6 epoch 4/5 | patches=294 | loss=1.0493 +[12:04:43][ 3669s][FT] round 6/6 epoch 5/5 | patches=294 | loss=0.9931 +[12:04:48][ 3674s][AL] round 6/6 | labeled=294 | selection_seconds=30.1000 round_seconds=481.0000 pseudo_student_dice=0.6501 pseudo_ensemble_dice=0.2981 consistency_scale=0.5000 val_dice=0.7047 val_vi=0.3934 val_ece=0.0079 +[12:04:59][ 3685s][CONS] final consolidation: pseudo-labels refreshed from the student +[12:05:17][ 3703s][FT] final 1/10 epoch 1/1 | patches=294 | loss=0.9656 +[12:05:35][ 3721s][FT] final 2/10 epoch 1/1 | patches=294 | loss=0.9334 +[12:05:53][ 3739s][FT] final 3/10 epoch 1/1 | patches=294 | loss=0.9946 +[12:06:11][ 3757s][FT] final 4/10 epoch 1/1 | patches=294 | loss=0.9482 +[12:06:29][ 3775s][FT] final 5/10 epoch 1/1 | patches=294 | loss=0.9485 +[12:06:47][ 3793s][FT] final 6/10 epoch 1/1 | patches=294 | loss=0.9475 +[12:07:05][ 3811s][FT] final 7/10 epoch 1/1 | patches=294 | loss=0.8823 +[12:07:23][ 3829s][FT] final 8/10 epoch 1/1 | patches=294 | loss=0.8645 +[12:07:41][ 3847s][FT] final 9/10 epoch 1/1 | patches=294 | loss=0.8566 +[12:07:59][ 3865s][FT] final 10/10 epoch 1/1 | patches=294 | loss=0.8279 +[12:08:05][ 3871s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=190.4000 val_dice=0.7156 val_vi=0.3905 val_ece=0.0081 +[12:08:05][ 3871s][WEIGHTS] final: c-elegans=0.013, cerebellum=0.509, flywire=0.004, berghia=0.035, microns=0.120, oct-glia=0.123, oct-sfl=0.176, snemi=0.020 +[12:09:06][ 3932s][TEST] raw student (no EMA, no TTA): patch_dice=0.7055 patch_vi=0.3509 +[12:09:13][ 3939s][TEST] final model without TTA: patch_dice=0.7055 patch_vi=0.3509 +[12:09:13][ 3939s][RESOURCES] {"phases": {"setup": {"seconds": 1.48, "peak_cpu_rss_mb": 2192.2, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 84.15, "peak_cpu_rss_mb": 4183.1, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 1122.94, "peak_cpu_rss_mb": 5940.4, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 2655.43, "peak_cpu_rss_mb": 5940.4, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 68.47, "peak_cpu_rss_mb": 5940.4, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 1300, "images": 52, "patches_per_second": 19.0, "seconds_per_image": 1.317}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5940.4, "total_seconds": 3932.5} +[12:09:13][ 3939s][TEST] patch_dice=0.7241 patch_iou=0.6435 patch_precision=0.7535 patch_recall=0.8265 patch_vi=0.3296 image_dice=0.8312 image_iou=0.7270 image_precision=0.8134 image_recall=0.8774 image_vi=0.3823 cal_ece=0.0033 cal_nll=0.1784 cal_brier=0.0514 cal_mean_confidence=0.9305 cal_accuracy=0.9315 cal_predicted_prevalence=0.2275 cal_true_prevalence=0.2108 cal_mean_probability=0.2430 +[12:09:14][ 3940s][DONE] whole-mouse-brain macs-v11-norestart-acq-entropy b5 s2 | test patch Dice=0.7241 VI=0.3296 | image Dice=0.8312 VI=0.3823 | 3939s diff --git a/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/result.json b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..51086449648571116dbc5f65561d99a315c561d6 --- /dev/null +++ b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/result.json @@ -0,0 +1,785 @@ +{ + "target": "whole-mouse-brain", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 2, + "config": { + "task": "membrane", + "target": "whole-mouse-brain", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "whole-mouse-brain", + "--variants", + "macs-v11-norestart-acq-entropy", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.00761798355588039, + 0.2559331647259477, + 0.03865987551029595, + 0.0753340114213908, + 0.3764144201177104, + 0.13449626142769927, + 0.09581294628140101, + 0.015731336959674627 + ], + "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": [ + 0.00761798355588039, + 0.2559331647259477, + 0.03865987551029595, + 0.0753340114213908, + 0.3764144201177104, + 0.13449626142769927, + 0.09581294628140101, + 0.015731336959674627 + ], + "per_source": [ + 0.14507434834977073, + 0.041950586009985594, + 0.0008814746660836073, + 0.14183397467001313, + 0.02035313133415202, + 0.40237139544038975, + 0.24338615668507596, + 0.0041489328445293286 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.11311692308315584, + "mmd2_mixture": 0.12945845805399875, + "mmd2_uniform": 0.15528887887137932, + "mmd2_single_source": [ + 0.38071558003618083, + 0.17697908713268734, + 0.3194072222380283, + 0.24037631453841346, + 0.16266629855891124, + 0.27345204518121097, + 0.2563983777819251, + 0.2660430684822974 + ], + "median_squared_distance": 0.1859244048719758, + "gammas": [ + 1.3446325143390696, + 2.689265028678139, + 5.378530057356278, + 10.757060114712557, + 21.514120229425114 + ], + "max_weight": 0.3764144201177104, + "entropy": 1.6341534970843021, + "effective_sources": 5.125117733158912, + "min_weight": 0.00761798355588039 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.3322006978705003, + 0.46442575425639476, + 0.8760620010344873, + 0.3346079883794868, + 0.5415020997962443, + 0.22348725249801393, + 0.277061938603162, + 0.710986599609967 + ], + "standardized": [ + -0.6467264485566062, + -0.026349470463598257, + 1.9049762391239466, + -0.6354318599803043, + 0.33527934358594014, + -1.1567910351159896, + -0.9054279215228022, + 1.130471152929414 + ], + "tau": 0.5, + "max_weight": 0.40237139544038975, + "entropy": 1.508582930948667, + "effective_sources": 4.5203206470852155, + "min_weight": 0.0008814746660836073 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "seconds": 6.6 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.3857342957712184, + 0.3164308639401021, + 0.10997194512386699, + 0.2588511091009309, + 0.32881765823766074, + 0.40721108437406883, + 0.43978952859529485, + 0.30535964718503905 + ], + "ensemble_val_dice": { + "mixture": 0.3845120898706078, + "mixture@raw": 0.3998895285882327, + "mixture@logit": 0.38020955262748224, + "per_source": 0.4199047055822174, + "per_source@raw": 0.42186208777273215, + "per_source@logit": 0.41929804644233215, + "uniform": 0.41282903446624, + "uniform@raw": 0.3877116419741835, + "uniform@logit": 0.4054966159612228 + }, + "val_patches": 1024 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 2.642308473587036, + 3.547258138656616 + ], + "top_source": "microns_basil", + "weight_stats": { + "max_weight": 0.3764144201177104, + "entropy": 1.6341534970843021, + "effective_sources": 5.125117733158912, + "min_weight": 0.00761798355588039 + }, + "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": 1122.9, + "round0_val": { + "val_dice": 0.39316518885694673, + "val_vi": 0.5071037992390497, + "val_ece": 0.13929538678771358 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 49, + "selection_seconds": 113.6, + "round_seconds": 573.4, + "teacher_dice_labeled": [ + 0.41667508794746294, + 0.3592800679911643, + 0.15312328284483975, + 0.3102883118187754, + 0.24828297584066997, + 0.40443558048329065, + 0.45309956020060516, + 0.3151859425557686 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 745730, + "evidence_nonempty_patches": 43, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.2919523581210136, + "restart_teacher_dice": 0.45309956020060516, + "weights": [ + 0.052345161843924994, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.947654838156075, + 0.0 + ], + "ensemble_choice": "top2", + "candidate_dice": { + "mixture": 0.41707512159255195, + "uniform": 0.39628150854804123, + "top1": 0.45309956020060516, + "top2": 0.4560005583432433, + "top3": 0.42428559646215114 + }, + "val_dice": 0.6299640712246265, + "val_vi": 0.5106772900495546, + "val_ece": 0.03300236752641396 + }, + { + "round": 2, + "n_labeled": 98, + "selection_seconds": 31.1, + "round_seconds": 78.3, + "teacher_dice_labeled": [ + 0.3249594717244935, + 0.2618002400057593, + 0.11626361986650778, + 0.2682125036278144, + 0.19005829309477, + 0.3157203697046871, + 0.357571670301185, + 0.26891050108232195 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1155835, + "evidence_nonempty_patches": 80, + "first_correction": false, + "restart": "", + "weights": [ + 0.0542688419808356, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.9457311580191643, + 0.0 + ], + "ensemble_choice": "top2", + "candidate_dice": { + "mixture": 0.30164075804164814, + "uniform": 0.30147095589245054, + "top1": 0.357571670301185, + "top2": 0.35967682146094637, + "top3": 0.32455408079063386 + }, + "pseudo_student_dice": 0.33799498785878834, + "pseudo_ensemble_dice": 0.12497048992244641, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.6671426429611568, + "val_vi": 0.42133914138201256, + "val_ece": 0.015966444291345954 + }, + { + "round": 3, + "n_labeled": 147, + "selection_seconds": 30.8, + "round_seconds": 394.7, + "teacher_dice_labeled": [ + 0.3362206587862502, + 0.33035776451296633, + 0.11570615054827438, + 0.25699455697450335, + 0.23065137965115518, + 0.34275950768617824, + 0.4148087923555059, + 0.29939804919802215 + ], + "teacher_dice_pool": [ + 0.38312200607089864, + 0.37357687854150773, + 0.1536548607926944, + 0.2644659774961156, + 0.33231233949642547, + 0.44608789466873955, + 0.5548695967709021, + 0.29333183491515874 + ], + "evidence_fg_pixels": 1901791, + "evidence_nonempty_patches": 106, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.3647942449410265, + "uniform": 0.3424460851717929, + "top1": 0.4148087923555059, + "top2": 0.3811854830612246, + "top3": 0.3826582850005112 + }, + "pseudo_student_dice": 0.4902302504374955, + "pseudo_ensemble_dice": 0.11578052428171628, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.6500437319549233, + "val_vi": 0.46661771310585776, + "val_ece": 0.026616772063627447 + }, + { + "round": 4, + "n_labeled": 196, + "selection_seconds": 30.6, + "round_seconds": 438.1, + "teacher_dice_labeled": [ + 0.3519483730790272, + 0.3550619667303971, + 0.11032225058098022, + 0.2531933417507679, + 0.2190875017687, + 0.3241839148029759, + 0.3952944285319471, + 0.32148791585466907 + ], + "teacher_dice_pool": [ + 0.4121643222363875, + 0.37964937638880697, + 0.15275562810191143, + 0.3040249259871205, + 0.3021430411408676, + 0.4332669406483501, + 0.5293982426240719, + 0.3288904127260634 + ], + "evidence_fg_pixels": 2568324, + "evidence_nonempty_patches": 139, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.36402664300994514, + "uniform": 0.3222085755550682, + "top1": 0.3952944285319471, + "top2": 0.3887335829311641, + "top3": 0.3864764612355511 + }, + "pseudo_student_dice": 0.53600076651133, + "pseudo_ensemble_dice": 0.22566554473495054, + "pseudo_source": "student", + "consistency_scale": 0.12, + "val_dice": 0.6867063671616198, + "val_vi": 0.4334551684265923, + "val_ece": 0.006732057576624399 + }, + { + "round": 5, + "n_labeled": 245, + "selection_seconds": 30.3, + "round_seconds": 458.7, + "teacher_dice_labeled": [ + 0.3584230095199944, + 0.38096205524839705, + 0.11117752910496465, + 0.2439638336867353, + 0.21809780395525596, + 0.31570369170261414, + 0.3941231445083987, + 0.33228651701690143 + ], + "teacher_dice_pool": [ + 0.3875614767122367, + 0.400028641768668, + 0.15084672589212084, + 0.2667770048170303, + 0.31216759610745726, + 0.4302824637590289, + 0.5282474877153657, + 0.31952189062440434 + ], + "evidence_fg_pixels": 3582198, + "evidence_nonempty_patches": 178, + "first_correction": false, + "restart": "", + "weights": [ + 0.01637661851566227, + 0.6892814651914534, + 0.0, + 0.0, + 0.0, + 0.0, + 0.29434191629288425, + 0.0 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.3906889930951955, + "uniform": 0.31898344193918804, + "top1": 0.3941231445083987, + "top2": 0.40592475485919916, + "top3": 0.40537019629518667 + }, + "pseudo_student_dice": 0.6111341036273602, + "pseudo_ensemble_dice": 0.448249058234845, + "pseudo_source": "student", + "consistency_scale": 0.37, + "val_dice": 0.7010087760930706, + "val_vi": 0.4148022188492324, + "val_ece": 0.009600339954623855 + }, + { + "round": 6, + "n_labeled": 294, + "selection_seconds": 30.1, + "round_seconds": 481.0, + "teacher_dice_labeled": [ + 0.37813906679059617, + 0.3929658426455666, + 0.10759623232507429, + 0.2470899709358254, + 0.2097623786143264, + 0.3154036114080012, + 0.3847248817593051, + 0.3467007071990453 + ], + "teacher_dice_pool": [ + 0.4000509789602617, + 0.39176571714136516, + 0.1482235973977921, + 0.27994162679950685, + 0.2977878407705155, + 0.4326847309976392, + 0.5262451617634437, + 0.3340916004034753 + ], + "evidence_fg_pixels": 4504610, + "evidence_nonempty_patches": 224, + "first_correction": false, + "restart": "", + "weights": [ + 0.013069321687079801, + 0.5092504873046682, + 0.004433236159290681, + 0.03485500253066859, + 0.11990260091834273, + 0.1232158588333893, + 0.1755654455535038, + 0.01970804701305701 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.40943043039142457, + "uniform": 0.30811234119371467, + "top1": 0.3929658426455666, + "top2": 0.41077618458479637, + "top3": 0.41068862692388175 + }, + "pseudo_student_dice": 0.6500897437229997, + "pseudo_ensemble_dice": 0.2980986095351606, + "pseudo_source": "student", + "consistency_scale": 0.5, + "val_dice": 0.7046566133521246, + "val_vi": 0.39338360372380154, + "val_ece": 0.007922761299779176 + }, + { + "round": 7, + "n_labeled": 294, + "selection_seconds": 0.0, + "round_seconds": 190.4, + "final_consolidation": true, + "val_dice": 0.715600054385459, + "val_vi": 0.3905030362479474, + "val_ece": 0.0081281918504199 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.2505878322752116, + "auroc_top20_error": 0.3924166090497315 + }, + "al_seconds": 2655.4, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.013069321687079801, + 0.5092504873046682, + 0.004433236159290681, + 0.03485500253066859, + 0.11990260091834273, + 0.1232158588333893, + 0.1755654455535038, + 0.01970804701305701 + ], + "weight_stats_final": { + "max_weight": 0.5092504873046682, + "entropy": 1.4364938946277095, + "effective_sources": 4.205923524457596, + "min_weight": 0.004433236159290681 + }, + "top_source_final": "dhanyasi-P14-mouse-cerebellum", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.7054563808799655, + "patch_iou": 0.6213611844092362, + "patch_precision": 0.7248983941572255, + "patch_recall": 0.8271769452900526, + "patch_vi": 0.35090675928415266, + "n_patches": 1300, + "cal_ece": 0.00821402980845419, + "cal_nll": 0.19145299976349267, + "cal_brier": 0.05514656907793308, + "cal_mean_confidence": 0.9346074646634128, + "cal_accuracy": 0.9264100060096154, + "cal_predicted_prevalence": 0.23219651442307693, + "cal_true_prevalence": 0.21078049879807692, + "cal_mean_probability": 0.244326443916351 + }, + "test_notta": { + "patch_dice": 0.7054563808799655, + "patch_iou": 0.6213611844092362, + "patch_precision": 0.7248983941572255, + "patch_recall": 0.8271769452900526, + "patch_vi": 0.35090675928415266, + "n_patches": 1300, + "cal_ece": 0.00821402980845419, + "cal_nll": 0.19145299976349267, + "cal_brier": 0.05514656907793308, + "cal_mean_confidence": 0.9346074646634128, + "cal_accuracy": 0.9264100060096154, + "cal_predicted_prevalence": 0.23219651442307693, + "cal_true_prevalence": 0.21078049879807692, + "cal_mean_probability": 0.244326443916351 + }, + "test": { + "patch_dice": 0.7240810547947243, + "patch_iou": 0.6434686049972183, + "patch_precision": 0.7535363328863108, + "patch_recall": 0.8265486522250965, + "patch_vi": 0.3295530513344462, + "n_patches": 1300, + "image_dice": 0.8312241136800578, + "image_iou": 0.726982790269668, + "image_precision": 0.8133640751108497, + "image_recall": 0.8773653978839356, + "image_vi": 0.38227195684253606, + "n_images": 52, + "cal_ece": 0.00327916983898916, + "cal_nll": 0.17840624343436598, + "cal_brier": 0.05143879160628608, + "cal_mean_confidence": 0.9305477422689723, + "cal_accuracy": 0.9314761117788461, + "cal_predicted_prevalence": 0.2275420673076923, + "cal_true_prevalence": 0.21078049879807692, + "cal_mean_probability": 0.2429929328227741 + }, + "eval_seconds": 68.5, + "total_seconds": 3939.2, + "resources": { + "phases": { + "setup": { + "seconds": 1.48, + "peak_cpu_rss_mb": 2192.2, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 84.15, + "peak_cpu_rss_mb": 4183.1, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 1122.94, + "peak_cpu_rss_mb": 5940.4, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4618.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2655.43, + "peak_cpu_rss_mb": 5940.4, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4626.0, + "rounds": 6 + }, + "inference": { + "seconds": 68.47, + "peak_cpu_rss_mb": 5940.4, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4626.0, + "patches": 1300, + "images": 52, + "patches_per_second": 19.0, + "seconds_per_image": 1.317 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5940.4, + "total_seconds": 3932.5 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/whole-mouse-brain/macs-v11-norestart-acq-entropy_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-entropy/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..609806470b0df6ccf574d9d144618f754020625b --- /dev/null +++ b/macs_v2_r26/runs/membrane/whole-mouse-brain/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 +1,49,113.6,573.4,"[0.41667508794746294, 0.3592800679911643, 0.15312328284483975, 0.3102883118187754, 0.24828297584066997, 0.40443558048329065, 0.45309956020060516, 0.3151859425557686]",,745730,43,True,kept,0.2919523581210136,0.45309956020060516,"[0.052345161843924994, 0.0, 0.0, 0.0, 0.0, 0.0, 0.947654838156075, 0.0]",top2,"{'mixture': 0.41707512159255195, 'uniform': 0.39628150854804123, 'top1': 0.45309956020060516, 'top2': 0.4560005583432433, 'top3': 0.42428559646215114}",0.6299640712246265,0.5106772900495546,0.03300236752641396,,,,, +2,98,31.1,78.3,"[0.3249594717244935, 0.2618002400057593, 0.11626361986650778, 0.2682125036278144, 0.19005829309477, 0.3157203697046871, 0.357571670301185, 0.26891050108232195]",,1155835,80,False,,,,"[0.0542688419808356, 0.0, 0.0, 0.0, 0.0, 0.0, 0.9457311580191643, 0.0]",top2,"{'mixture': 0.30164075804164814, 'uniform': 0.30147095589245054, 'top1': 0.357571670301185, 'top2': 0.35967682146094637, 'top3': 0.32455408079063386}",0.6671426429611568,0.42133914138201256,0.015966444291345954,0.33799498785878834,0.12497048992244641,student,0.0, +3,147,30.8,394.7,"[0.3362206587862502, 0.33035776451296633, 0.11570615054827438, 0.25699455697450335, 0.23065137965115518, 0.34275950768617824, 0.4148087923555059, 0.29939804919802215]","[0.38312200607089864, 0.37357687854150773, 0.1536548607926944, 0.2644659774961156, 0.33231233949642547, 0.44608789466873955, 0.5548695967709021, 0.29333183491515874]",1901791,106,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.3647942449410265, 'uniform': 0.3424460851717929, 'top1': 0.4148087923555059, 'top2': 0.3811854830612246, 'top3': 0.3826582850005112}",0.6500437319549233,0.46661771310585776,0.026616772063627447,0.4902302504374955,0.11578052428171628,student,0.0, +4,196,30.6,438.1,"[0.3519483730790272, 0.3550619667303971, 0.11032225058098022, 0.2531933417507679, 0.2190875017687, 0.3241839148029759, 0.3952944285319471, 0.32148791585466907]","[0.4121643222363875, 0.37964937638880697, 0.15275562810191143, 0.3040249259871205, 0.3021430411408676, 0.4332669406483501, 0.5293982426240719, 0.3288904127260634]",2568324,139,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.36402664300994514, 'uniform': 0.3222085755550682, 'top1': 0.3952944285319471, 'top2': 0.3887335829311641, 'top3': 0.3864764612355511}",0.6867063671616198,0.4334551684265923,0.006732057576624399,0.53600076651133,0.22566554473495054,student,0.12, +5,245,30.3,458.7,"[0.3584230095199944, 0.38096205524839705, 0.11117752910496465, 0.2439638336867353, 0.21809780395525596, 0.31570369170261414, 0.3941231445083987, 0.33228651701690143]","[0.3875614767122367, 0.400028641768668, 0.15084672589212084, 0.2667770048170303, 0.31216759610745726, 0.4302824637590289, 0.5282474877153657, 0.31952189062440434]",3582198,178,False,,,,"[0.01637661851566227, 0.6892814651914534, 0.0, 0.0, 0.0, 0.0, 0.29434191629288425, 0.0]",top3,"{'mixture': 0.3906889930951955, 'uniform': 0.31898344193918804, 'top1': 0.3941231445083987, 'top2': 0.40592475485919916, 'top3': 0.40537019629518667}",0.7010087760930706,0.4148022188492324,0.009600339954623855,0.6111341036273602,0.448249058234845,student,0.37, +6,294,30.1,481.0,"[0.37813906679059617, 0.3929658426455666, 0.10759623232507429, 0.2470899709358254, 0.2097623786143264, 0.3154036114080012, 0.3847248817593051, 0.3467007071990453]","[0.4000509789602617, 0.39176571714136516, 0.1482235973977921, 0.27994162679950685, 0.2977878407705155, 0.4326847309976392, 0.5262451617634437, 0.3340916004034753]",4504610,224,False,,,,"[0.013069321687079801, 0.5092504873046682, 0.004433236159290681, 0.03485500253066859, 0.11990260091834273, 0.1232158588333893, 0.1755654455535038, 0.01970804701305701]",mixture,"{'mixture': 0.40943043039142457, 'uniform': 0.30811234119371467, 'top1': 0.3929658426455666, 'top2': 0.41077618458479637, 'top3': 0.41068862692388175}",0.7046566133521246,0.39338360372380154,0.007922761299779176,0.6500897437229997,0.2980986095351606,student,0.5, +7,294,0.0,190.4,,,,,,,,,,,,0.715600054385459,0.3905030362479474,0.0081281918504199,,,,,True diff --git a/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..a5f47ace9eea71f2ee3cfe53c56b23d9792ea004 --- /dev/null +++ b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt @@ -0,0 +1,103 @@ +[02:45:19][ 0s][START] target=whole-mouse-brain variant=macs-v11-norestart-acq-random budget=5.0% seed=1 | {"task": "membrane", "target": "whole-mouse-brain", "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": 8, "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} +[02:45:19][ 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. +[02:45:21][ 2s][DATA] whole-mouse-brain: train=5925 val=1250 test=1300 patches +[02:45:23][ 4s][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 +[02:45:23][ 4s][FEATURES] 72 domain x encoder feature sets (72 cached) | n=512/domain | 0s +[02:45:29][ 11s][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 +[02:45:29][ 11s][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 +[02:46:44][ 85s][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 +[02:46:44][ 85s][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 +[02:46:44][ 85s][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 +[02:46:44][ 85s][INIT] student initialised from teacher microns (weight 0.376) +[02:46:59][ 100s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[02:46:59][ 100s][KD] distilling the mixture ensemble (prob space) into the student +[02:51:00][ 341s][KD] soft labels for 5925 patches from 8 teachers in 241s +[02:53:56][ 517s][KD] epoch 1/5 | patches=5925 | loss=0.6724 | 176s +[02:56:52][ 694s][KD] epoch 2/5 | patches=5925 | loss=0.4930 | 176s +[02:59:49][ 870s][KD] epoch 3/5 | patches=5925 | loss=0.4658 | 176s +[03:02:45][ 1046s][KD] epoch 4/5 | patches=5925 | loss=0.4598 | 176s +[03:05:41][ 1222s][KD] epoch 5/5 | patches=5925 | loss=0.4555 | 176s +[03:05:47][ 1228s][ROUND0] before any target label: val_dice=0.3878 val_vi=0.5059 +[03:05:47][ 1228s][AL] patches=5925 (pool 5925 after dropping blank; free blank negatives: True) | budget=296 (5.0%) | rounds=6 | per_round=49 | acquisition=random | first_round=stratified +[03:06:18][ 1259s][AL] round-1 uncertainty quality on 5925 patches: {'spearman_uncertainty_error': -0.006734909767982264, 'auroc_top20_error': 0.5056221403265146} +[03:07:40][ 1341s][AL] round 1: stratified cold start: 25 patches from the membrane-rich half (ensemble prevalence >= 0.214) + 24 from the membrane-poor half of 2000 candidates +[03:07:48][ 1349s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.415, uniform=0.393, top1=0.460, top2=0.464, top3=0.490 -> using top3 +[03:07:48][ 1349s][WEIGHTS] round 1: student kept (labeled Dice 0.314 vs best teacher oct-sfl 0.460); distilling the corrected ensemble into it +[03:07:48][ 1349s][WEIGHTS] round 1: teacher Dice on 49 labeled patches + 0 blank (774194 fg px): c-elegans=0.431, cerebellum=0.372, flywire=0.134, berghia=0.291, microns=0.284, oct-glia=0.371, oct-sfl=0.460, snemi=0.295 | weights: c-elegans=0.027, cerebellum=0.511, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.462, snemi=0.000 +[03:09:19][ 1440s][KD] soft labels for 5876 patches from 8 teachers in 90s +[03:12:13][ 1615s][KD] epoch 1/2 | patches=5876 | loss=0.4892 | 175s +[03:15:08][ 1789s][KD] epoch 2/2 | patches=5876 | loss=0.4853 | 174s +[03:15:39][ 1820s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 31s +[03:15:43][ 1824s][FT] round 1/6 epoch 1/5 | patches=49 | loss=1.3866 +[03:15:46][ 1827s][FT] round 1/6 epoch 2/5 | patches=49 | loss=1.2764 +[03:15:50][ 1831s][FT] round 1/6 epoch 3/5 | patches=49 | loss=1.2397 +[03:15:54][ 1835s][FT] round 1/6 epoch 4/5 | patches=49 | loss=1.2255 +[03:15:57][ 1838s][FT] round 1/6 epoch 5/5 | patches=49 | loss=1.2378 +[03:16:03][ 1844s][AL] round 1/6 | labeled=49 | selection_seconds=112.9000 round_seconds=610.4000 restart_student_dice=0.3142 restart_teacher_dice=0.4601 val_dice=0.6513 val_vi=0.4647 val_ece=0.0896 +[03:16:18][ 1859s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.436, uniform=0.403, top1=0.475, top2=0.478, top3=0.452 -> using top2 +[03:16:18][ 1859s][WEIGHTS] round 2: teacher Dice on 98 labeled patches + 0 blank (1631452 fg px): c-elegans=0.442, cerebellum=0.374, flywire=0.140, berghia=0.295, microns=0.276, oct-glia=0.378, oct-sfl=0.475, snemi=0.324 | weights: c-elegans=0.054, cerebellum=0.000, flywire=0.000, berghia=0.000, microns=0.000, oct-glia=0.000, oct-sfl=0.946, snemi=0.000 +[03:16:34][ 1875s][CONS] round 2: on the 49 new patches student Dice=0.828 vs ensemble 0.491 -> pseudo-labels from the student | consistency x1.00 +[03:16:40][ 1881s][FT] round 2/6 epoch 1/5 | patches=98 | loss=1.0456 +[03:16:47][ 1888s][FT] round 2/6 epoch 2/5 | patches=98 | loss=0.9731 +[03:16:53][ 1894s][FT] round 2/6 epoch 3/5 | patches=98 | loss=0.9812 +[03:17:00][ 1901s][FT] round 2/6 epoch 4/5 | patches=98 | loss=0.8799 +[03:17:06][ 1907s][FT] round 2/6 epoch 5/5 | patches=98 | loss=0.9492 +[03:17:12][ 1913s][AL] round 2/6 | labeled=98 | selection_seconds=0.2000 round_seconds=63.2000 pseudo_student_dice=0.8275 pseudo_ensemble_dice=0.4908 consistency_scale=1.0000 val_dice=0.6900 val_vi=0.3935 val_ece=0.0466 +[03:19:59][ 2080s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.396, cerebellum=0.379, flywire=0.160, berghia=0.279, microns=0.333, oct-glia=0.456, oct-sfl=0.545, snemi=0.289 | blended with labeled Dice at 0.50 +[03:22:36][ 2237s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.461, uniform=0.447, top1=0.505, top2=0.477, top3=0.480 -> using top1 +[03:22:36][ 2237s][WEIGHTS] round 3: teacher Dice on 147 labeled patches + 0 blank (2470847 fg px): c-elegans=0.421, cerebellum=0.375, flywire=0.151, berghia=0.280, microns=0.307, oct-glia=0.421, oct-sfl=0.505, snemi=0.310 | 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 +[03:22:51][ 2253s][CONS] round 3: on the 49 new patches student Dice=0.848 vs ensemble 0.445 -> pseudo-labels from the student | consistency x1.00 +[03:23:01][ 2262s][FT] round 3/6 epoch 1/5 | patches=147 | loss=0.8251 +[03:23:10][ 2271s][FT] round 3/6 epoch 2/5 | patches=147 | loss=0.8697 +[03:23:19][ 2281s][FT] round 3/6 epoch 3/5 | patches=147 | loss=0.8617 +[03:23:29][ 2290s][FT] round 3/6 epoch 4/5 | patches=147 | loss=0.8097 +[03:23:38][ 2299s][FT] round 3/6 epoch 5/5 | patches=147 | loss=0.8147 +[03:23:44][ 2305s][AL] round 3/6 | labeled=147 | selection_seconds=0.2000 round_seconds=386.1000 pseudo_student_dice=0.8483 pseudo_ensemble_dice=0.4447 consistency_scale=1.0000 val_dice=0.6998 val_vi=0.3811 val_ece=0.0066 +[03:26:34][ 2475s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.394, cerebellum=0.379, flywire=0.159, berghia=0.289, microns=0.326, oct-glia=0.460, oct-sfl=0.534, snemi=0.295 | blended with labeled Dice at 0.50 +[03:29:13][ 2635s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.456, uniform=0.446, top1=0.497, top2=0.467, top3=0.469 -> using top1 +[03:29:13][ 2635s][WEIGHTS] round 4: teacher Dice on 196 labeled patches + 0 blank (3255460 fg px): c-elegans=0.417, cerebellum=0.378, flywire=0.151, berghia=0.283, microns=0.306, oct-glia=0.423, oct-sfl=0.497, snemi=0.314 | 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 +[03:29:29][ 2650s][CONS] round 4: on the 49 new patches student Dice=0.889 vs ensemble 0.444 -> pseudo-labels from the student | consistency x1.00 +[03:29:41][ 2662s][FT] round 4/6 epoch 1/5 | patches=196 | loss=0.7037 +[03:29:53][ 2674s][FT] round 4/6 epoch 2/5 | patches=196 | loss=0.6993 +[03:30:05][ 2687s][FT] round 4/6 epoch 3/5 | patches=196 | loss=0.6914 +[03:30:18][ 2699s][FT] round 4/6 epoch 4/5 | patches=196 | loss=0.6840 +[03:30:30][ 2711s][FT] round 4/6 epoch 5/5 | patches=196 | loss=0.6446 +[03:30:36][ 2717s][AL] round 4/6 | labeled=196 | selection_seconds=0.2000 round_seconds=405.8000 pseudo_student_dice=0.8894 pseudo_ensemble_dice=0.4436 consistency_scale=1.0000 val_dice=0.7284 val_vi=0.3506 val_ece=0.0106 +[03:33:29][ 2890s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.397, cerebellum=0.379, flywire=0.154, berghia=0.284, microns=0.330, oct-glia=0.463, oct-sfl=0.543, snemi=0.305 | blended with labeled Dice at 0.50 +[03:36:12][ 3054s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.462, uniform=0.453, top1=0.509, top2=0.477, top3=0.479 -> using top1 +[03:36:12][ 3054s][WEIGHTS] round 5: teacher Dice on 245 labeled patches + 0 blank (3796031 fg px): c-elegans=0.407, cerebellum=0.378, flywire=0.149, berghia=0.280, microns=0.311, oct-glia=0.434, oct-sfl=0.509, snemi=0.312 | 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 +[03:36:28][ 3069s][CONS] round 5: on the 49 new patches student Dice=0.815 vs ensemble 0.562 -> pseudo-labels from the student | consistency x1.00 +[03:36:43][ 3084s][FT] round 5/6 epoch 1/5 | patches=245 | loss=0.6626 +[03:36:58][ 3099s][FT] round 5/6 epoch 2/5 | patches=245 | loss=0.6472 +[03:37:13][ 3114s][FT] round 5/6 epoch 3/5 | patches=245 | loss=0.6301 +[03:37:28][ 3129s][FT] round 5/6 epoch 4/5 | patches=245 | loss=0.6262 +[03:37:43][ 3144s][FT] round 5/6 epoch 5/5 | patches=245 | loss=0.6015 +[03:37:49][ 3150s][AL] round 5/6 | labeled=245 | selection_seconds=0.2000 round_seconds=427.2000 pseudo_student_dice=0.8150 pseudo_ensemble_dice=0.5618 consistency_scale=1.0000 val_dice=0.7361 val_vi=0.3468 val_ece=0.0179 +[03:40:46][ 3327s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: c-elegans=0.411, cerebellum=0.385, flywire=0.156, berghia=0.288, microns=0.321, oct-glia=0.459, oct-sfl=0.540, snemi=0.320 | blended with labeled Dice at 0.50 +[03:43:33][ 3494s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.461, uniform=0.448, top1=0.509, top2=0.476, top3=0.478 -> using top1 +[03:43:33][ 3494s][WEIGHTS] round 6: teacher Dice on 294 labeled patches + 0 blank (4607078 fg px): c-elegans=0.413, cerebellum=0.381, flywire=0.150, berghia=0.286, microns=0.307, oct-glia=0.431, oct-sfl=0.509, snemi=0.321 | 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 +[03:43:48][ 3509s][CONS] round 6: on the 49 new patches student Dice=0.801 vs ensemble 0.486 -> pseudo-labels from the student | consistency x1.00 +[03:44:06][ 3527s][FT] round 6/6 epoch 1/5 | patches=294 | loss=0.6618 +[03:44:24][ 3545s][FT] round 6/6 epoch 2/5 | patches=294 | loss=0.6521 +[03:44:41][ 3563s][FT] round 6/6 epoch 3/5 | patches=294 | loss=0.5995 +[03:44:59][ 3581s][FT] round 6/6 epoch 4/5 | patches=294 | loss=0.5934 +[03:45:17][ 3598s][FT] round 6/6 epoch 5/5 | patches=294 | loss=0.5715 +[03:45:23][ 3604s][AL] round 6/6 | labeled=294 | selection_seconds=0.2000 round_seconds=448.7000 pseudo_student_dice=0.8012 pseudo_ensemble_dice=0.4864 consistency_scale=1.0000 val_dice=0.7406 val_vi=0.3158 val_ece=0.0196 +[03:45:33][ 3614s][CONS] final consolidation: pseudo-labels refreshed from the student +[03:45:51][ 3632s][FT] final 1/10 epoch 1/1 | patches=294 | loss=0.5299 +[03:46:09][ 3650s][FT] final 2/10 epoch 1/1 | patches=294 | loss=0.5280 +[03:46:27][ 3668s][FT] final 3/10 epoch 1/1 | patches=294 | loss=0.5384 +[03:46:45][ 3686s][FT] final 4/10 epoch 1/1 | patches=294 | loss=0.5264 +[03:47:03][ 3704s][FT] final 5/10 epoch 1/1 | patches=294 | loss=0.5097 +[03:47:20][ 3722s][FT] final 6/10 epoch 1/1 | patches=294 | loss=0.5057 +[03:47:38][ 3740s][FT] final 7/10 epoch 1/1 | patches=294 | loss=0.4599 +[03:47:56][ 3757s][FT] final 8/10 epoch 1/1 | patches=294 | loss=0.4586 +[03:48:14][ 3775s][FT] final 9/10 epoch 1/1 | patches=294 | loss=0.4521 +[03:48:32][ 3793s][FT] final 10/10 epoch 1/1 | patches=294 | loss=0.4527 +[03:48:38][ 3799s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=189.0000 val_dice=0.7512 val_vi=0.3181 val_ece=0.0216 +[03:48:38][ 3799s][WEIGHTS] final: 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 +[03:49:39][ 3860s][TEST] raw student (no EMA, no TTA): patch_dice=0.7474 patch_vi=0.2821 +[03:49:47][ 3868s][TEST] final model without TTA: patch_dice=0.7474 patch_vi=0.2821 +[03:49:47][ 3868s][RESOURCES] {"phases": {"setup": {"seconds": 1.47, "peak_cpu_rss_mb": 2183.9, "peak_gpu_alloc_mb": 1066.3, "peak_gpu_reserved_mb": 1086.0}, "features_and_weights": {"seconds": 95.65, "peak_cpu_rss_mb": 4183.8, "peak_gpu_alloc_mb": 2820.1, "peak_gpu_reserved_mb": 3212.0}, "distillation": {"seconds": 1122.62, "peak_cpu_rss_mb": 5939.8, "peak_gpu_alloc_mb": 4300.6, "peak_gpu_reserved_mb": 4618.0, "kd_epochs": 5}, "active_learning": {"seconds": 2570.92, "peak_cpu_rss_mb": 5939.8, "peak_gpu_alloc_mb": 4420.9, "peak_gpu_reserved_mb": 4626.0, "rounds": 6}, "inference": {"seconds": 68.69, "peak_cpu_rss_mb": 5939.8, "peak_gpu_alloc_mb": 2623.4, "peak_gpu_reserved_mb": 4626.0, "patches": 1300, "images": 52, "patches_per_second": 18.9, "seconds_per_image": 1.321}}, "peak_gpu_alloc_mb": 4420.9, "peak_cpu_rss_mb": 5939.8, "total_seconds": 3859.3} +[03:49:47][ 3868s][TEST] patch_dice=0.7625 patch_iou=0.6916 patch_precision=0.8210 patch_recall=0.8202 patch_vi=0.2623 image_dice=0.8727 image_iou=0.7844 image_precision=0.8902 image_recall=0.8710 image_vi=0.3002 cal_ece=0.0089 cal_nll=0.1345 cal_brier=0.0377 cal_mean_confidence=0.9586 cal_accuracy=0.9497 cal_predicted_prevalence=0.2075 cal_true_prevalence=0.2108 cal_mean_probability=0.2073 +[03:49:47][ 3869s][DONE] whole-mouse-brain macs-v11-norestart-acq-random b5 s1 | test patch Dice=0.7625 VI=0.2623 | image Dice=0.8727 VI=0.3002 | 3868s diff --git a/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/result.json b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..2d28ae43952412ab946e44dc1a9ae5efc4d2bb3c --- /dev/null +++ b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/result.json @@ -0,0 +1,785 @@ +{ + "target": "whole-mouse-brain", + "variant": "macs-v11-norestart-acq-random", + "budget": 5.0, + "seed": 1, + "config": { + "task": "membrane", + "target": "whole-mouse-brain", + "variant": "macs-v11-norestart-acq-random", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "whole-mouse-brain", + "--variants", + "macs-v11-norestart-acq-random", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "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": [ + 0.00761798355588039, + 0.2559331647259477, + 0.03865987551029595, + 0.0753340114213908, + 0.3764144201177104, + 0.13449626142769927, + 0.09581294628140101, + 0.015731336959674627 + ], + "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": [ + 0.00761798355588039, + 0.2559331647259477, + 0.03865987551029595, + 0.0753340114213908, + 0.3764144201177104, + 0.13449626142769927, + 0.09581294628140101, + 0.015731336959674627 + ], + "per_source": [ + 0.14507434834977073, + 0.041950586009985594, + 0.0008814746660836073, + 0.14183397467001313, + 0.02035313133415202, + 0.40237139544038975, + 0.24338615668507596, + 0.0041489328445293286 + ], + "uniform": [ + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125, + 0.125 + ], + "mixture_info": { + "objective": 0.11311692308315584, + "mmd2_mixture": 0.12945845805399875, + "mmd2_uniform": 0.15528887887137932, + "mmd2_single_source": [ + 0.38071558003618083, + 0.17697908713268734, + 0.3194072222380283, + 0.24037631453841346, + 0.16266629855891124, + 0.27345204518121097, + 0.2563983777819251, + 0.2660430684822974 + ], + "median_squared_distance": 0.1859244048719758, + "gammas": [ + 1.3446325143390696, + 2.689265028678139, + 5.378530057356278, + 10.757060114712557, + 21.514120229425114 + ], + "max_weight": 0.3764144201177104, + "entropy": 1.6341534970843021, + "effective_sources": 5.125117733158912, + "min_weight": 0.00761798355588039 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.3322006978705003, + 0.46442575425639476, + 0.8760620010344873, + 0.3346079883794868, + 0.5415020997962443, + 0.22348725249801393, + 0.277061938603162, + 0.710986599609967 + ], + "standardized": [ + -0.6467264485566062, + -0.026349470463598257, + 1.9049762391239466, + -0.6354318599803043, + 0.33527934358594014, + -1.1567910351159896, + -0.9054279215228022, + 1.130471152929414 + ], + "tau": 0.5, + "max_weight": 0.40237139544038975, + "entropy": 1.508582930948667, + "effective_sources": 4.5203206470852155, + "min_weight": 0.0008814746660836073 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "octopus-vulgaris-vertical-lobe-glia-deep-neuropil", + "seconds": 6.6 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.3857342957712184, + 0.3164308639401021, + 0.10997194512386699, + 0.2588511091009309, + 0.32881765823766074, + 0.40721108437406883, + 0.43978952859529485, + 0.30535964718503905 + ], + "ensemble_val_dice": { + "mixture": 0.3845120898706078, + "mixture@raw": 0.3998895285882327, + "mixture@logit": 0.38020955262748224, + "per_source": 0.4199047055822174, + "per_source@raw": 0.42186208777273215, + "per_source@logit": 0.41929804644233215, + "uniform": 0.41282903446624, + "uniform@raw": 0.3877116419741835, + "uniform@logit": 0.4054966159612228 + }, + "val_patches": 1024 + }, + "temperatures": [ + 2.575598955154419, + 4.557764530181885, + 1.9903231859207153, + 8.0, + 1.9153105020523071, + 2.5512874126434326, + 2.642308473587036, + 3.547258138656616 + ], + "top_source": "microns_basil", + "weight_stats": { + "max_weight": 0.3764144201177104, + "entropy": 1.6341534970843021, + "effective_sources": 5.125117733158912, + "min_weight": 0.00761798355588039 + }, + "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": 1122.6, + "round0_val": { + "val_dice": 0.38775386975069404, + "val_vi": 0.5058585267638911, + "val_ece": 0.13684088848086562 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 49, + "selection_seconds": 112.9, + "round_seconds": 610.4, + "teacher_dice_labeled": [ + 0.4308437371432156, + 0.3720464899554072, + 0.13400588777866665, + 0.2906832527572902, + 0.28441244487328066, + 0.3714445937083196, + 0.46009083351660035, + 0.2953659731768761 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 774194, + "evidence_nonempty_patches": 43, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.3141552482233243, + "restart_teacher_dice": 0.46009083351660035, + "weights": [ + 0.027388975830910146, + 0.511103822265685, + 0.0, + 0.0, + 0.0, + 0.0, + 0.46150720190340494, + 0.0 + ], + "ensemble_choice": "top3", + "candidate_dice": { + "mixture": 0.4149047036603486, + "uniform": 0.3931444911550303, + "top1": 0.46009083351660035, + "top2": 0.46380715367111625, + "top3": 0.49022779815240064 + }, + "val_dice": 0.6512893174377166, + "val_vi": 0.4646900588599244, + "val_ece": 0.0896417007453064 + }, + { + "round": 2, + "n_labeled": 98, + "selection_seconds": 0.2, + "round_seconds": 63.2, + "teacher_dice_labeled": [ + 0.4424330801851372, + 0.37408893009446964, + 0.13996203868108806, + 0.29503398267841263, + 0.2762732338980843, + 0.3782590298845219, + 0.47481626651490355, + 0.3238189071307628 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1631452, + "evidence_nonempty_patches": 85, + "first_correction": false, + "restart": "", + "weights": [ + 0.0543864996701088, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.9456135003298912, + 0.0 + ], + "ensemble_choice": "top2", + "candidate_dice": { + "mixture": 0.4356210342800381, + "uniform": 0.40298612721687643, + "top1": 0.47481626651490355, + "top2": 0.47797929634414377, + "top3": 0.4516145396104275 + }, + "pseudo_student_dice": 0.8275029820365678, + "pseudo_ensemble_dice": 0.49077896830410456, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.6899674120835443, + "val_vi": 0.3934530539606663, + "val_ece": 0.046598598690968274 + }, + { + "round": 3, + "n_labeled": 147, + "selection_seconds": 0.2, + "round_seconds": 386.1, + "teacher_dice_labeled": [ + 0.4208794202637337, + 0.37488266988173913, + 0.150732512981738, + 0.280229848328845, + 0.30736600511815093, + 0.42137673253425356, + 0.504630320454826, + 0.3103878885056744 + ], + "teacher_dice_pool": [ + 0.3964222900856918, + 0.37934827125456433, + 0.15986144969132293, + 0.2785786219247478, + 0.33299853995204853, + 0.45638475100222214, + 0.5447891954337357, + 0.2891779812953886 + ], + "evidence_fg_pixels": 2470847, + "evidence_nonempty_patches": 129, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.46107124265136823, + "uniform": 0.4470064837241954, + "top1": 0.504630320454826, + "top2": 0.4774312960056267, + "top3": 0.47954494564006567 + }, + "pseudo_student_dice": 0.8483273711237067, + "pseudo_ensemble_dice": 0.4447035907635971, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.6997838980735469, + "val_vi": 0.38111564948839316, + "val_ece": 0.006551502400643054 + }, + { + "round": 4, + "n_labeled": 196, + "selection_seconds": 0.2, + "round_seconds": 405.8, + "teacher_dice_labeled": [ + 0.41735012268474536, + 0.37774968729190495, + 0.15129542306014881, + 0.2833705136853922, + 0.3061931648198789, + 0.4232066747177724, + 0.49688937580258463, + 0.314315337148227 + ], + "teacher_dice_pool": [ + 0.3943931472745527, + 0.37855105448483334, + 0.15943182820909615, + 0.2892929128772673, + 0.3257258256062749, + 0.4601414870055501, + 0.5342473085263869, + 0.2950977113256563 + ], + "evidence_fg_pixels": 3255460, + "evidence_nonempty_patches": 173, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.45595241469488296, + "uniform": 0.44581060629678865, + "top1": 0.49688937580258463, + "top2": 0.46677430571359046, + "top3": 0.4690154172263659 + }, + "pseudo_student_dice": 0.8894096619935941, + "pseudo_ensemble_dice": 0.4435759384979648, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.7284449934265951, + "val_vi": 0.35062932040639544, + "val_ece": 0.01061670729460243 + }, + { + "round": 5, + "n_labeled": 245, + "selection_seconds": 0.2, + "round_seconds": 427.2, + "teacher_dice_labeled": [ + 0.40692426871897414, + 0.3776928989394102, + 0.14920549941794428, + 0.2795811790533572, + 0.31140825070370776, + 0.43368676727852196, + 0.5091293274911578, + 0.3121180397885524 + ], + "teacher_dice_pool": [ + 0.3970244948724663, + 0.37928996723211816, + 0.15356131391556158, + 0.28419405601396547, + 0.3296259383047535, + 0.46253273706373266, + 0.542636728401153, + 0.3050203065494246 + ], + "evidence_fg_pixels": 3796031, + "evidence_nonempty_patches": 213, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.4624701194922756, + "uniform": 0.45307112409178935, + "top1": 0.5091293274911578, + "top2": 0.4770389469575239, + "top3": 0.4788482169595161 + }, + "pseudo_student_dice": 0.8150319479999724, + "pseudo_ensemble_dice": 0.5618041860169222, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.7360976720287663, + "val_vi": 0.3467971465172063, + "val_ece": 0.01788726699061594 + }, + { + "round": 6, + "n_labeled": 294, + "selection_seconds": 0.2, + "round_seconds": 448.7, + "teacher_dice_labeled": [ + 0.4131523804861451, + 0.38148291144188595, + 0.14959132590158677, + 0.2859449485429898, + 0.3070357015540196, + 0.4312325733658978, + 0.5087118932311401, + 0.321087704901554 + ], + "teacher_dice_pool": [ + 0.4114461558819054, + 0.3849032866661021, + 0.1555940064347239, + 0.2883347744854101, + 0.32091747898050904, + 0.4585541603506631, + 0.5398959859343434, + 0.31990600159271626 + ], + "evidence_fg_pixels": 4607078, + "evidence_nonempty_patches": 257, + "first_correction": false, + "restart": "", + "weights": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.46080784643360345, + "uniform": 0.4478514590552423, + "top1": 0.5087118932311401, + "top2": 0.47603001745833207, + "top3": 0.47780834110988624 + }, + "pseudo_student_dice": 0.8012167018453978, + "pseudo_ensemble_dice": 0.4863777827338713, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.7405788014844665, + "val_vi": 0.31581319789840717, + "val_ece": 0.019612879670904602 + }, + { + "round": 7, + "n_labeled": 294, + "selection_seconds": 0.0, + "round_seconds": 189.0, + "final_consolidation": true, + "val_dice": 0.7512189727699001, + "val_vi": 0.3181009168739638, + "val_ece": 0.021623735484845367 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.006734909767982264, + "auroc_top20_error": 0.5056221403265146 + }, + "al_seconds": 2570.9, + "source_archs": [ + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 1.0, + 0.0 + ], + "weight_stats_final": { + "max_weight": 1.0, + "entropy": -0.0, + "effective_sources": 1.0, + "min_weight": 0.0 + }, + "top_source_final": "octopus-vulgaris-vertical-lobe-sfltract", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.7474120376241677, + "patch_iou": 0.6728028364833846, + "patch_precision": 0.7935101241758076, + "patch_recall": 0.8178482491923542, + "patch_vi": 0.28209865954645047, + "n_patches": 1300, + "cal_ece": 0.0176318543517363, + "cal_nll": 0.1534192941139603, + "cal_brier": 0.04106024181296851, + "cal_mean_confidence": 0.9636918002651976, + "cal_accuracy": 0.9460599459134615, + "cal_predicted_prevalence": 0.2094666466346154, + "cal_true_prevalence": 0.21078049879807692, + "cal_mean_probability": 0.20719564276506328 + }, + "test_notta": { + "patch_dice": 0.7474120376241677, + "patch_iou": 0.6728028364833846, + "patch_precision": 0.7935101241758076, + "patch_recall": 0.8178482491923542, + "patch_vi": 0.28209865954645047, + "n_patches": 1300, + "cal_ece": 0.0176318543517363, + "cal_nll": 0.1534192941139603, + "cal_brier": 0.04106024181296851, + "cal_mean_confidence": 0.9636918002651976, + "cal_accuracy": 0.9460599459134615, + "cal_predicted_prevalence": 0.2094666466346154, + "cal_true_prevalence": 0.21078049879807692, + "cal_mean_probability": 0.20719564276506328 + }, + "test": { + "patch_dice": 0.7625086122496587, + "patch_iou": 0.6915794905331696, + "patch_precision": 0.8210280221083377, + "patch_recall": 0.8201574493133729, + "patch_vi": 0.26227260743302666, + "n_patches": 1300, + "image_dice": 0.8726720894721708, + "image_iou": 0.784447861167505, + "image_precision": 0.8901989738983472, + "image_recall": 0.87096714633107, + "image_vi": 0.3001978081447007, + "n_images": 52, + "cal_ece": 0.008869097294647845, + "cal_nll": 0.1345410909698496, + "cal_brier": 0.037687427717699104, + "cal_mean_confidence": 0.9585626016781935, + "cal_accuracy": 0.9497438401442307, + "cal_predicted_prevalence": 0.20746243990384616, + "cal_true_prevalence": 0.21078049879807692, + "cal_mean_probability": 0.2073494734025003 + }, + "eval_seconds": 68.7, + "total_seconds": 3867.7, + "resources": { + "phases": { + "setup": { + "seconds": 1.47, + "peak_cpu_rss_mb": 2183.9, + "peak_gpu_alloc_mb": 1066.3, + "peak_gpu_reserved_mb": 1086.0 + }, + "features_and_weights": { + "seconds": 95.65, + "peak_cpu_rss_mb": 4183.8, + "peak_gpu_alloc_mb": 2820.1, + "peak_gpu_reserved_mb": 3212.0 + }, + "distillation": { + "seconds": 1122.62, + "peak_cpu_rss_mb": 5939.8, + "peak_gpu_alloc_mb": 4300.6, + "peak_gpu_reserved_mb": 4618.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2570.92, + "peak_cpu_rss_mb": 5939.8, + "peak_gpu_alloc_mb": 4420.9, + "peak_gpu_reserved_mb": 4626.0, + "rounds": 6 + }, + "inference": { + "seconds": 68.69, + "peak_cpu_rss_mb": 5939.8, + "peak_gpu_alloc_mb": 2623.4, + "peak_gpu_reserved_mb": 4626.0, + "patches": 1300, + "images": 52, + "patches_per_second": 18.9, + "seconds_per_image": 1.321 + } + }, + "peak_gpu_alloc_mb": 4420.9, + "peak_cpu_rss_mb": 5939.8, + "total_seconds": 3859.3 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/whole-mouse-brain/macs-v11-norestart-acq-random_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/membrane/whole-mouse-brain/macs-v11-norestart-acq-random/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..fc9ecdaa46802545a321cc83509d69692a69a151 --- /dev/null +++ b/macs_v2_r26/runs/membrane/whole-mouse-brain/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 +1,49,112.9,610.4,"[0.4308437371432156, 0.3720464899554072, 0.13400588777866665, 0.2906832527572902, 0.28441244487328066, 0.3714445937083196, 0.46009083351660035, 0.2953659731768761]",,774194,43,True,kept,0.3141552482233243,0.46009083351660035,"[0.027388975830910146, 0.511103822265685, 0.0, 0.0, 0.0, 0.0, 0.46150720190340494, 0.0]",top3,"{'mixture': 0.4149047036603486, 'uniform': 0.3931444911550303, 'top1': 0.46009083351660035, 'top2': 0.46380715367111625, 'top3': 0.49022779815240064}",0.6512893174377166,0.4646900588599244,0.0896417007453064,,,,, +2,98,0.2,63.2,"[0.4424330801851372, 0.37408893009446964, 0.13996203868108806, 0.29503398267841263, 0.2762732338980843, 0.3782590298845219, 0.47481626651490355, 0.3238189071307628]",,1631452,85,False,,,,"[0.0543864996701088, 0.0, 0.0, 0.0, 0.0, 0.0, 0.9456135003298912, 0.0]",top2,"{'mixture': 0.4356210342800381, 'uniform': 0.40298612721687643, 'top1': 0.47481626651490355, 'top2': 0.47797929634414377, 'top3': 0.4516145396104275}",0.6899674120835443,0.3934530539606663,0.046598598690968274,0.8275029820365678,0.49077896830410456,student,1.0, +3,147,0.2,386.1,"[0.4208794202637337, 0.37488266988173913, 0.150732512981738, 0.280229848328845, 0.30736600511815093, 0.42137673253425356, 0.504630320454826, 0.3103878885056744]","[0.3964222900856918, 0.37934827125456433, 0.15986144969132293, 0.2785786219247478, 0.33299853995204853, 0.45638475100222214, 0.5447891954337357, 0.2891779812953886]",2470847,129,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.46107124265136823, 'uniform': 0.4470064837241954, 'top1': 0.504630320454826, 'top2': 0.4774312960056267, 'top3': 0.47954494564006567}",0.6997838980735469,0.38111564948839316,0.006551502400643054,0.8483273711237067,0.4447035907635971,student,1.0, +4,196,0.2,405.8,"[0.41735012268474536, 0.37774968729190495, 0.15129542306014881, 0.2833705136853922, 0.3061931648198789, 0.4232066747177724, 0.49688937580258463, 0.314315337148227]","[0.3943931472745527, 0.37855105448483334, 0.15943182820909615, 0.2892929128772673, 0.3257258256062749, 0.4601414870055501, 0.5342473085263869, 0.2950977113256563]",3255460,173,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.45595241469488296, 'uniform': 0.44581060629678865, 'top1': 0.49688937580258463, 'top2': 0.46677430571359046, 'top3': 0.4690154172263659}",0.7284449934265951,0.35062932040639544,0.01061670729460243,0.8894096619935941,0.4435759384979648,student,1.0, +5,245,0.2,427.2,"[0.40692426871897414, 0.3776928989394102, 0.14920549941794428, 0.2795811790533572, 0.31140825070370776, 0.43368676727852196, 0.5091293274911578, 0.3121180397885524]","[0.3970244948724663, 0.37928996723211816, 0.15356131391556158, 0.28419405601396547, 0.3296259383047535, 0.46253273706373266, 0.542636728401153, 0.3050203065494246]",3796031,213,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.4624701194922756, 'uniform': 0.45307112409178935, 'top1': 0.5091293274911578, 'top2': 0.4770389469575239, 'top3': 0.4788482169595161}",0.7360976720287663,0.3467971465172063,0.01788726699061594,0.8150319479999724,0.5618041860169222,student,1.0, +6,294,0.2,448.7,"[0.4131523804861451, 0.38148291144188595, 0.14959132590158677, 0.2859449485429898, 0.3070357015540196, 0.4312325733658978, 0.5087118932311401, 0.321087704901554]","[0.4114461558819054, 0.3849032866661021, 0.1555940064347239, 0.2883347744854101, 0.32091747898050904, 0.4585541603506631, 0.5398959859343434, 0.31990600159271626]",4607078,257,False,,,,"[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0]",top1,"{'mixture': 0.46080784643360345, 'uniform': 0.4478514590552423, 'top1': 0.5087118932311401, 'top2': 0.47603001745833207, 'top3': 0.47780834110988624}",0.7405788014844665,0.31581319789840717,0.019612879670904602,0.8012167018453978,0.4863777827338713,student,1.0, +7,294,0.0,189.0,,,,,,,,,,,,0.7512189727699001,0.3181009168739638,0.021623735484845367,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..35a7c1be4b794b94ae03a220ac9db79e45396d25 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt @@ -0,0 +1,105 @@ +[23:08:10][ 0s][START] target=drosophila-vnc variant=macs-v11-norestart-acq-laplace budget=5.0% seed=1 | {"task": "mitochondria", "target": "drosophila-vnc", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[23:08:10][ 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. +[23:08:11][ 0s][DATA] drosophila-vnc: train=686 val=147 test=147 patches +[23:08:20][ 10s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, multiclass=2.60 +[23:08:27][ 17s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 7s +[23:08:28][ 18s][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 +[23:08:28][ 18s][WEIGHTS] per-source: fly-mito=0.039, kasthuri=0.943, lucchi=0.014, multiclass=0.004 | min-MMD source (NeuroADDA rule) = kasthuri +[23:08:35][ 24s][TEACHERS] target-val Dice: fly-mito=0.278, kasthuri=0.183, lucchi=0.058, multiclass=0.017 +[23:08:35][ 24s][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 +[23:08:35][ 24s][WEIGHTS] using mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020 +[23:08:35][ 24s][INIT] student initialised from teacher kasthuri (weight 0.838) +[23:08:38][ 27s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[23:08:38][ 27s][KD] distilling the mixture ensemble (prob space) into the student +[23:08:51][ 41s][KD] soft labels for 686 patches from 4 teachers in 14s +[23:09:12][ 61s][KD] epoch 1/5 | patches=686 | loss=0.2005 | 20s +[23:09:32][ 81s][KD] epoch 2/5 | patches=686 | loss=0.1322 | 20s +[23:09:52][ 101s][KD] epoch 3/5 | patches=686 | loss=0.1259 | 20s +[23:10:12][ 121s][KD] epoch 4/5 | patches=686 | loss=0.1216 | 20s +[23:10:32][ 142s][KD] epoch 5/5 | patches=686 | loss=0.1204 | 20s +[23:10:33][ 143s][ROUND0] before any target label: val_dice=0.1919 val_vi=0.2213 +[23:10:33][ 143s][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 +[23:10:40][ 150s][AL] round-1 uncertainty quality on 686 patches: {'spearman_uncertainty_error': 0.1281956487029202, 'auroc_top20_error': 0.6150618810253443} +[23:10:54][ 164s][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 +[23:10:55][ 165s][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 +[23:10:55][ 165s][WEIGHTS] round 1: student kept (labeled Dice 0.123 vs best teacher fly-mito 0.312); distilling the corrected ensemble into it +[23:10:55][ 165s][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 +[23:10:59][ 169s][KD] soft labels for 681 patches from 4 teachers in 4s +[23:11:19][ 189s][KD] epoch 1/2 | patches=681 | loss=0.2934 | 20s +[23:11:39][ 209s][KD] epoch 2/2 | patches=681 | loss=0.2402 | 20s +[23:11:43][ 212s][CONS] pseudo-labels for 681 unlabeled patches from the current ensemble in 4s +[23:11:43][ 213s][FT] round 1/6 epoch 1/5 | patches=5 | loss=1.4014 +[23:11:44][ 214s][FT] round 1/6 epoch 2/5 | patches=5 | loss=1.3138 +[23:11:45][ 215s][FT] round 1/6 epoch 3/5 | patches=5 | loss=1.2332 +[23:11:46][ 216s][FT] round 1/6 epoch 4/5 | patches=5 | loss=1.1817 +[23:11:47][ 217s][FT] round 1/6 epoch 5/5 | patches=5 | loss=1.2248 +[23:11:48][ 218s][AL] round 1/6 | labeled=5 | selection_seconds=21.2000 round_seconds=74.2000 restart_student_dice=0.1231 restart_teacher_dice=0.3119 val_dice=0.2369 val_vi=0.2879 val_ece=0.0156 +[23:11:55][ 225s][AL] round 2: 14/681 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:11:57][ 226s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.141, uniform=0.025, top1=0.172, top2=0.142, top3=0.142 -> using top1 +[23:11:57][ 226s][WEIGHTS] round 2: teacher Dice on 10 labeled patches + 0 blank (17002 fg px): fly-mito=0.172, kasthuri=0.135, lucchi=0.001, multiclass=0.003 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[23:11:58][ 227s][CONS] round 2: on the 5 new patches student Dice=0.042 vs ensemble 0.040 -> pseudo-labels from the ensemble | consistency x1.00 | judgement set below the foreground gate: source kept +[23:11:59][ 229s][FT] round 2/6 epoch 1/5 | patches=10 | loss=1.1973 +[23:12:00][ 230s][FT] round 2/6 epoch 2/5 | patches=10 | loss=1.1801 +[23:12:01][ 231s][FT] round 2/6 epoch 3/5 | patches=10 | loss=1.1914 +[23:12:03][ 232s][FT] round 2/6 epoch 4/5 | patches=10 | loss=1.1851 +[23:12:04][ 234s][FT] round 2/6 epoch 5/5 | patches=10 | loss=1.1409 +[23:12:05][ 235s][AL] round 2/6 | labeled=10 | selection_seconds=7.4000 round_seconds=16.0000 pseudo_student_dice=0.0421 pseudo_ensemble_dice=0.0401 consistency_scale=1.0000 val_dice=0.2866 val_vi=0.2713 val_ece=0.0474 +[23:12:12][ 242s][AL] round 3: 31/676 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:12:14][ 244s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.146, uniform=0.018, top1=0.182, top2=0.147, top3=0.146 -> using top1 +[23:12:14][ 244s][WEIGHTS] round 3: teacher Dice on 15 labeled patches + 0 blank (24575 fg px): fly-mito=0.182, kasthuri=0.137, lucchi=0.001, multiclass=0.002 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[23:12:15][ 245s][CONS] round 3: on the 5 new patches student Dice=0.203 vs ensemble 0.197 -> pseudo-labels from the ensemble | consistency x1.00 +[23:12:16][ 246s][FT] round 3/6 epoch 1/5 | patches=15 | loss=1.1297 +[23:12:18][ 248s][FT] round 3/6 epoch 2/5 | patches=15 | loss=1.1117 +[23:12:19][ 249s][FT] round 3/6 epoch 3/5 | patches=15 | loss=1.0964 +[23:12:21][ 250s][FT] round 3/6 epoch 4/5 | patches=15 | loss=1.0400 +[23:12:22][ 252s][FT] round 3/6 epoch 5/5 | patches=15 | loss=1.0601 +[23:12:23][ 253s][AL] round 3/6 | labeled=15 | selection_seconds=7.3000 round_seconds=17.0000 pseudo_student_dice=0.2030 pseudo_ensemble_dice=0.1971 consistency_scale=1.0000 val_dice=0.4354 val_vi=0.3016 val_ece=0.0098 +[23:12:30][ 260s][AL] round 4: 15/671 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:12:33][ 262s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.128, uniform=0.017, top1=0.147, top2=0.129, top3=0.129 -> using top1 +[23:12:33][ 262s][WEIGHTS] round 4: teacher Dice on 20 labeled patches + 0 blank (25508 fg px): fly-mito=0.147, kasthuri=0.122, lucchi=0.001, multiclass=0.002 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[23:12:33][ 263s][CONS] round 4: on the 5 new patches student Dice=0.026 vs ensemble 0.011 -> pseudo-labels from the ensemble | consistency x1.00 | judgement set below the foreground gate: source kept +[23:12:35][ 265s][FT] round 4/6 epoch 1/5 | patches=20 | loss=1.0636 +[23:12:37][ 267s][FT] round 4/6 epoch 2/5 | patches=20 | loss=1.0478 +[23:12:39][ 268s][FT] round 4/6 epoch 3/5 | patches=20 | loss=1.0256 +[23:12:41][ 270s][FT] round 4/6 epoch 4/5 | patches=20 | loss=1.0184 +[23:12:42][ 272s][FT] round 4/6 epoch 5/5 | patches=20 | loss=1.0131 +[23:12:43][ 273s][AL] round 4/6 | labeled=20 | selection_seconds=7.3000 round_seconds=19.3000 pseudo_student_dice=0.0263 pseudo_ensemble_dice=0.0112 consistency_scale=1.0000 val_dice=0.5388 val_vi=0.1580 val_ece=0.0072 +[23:12:51][ 280s][AL] round 5: 99/666 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:12:53][ 283s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.117, uniform=0.011, top1=0.150, top2=0.118, top3=0.117 -> using top1 +[23:12:53][ 283s][WEIGHTS] round 5: teacher Dice on 25 labeled patches + 0 blank (41206 fg px): fly-mito=0.150, kasthuri=0.112, lucchi=0.001, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[23:12:58][ 287s][CONS] round 5: on the 5 new patches student Dice=0.588 vs ensemble 0.157 -> pseudo-labels from the student | consistency x0.29 +[23:13:00][ 290s][FT] round 5/6 epoch 1/5 | patches=25 | loss=0.7865 +[23:13:02][ 292s][FT] round 5/6 epoch 2/5 | patches=25 | loss=0.7621 +[23:13:04][ 294s][FT] round 5/6 epoch 3/5 | patches=25 | loss=0.7242 +[23:13:06][ 296s][FT] round 5/6 epoch 4/5 | patches=25 | loss=0.6684 +[23:13:09][ 298s][FT] round 5/6 epoch 5/5 | patches=25 | loss=0.6240 +[23:13:10][ 299s][AL] round 5/6 | labeled=25 | selection_seconds=7.2000 round_seconds=25.2000 pseudo_student_dice=0.5880 pseudo_ensemble_dice=0.1565 consistency_scale=0.2930 val_dice=0.6531 val_vi=0.1316 val_ece=0.0065 +[23:13:17][ 307s][AL] round 6: 116/661 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:13:45][ 335s][WEIGHTS] round 6: teacher agreement with the student on 656 unlabeled patches: fly-mito=0.387, kasthuri=0.467, lucchi=0.009, multiclass=0.001 | blended with labeled Dice at 0.50 +[23:14:11][ 360s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.284, uniform=0.034, top1=0.288, top2=0.284, top3=0.284 -> using mixture +[23:14:11][ 360s][WEIGHTS] round 6: teacher Dice on 30 labeled patches + 0 blank (67637 fg px): fly-mito=0.288, kasthuri=0.281, lucchi=0.005, multiclass=0.001 | weights: fly-mito=0.130, kasthuri=0.867, lucchi=0.002, multiclass=0.001 +[23:14:15][ 365s][CONS] round 6: on the 5 new patches student Dice=0.411 vs ensemble 0.063 -> pseudo-labels from the student | consistency x0.00 +[23:14:16][ 366s][FT] round 6/6 epoch 1/5 | patches=30 | loss=0.6123 +[23:14:17][ 367s][FT] round 6/6 epoch 2/5 | patches=30 | loss=0.5794 +[23:14:19][ 368s][FT] round 6/6 epoch 3/5 | patches=30 | loss=0.5905 +[23:14:20][ 370s][FT] round 6/6 epoch 4/5 | patches=30 | loss=0.5808 +[23:14:21][ 371s][FT] round 6/6 epoch 5/5 | patches=30 | loss=0.5486 +[23:14:22][ 372s][AL] round 6/6 | labeled=30 | selection_seconds=7.2000 round_seconds=71.4000 pseudo_student_dice=0.4114 pseudo_ensemble_dice=0.0634 consistency_scale=0.0000 val_dice=0.6087 val_vi=0.1785 val_ece=0.0197 +[23:14:26][ 375s][CONS] final consolidation: pseudo-labels refreshed from the student +[23:14:27][ 377s][FT] final 1/10 epoch 1/1 | patches=30 | loss=0.5766 +[23:14:28][ 378s][FT] final 2/10 epoch 1/1 | patches=30 | loss=0.4982 +[23:14:29][ 379s][FT] final 3/10 epoch 1/1 | patches=30 | loss=0.5141 +[23:14:31][ 380s][FT] final 4/10 epoch 1/1 | patches=30 | loss=0.5012 +[23:14:32][ 381s][FT] final 5/10 epoch 1/1 | patches=30 | loss=0.4460 +[23:14:33][ 383s][FT] final 6/10 epoch 1/1 | patches=30 | loss=0.4272 +[23:14:34][ 384s][FT] final 7/10 epoch 1/1 | patches=30 | loss=0.4768 +[23:14:35][ 385s][FT] final 8/10 epoch 1/1 | patches=30 | loss=0.4484 +[23:14:37][ 386s][FT] final 9/10 epoch 1/1 | patches=30 | loss=0.4726 +[23:14:38][ 387s][FT] final 10/10 epoch 1/1 | patches=30 | loss=0.4304 +[23:14:39][ 389s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=15.6000 val_dice=0.6910 val_vi=0.1376 val_ece=0.0151 +[23:14:39][ 389s][WEIGHTS] final: fly-mito=0.130, kasthuri=0.867, lucchi=0.002, multiclass=0.001 +[23:14:46][ 396s][TEST] raw student (no EMA, no TTA): patch_dice=0.7909 patch_vi=0.1394 +[23:14:47][ 397s][TEST] final model without TTA: patch_dice=0.7909 patch_vi=0.1394 +[23:14:47][ 397s][RESOURCES] {"phases": {"setup": {"seconds": 8.5, "peak_cpu_rss_mb": 2190.7, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 17.69, "peak_cpu_rss_mb": 4161.9, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 114.25, "peak_cpu_rss_mb": 4369.3, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4168.0, "kd_epochs": 5}, "active_learning": {"seconds": 246.0, "peak_cpu_rss_mb": 4472.1, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4178.0, "rounds": 6}, "inference": {"seconds": 8.46, "peak_cpu_rss_mb": 4472.1, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4178.0, "patches": 147, "images": 3, "patches_per_second": 17.4, "seconds_per_image": 2.82}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 4472.1, "total_seconds": 394.9} +[23:14:47][ 397s][TEST] patch_dice=0.7877 patch_iou=0.6951 patch_precision=0.7281 patch_recall=0.9348 patch_vi=0.1348 image_dice=0.8425 image_iou=0.7285 image_precision=0.7509 image_recall=0.9602 image_vi=0.1642 cal_ece=0.0075 cal_nll=0.0525 cal_brier=0.0154 cal_mean_confidence=0.9862 cal_accuracy=0.9787 cal_predicted_prevalence=0.0745 cal_true_prevalence=0.0594 cal_mean_probability=0.0762 +[23:14:48][ 397s][DONE] drosophila-vnc macs-v11-norestart-acq-laplace b5 s1 | test patch Dice=0.7877 VI=0.1348 | image Dice=0.8425 VI=0.1642 | 397s diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..abdfccd063353c4c1ff12978477c7c1e010deb3e --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json @@ -0,0 +1,652 @@ +{ + "target": "drosophila-vnc", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "config": { + "task": "mitochondria", + "target": "drosophila-vnc", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "fce9b15", + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "drosophila-vnc", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "fce9b15", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "multiclass" + ], + "alphas": [ + 0.11690492516893997, + 0.8375897330633602, + 0.0256027087898516, + 0.019902632977848233 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "multiclass" + ], + "mixture": [ + 0.11690492516893997, + 0.8375897330633602, + 0.0256027087898516, + 0.019902632977848233 + ], + "per_source": [ + 0.038599207620062845, + 0.9432861284385984, + 0.013712296406123614, + 0.0044023675352152245 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.1917820351790263, + "mmd2_mixture": 0.19749363071106324, + "mmd2_uniform": 0.3743489659922599, + "mmd2_single_source": [ + 0.7042977370508033, + 0.20706860465273969, + 0.6307956892595754, + 0.6534362949819378 + ], + "median_squared_distance": 0.19645890982127545, + "gammas": [ + 1.272530730356961, + 2.545061460713922, + 5.090122921427844, + 10.180245842855689, + 20.360491685711377 + ], + "max_weight": 0.8375897330633602, + "entropy": 0.5711595532036933, + "effective_sources": 1.7703186403916105, + "min_weight": 0.019902632977848233 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.605076346545893, + 0.31101778520145595, + 0.7002952057460718, + 0.8048259839560096 + ], + "standardized": [ + -0.0012362667606482206, + -1.5993052227157551, + 0.5162331182950395, + 1.0843083711813646 + ], + "tau": 0.5, + "max_weight": 0.9432861284385984, + "entropy": 0.26340029394025066, + "effective_sources": 1.301347536237684, + "min_weight": 0.0044023675352152245 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "kasthuri-pp", + "seconds": 8.5 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.27840715557699913, + 0.18341499813983803, + 0.0577032083417681, + 0.01681466600081078 + ], + "ensemble_val_dice": { + "mixture": 0.18652277313599772, + "mixture@raw": 0.18548230728775292, + "mixture@logit": 0.19296074616265912, + "per_source": 0.18415443581907806, + "per_source@raw": 0.1835051571090219, + "per_source@logit": 0.18905485087217583, + "uniform": 0.0696943056822927, + "uniform@raw": 0.08847159266818436, + "uniform@logit": 0.06549734203813108 + }, + "val_patches": 147 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 3.80090069770813, + 2.597146987915039 + ], + "top_source": "kasthuri-pp", + "weight_stats": { + "max_weight": 0.8375897330633602, + "entropy": 0.5711595532036933, + "effective_sources": 1.7703186403916105, + "min_weight": 0.019902632977848233 + }, + "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": 114.3, + "round0_val": { + "val_dice": 0.19189352403793783, + "val_vi": 0.22130833251687648, + "val_ece": 0.02232954733842891 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 5, + "selection_seconds": 21.2, + "round_seconds": 74.2, + "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": 0.12307992202729044, + "restart_teacher_dice": 0.3118739095489208, + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.30330376940133036, + "uniform": 0.018090264826557255, + "top1": 0.3118739095489208, + "top2": 0.30483917067915484, + "top3": 0.3042653191818001 + }, + "val_dice": 0.23685093051502523, + "val_vi": 0.28794167838117374, + "val_ece": 0.015598701639074236 + }, + { + "round": 2, + "n_labeled": 10, + "selection_seconds": 7.4, + "round_seconds": 16.0, + "teacher_dice_labeled": [ + 0.17236378900771807, + 0.134718532257256, + 0.0011034322550670772, + 0.0025857610756766076 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 17002, + "evidence_nonempty_patches": 7, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.14129939786469667, + "uniform": 0.025314333612740988, + "top1": 0.17236378900771807, + "top2": 0.14178256808945655, + "top3": 0.14157927984838914 + }, + "pseudo_student_dice": 0.042085187539733, + "pseudo_ensemble_dice": 0.04012414510853405, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.2866201808181058, + "val_vi": 0.27125518727925535, + "val_ece": 0.04738608310345371 + }, + { + "round": 3, + "n_labeled": 15, + "selection_seconds": 7.3, + "round_seconds": 17.0, + "teacher_dice_labeled": [ + 0.1816465346770543, + 0.13722353870458137, + 0.0007644952319639481, + 0.0017971963736570949 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 24575, + "evidence_nonempty_patches": 12, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.14601875873079226, + "uniform": 0.017784931883318283, + "top1": 0.1816465346770543, + "top2": 0.14662073360149747, + "top3": 0.14634583948008453 + }, + "pseudo_student_dice": 0.2029553264604811, + "pseudo_ensemble_dice": 0.1971301728238465, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.43543202231132283, + "val_vi": 0.30158674255549045, + "val_ece": 0.009805639594878747 + }, + { + "round": 4, + "n_labeled": 20, + "selection_seconds": 7.3, + "round_seconds": 19.3, + "teacher_dice_labeled": [ + 0.14676982261976956, + 0.12248808874980222, + 0.0007346686257830021, + 0.002188183807439825 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 25508, + "evidence_nonempty_patches": 15, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.12848106904231626, + "uniform": 0.017156491440690803, + "top1": 0.14676982261976956, + "top2": 0.12918176006818666, + "top3": 0.1288194073915982 + }, + "pseudo_student_dice": 0.02631578947368421, + "pseudo_ensemble_dice": 0.01118521606224816, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.538829033552655, + "val_vi": 0.15799834303392746, + "val_ece": 0.007239701747245221 + }, + { + "round": 5, + "n_labeled": 25, + "selection_seconds": 7.2, + "round_seconds": 25.2, + "teacher_dice_labeled": [ + 0.14991183354172677, + 0.11196244467753076, + 0.0005527650267970872, + 0.0013616492678149112 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 41206, + "evidence_nonempty_patches": 20, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.11684489229345535, + "uniform": 0.011136968085106383, + "top1": 0.14991183354172677, + "top2": 0.11754230882560181, + "top3": 0.11732180427977626 + }, + "pseudo_student_dice": 0.5880473899154907, + "pseudo_ensemble_dice": 0.15650009421145017, + "pseudo_source": "student", + "consistency_scale": 0.293, + "val_dice": 0.6531419798130339, + "val_vi": 0.13160857087554498, + "val_ece": 0.006511332969372438 + }, + { + "round": 6, + "n_labeled": 30, + "selection_seconds": 7.2, + "round_seconds": 71.4, + "teacher_dice_labeled": [ + 0.28795011474485044, + 0.280563902813102, + 0.0054258056716156265, + 0.0008719495006685415 + ], + "teacher_dice_pool": [ + 0.3870062594431254, + 0.4669729462302163, + 0.009022441201141316, + 0.0009105291166930196 + ], + "evidence_fg_pixels": 67637, + "evidence_nonempty_patches": 24, + "first_correction": false, + "restart": "", + "weights": [ + 0.13025564297709144, + 0.8667965482101361, + 0.0016914617097016046, + 0.0012563471030710067 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.2837908050450364, + "uniform": 0.034392525491845516, + "top1": 0.28795011474485044, + "top2": 0.2838905935041172, + "top3": 0.2838305958183957 + }, + "pseudo_student_dice": 0.41143576121906894, + "pseudo_ensemble_dice": 0.06339725733876152, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.608747389325827, + "val_vi": 0.17846436977004534, + "val_ece": 0.01971239731023785 + }, + { + "round": 7, + "n_labeled": 30, + "selection_seconds": 0.0, + "round_seconds": 15.6, + "final_consolidation": true, + "val_dice": 0.6909546100135161, + "val_vi": 0.13761017851695848, + "val_ece": 0.015077464841051817 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.1281956487029202, + "auroc_top20_error": 0.6150618810253443 + }, + "al_seconds": 246.0, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.13025564297709144, + 0.8667965482101361, + 0.0016914617097016046, + 0.0012563471030710067 + ], + "weight_stats_final": { + "max_weight": 0.8667965482101361, + "entropy": 0.4085908200891099, + "effective_sources": 1.5046959031685392, + "min_weight": 0.0012563471030710067 + }, + "top_source_final": "kasthuri-pp", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.79094270337799, + "patch_iou": 0.6974377079183249, + "patch_precision": 0.7248810371901514, + "patch_recall": 0.9449957914805975, + "patch_vi": 0.13943115167621045, + "n_patches": 147, + "cal_ece": 0.014094231018419017, + "cal_nll": 0.061826090976241234, + "cal_brier": 0.016988086156188117, + "cal_mean_confidence": 0.9927206806975483, + "cal_accuracy": 0.9786484906462585, + "cal_predicted_prevalence": 0.07579985119047619, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.0756547042173509 + }, + "test_notta": { + "patch_dice": 0.79094270337799, + "patch_iou": 0.6974377079183249, + "patch_precision": 0.7248810371901514, + "patch_recall": 0.9449957914805975, + "patch_vi": 0.13943115167621045, + "n_patches": 147, + "cal_ece": 0.014094231018419017, + "cal_nll": 0.061826090976241234, + "cal_brier": 0.016988086156188117, + "cal_mean_confidence": 0.9927206806975483, + "cal_accuracy": 0.9786484906462585, + "cal_predicted_prevalence": 0.07579985119047619, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.0756547042173509 + }, + "test": { + "patch_dice": 0.7877385379966639, + "patch_iou": 0.6951161067329877, + "patch_precision": 0.7280763728942042, + "patch_recall": 0.9348244922204363, + "patch_vi": 0.1348154343767868, + "n_patches": 147, + "image_dice": 0.8425339902518142, + "image_iou": 0.728532153833435, + "image_precision": 0.7508811622248409, + "image_recall": 0.9602402210801045, + "image_vi": 0.1642342012294927, + "n_images": 3, + "cal_ece": 0.00746777378342397, + "cal_nll": 0.05250644355455002, + "cal_brier": 0.015442382133295839, + "cal_mean_confidence": 0.9861893405351245, + "cal_accuracy": 0.9787215667517006, + "cal_predicted_prevalence": 0.07450441113945579, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.07621637154438468 + }, + "eval_seconds": 8.5, + "total_seconds": 397.0, + "resources": { + "phases": { + "setup": { + "seconds": 8.5, + "peak_cpu_rss_mb": 2190.7, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 17.69, + "peak_cpu_rss_mb": 4161.9, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 114.25, + "peak_cpu_rss_mb": 4369.3, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4168.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 246.0, + "peak_cpu_rss_mb": 4472.1, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4178.0, + "rounds": 6 + }, + "inference": { + "seconds": 8.46, + "peak_cpu_rss_mb": 4472.1, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4178.0, + "patches": 147, + "images": 3, + "patches_per_second": 17.4, + "seconds_per_image": 2.82 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 4472.1, + "total_seconds": 394.9 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/drosophila-vnc/macs-v11-norestart-acq-laplace_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..3fe0edd5b4f29b0c67db414abd2ee41743996376 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/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 +1,5,21.2,74.2,"[0.3118739095489208, 0.0968435913812229, 0.0004512053628980276, 0.00288]",,15368,5,True,kept,0.12307992202729044,0.3118739095489208,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.30330376940133036, 'uniform': 0.018090264826557255, 'top1': 0.3118739095489208, 'top2': 0.30483917067915484, 'top3': 0.3042653191818001}",0.23685093051502523,0.28794167838117374,0.015598701639074236,,,,, +2,10,7.4,16.0,"[0.17236378900771807, 0.134718532257256, 0.0011034322550670772, 0.0025857610756766076]",,17002,7,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.14129939786469667, 'uniform': 0.025314333612740988, 'top1': 0.17236378900771807, 'top2': 0.14178256808945655, 'top3': 0.14157927984838914}",0.2866201808181058,0.27125518727925535,0.04738608310345371,0.042085187539733,0.04012414510853405,ensemble,1.0, +3,15,7.3,17.0,"[0.1816465346770543, 0.13722353870458137, 0.0007644952319639481, 0.0017971963736570949]",,24575,12,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.14601875873079226, 'uniform': 0.017784931883318283, 'top1': 0.1816465346770543, 'top2': 0.14662073360149747, 'top3': 0.14634583948008453}",0.43543202231132283,0.30158674255549045,0.009805639594878747,0.2029553264604811,0.1971301728238465,ensemble,1.0, +4,20,7.3,19.3,"[0.14676982261976956, 0.12248808874980222, 0.0007346686257830021, 0.002188183807439825]",,25508,15,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.12848106904231626, 'uniform': 0.017156491440690803, 'top1': 0.14676982261976956, 'top2': 0.12918176006818666, 'top3': 0.1288194073915982}",0.538829033552655,0.15799834303392746,0.007239701747245221,0.02631578947368421,0.01118521606224816,ensemble,1.0, +5,25,7.2,25.2,"[0.14991183354172677, 0.11196244467753076, 0.0005527650267970872, 0.0013616492678149112]",,41206,20,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.11684489229345535, 'uniform': 0.011136968085106383, 'top1': 0.14991183354172677, 'top2': 0.11754230882560181, 'top3': 0.11732180427977626}",0.6531419798130339,0.13160857087554498,0.006511332969372438,0.5880473899154907,0.15650009421145017,student,0.293, +6,30,7.2,71.4,"[0.28795011474485044, 0.280563902813102, 0.0054258056716156265, 0.0008719495006685415]","[0.3870062594431254, 0.4669729462302163, 0.009022441201141316, 0.0009105291166930196]",67637,24,False,,,,"[0.13025564297709144, 0.8667965482101361, 0.0016914617097016046, 0.0012563471030710067]",mixture,"{'mixture': 0.2837908050450364, 'uniform': 0.034392525491845516, 'top1': 0.28795011474485044, 'top2': 0.2838905935041172, 'top3': 0.2838305958183957}",0.608747389325827,0.17846436977004534,0.01971239731023785,0.41143576121906894,0.06339725733876152,student,0.0, +7,30,0.0,15.6,,,,,,,,,,,,0.6909546100135161,0.13761017851695848,0.015077464841051817,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..90fe83443c6b52cbaac3b15691eb156e29df561e --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt @@ -0,0 +1,105 @@ +[12:09:17][ 0s][START] target=drosophila-vnc variant=macs-v11-norestart-acq-laplace budget=5.0% seed=2 | {"task": "mitochondria", "target": "drosophila-vnc", "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": 8, "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} +[12:09:17][ 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. +[12:09:17][ 0s][DATA] drosophila-vnc: train=686 val=147 test=147 patches +[12:09:18][ 2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, multiclass=2.60 +[12:09:19][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[12:09:20][ 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 +[12:09:20][ 3s][WEIGHTS] per-source: fly-mito=0.039, kasthuri=0.943, lucchi=0.014, multiclass=0.004 | min-MMD source (NeuroADDA rule) = kasthuri +[12:09:26][ 10s][TEACHERS] target-val Dice: fly-mito=0.278, kasthuri=0.183, lucchi=0.058, multiclass=0.017 +[12:09:26][ 10s][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 +[12:09:26][ 10s][WEIGHTS] using mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020 +[12:09:26][ 10s][INIT] student initialised from teacher kasthuri (weight 0.838) +[12:09:30][ 13s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[12:09:30][ 13s][KD] distilling the mixture ensemble (prob space) into the student +[12:09:44][ 27s][KD] soft labels for 686 patches from 4 teachers in 14s +[12:10:05][ 48s][KD] epoch 1/5 | patches=686 | loss=0.1999 | 21s +[12:10:25][ 69s][KD] epoch 2/5 | patches=686 | loss=0.1335 | 21s +[12:10:46][ 89s][KD] epoch 3/5 | patches=686 | loss=0.1266 | 21s +[12:11:07][ 110s][KD] epoch 4/5 | patches=686 | loss=0.1231 | 21s +[12:11:28][ 131s][KD] epoch 5/5 | patches=686 | loss=0.1203 | 21s +[12:11:29][ 132s][ROUND0] before any target label: val_dice=0.2059 val_vi=0.2207 +[12:11:29][ 132s][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 +[12:11:36][ 140s][AL] round-1 uncertainty quality on 686 patches: {'spearman_uncertainty_error': 0.126648544840516, 'auroc_top20_error': 0.6094229345181423} +[12:11:51][ 154s][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 +[12:11:52][ 155s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.531, uniform=0.067, top1=0.531, top2=0.531, top3=0.531 -> using mixture +[12:11:52][ 156s][WEIGHTS] round 1: student kept (labeled Dice 0.583 vs best teacher kasthuri 0.531); distilling the corrected ensemble into it +[12:11:52][ 156s][WEIGHTS] round 1: teacher Dice on 5 labeled patches + 0 blank (19579 fg px): fly-mito=0.363, kasthuri=0.531, lucchi=0.014, multiclass=0.002 | weights: fly-mito=0.025, kasthuri=0.974, lucchi=0.000, multiclass=0.000 +[12:12:06][ 170s][KD] soft labels for 681 patches from 4 teachers in 14s +[12:12:27][ 190s][KD] epoch 1/2 | patches=681 | loss=0.0886 | 21s +[12:12:48][ 211s][KD] epoch 2/2 | patches=681 | loss=0.0859 | 21s +[12:13:02][ 225s][CONS] pseudo-labels for 681 unlabeled patches from the current ensemble in 14s +[12:13:03][ 226s][FT] round 1/6 epoch 1/5 | patches=5 | loss=0.9031 +[12:13:04][ 227s][FT] round 1/6 epoch 2/5 | patches=5 | loss=0.8632 +[12:13:05][ 228s][FT] round 1/6 epoch 3/5 | patches=5 | loss=0.8447 +[12:13:06][ 229s][FT] round 1/6 epoch 4/5 | patches=5 | loss=0.8470 +[12:13:07][ 230s][FT] round 1/6 epoch 5/5 | patches=5 | loss=0.8177 +[12:13:08][ 232s][AL] round 1/6 | labeled=5 | selection_seconds=22.1000 round_seconds=98.5000 restart_student_dice=0.5832 restart_teacher_dice=0.5310 val_dice=0.1982 val_vi=0.2106 val_ece=0.0303 +[12:13:16][ 239s][AL] round 2: 242/681 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:13:18][ 241s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.363, uniform=0.033, top1=0.364, top2=0.363, top3=0.363 -> using mixture +[12:13:18][ 241s][WEIGHTS] round 2: teacher Dice on 10 labeled patches + 0 blank (46033 fg px): fly-mito=0.334, kasthuri=0.364, lucchi=0.006, multiclass=0.001 | weights: fly-mito=0.094, kasthuri=0.905, lucchi=0.001, multiclass=0.001 +[12:13:19][ 242s][CONS] round 2: on the 5 new patches student Dice=0.153 vs ensemble 0.218 -> pseudo-labels from the ensemble | consistency x1.00 +[12:13:20][ 243s][FT] round 2/6 epoch 1/5 | patches=10 | loss=0.9990 +[12:13:21][ 245s][FT] round 2/6 epoch 2/5 | patches=10 | loss=0.9318 +[12:13:23][ 246s][FT] round 2/6 epoch 3/5 | patches=10 | loss=0.8588 +[12:13:24][ 248s][FT] round 2/6 epoch 4/5 | patches=10 | loss=0.7714 +[12:13:26][ 249s][FT] round 2/6 epoch 5/5 | patches=10 | loss=0.6310 +[12:13:27][ 250s][AL] round 2/6 | labeled=10 | selection_seconds=7.6000 round_seconds=17.6000 pseudo_student_dice=0.1526 pseudo_ensemble_dice=0.2182 consistency_scale=1.0000 val_dice=0.4146 val_vi=0.1946 val_ece=0.0208 +[12:13:35][ 258s][AL] round 3: 141/676 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:13:37][ 260s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.242, uniform=0.021, top1=0.334, top2=0.243, top3=0.243 -> using top1 +[12:13:37][ 260s][WEIGHTS] round 3: teacher Dice on 15 labeled patches + 0 blank (71466 fg px): fly-mito=0.334, kasthuri=0.243, lucchi=0.005, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[12:13:38][ 261s][CONS] round 3: on the 5 new patches student Dice=0.270 vs ensemble 0.334 -> pseudo-labels from the ensemble | consistency x1.00 +[12:13:39][ 262s][FT] round 3/6 epoch 1/5 | patches=15 | loss=0.9433 +[12:13:41][ 264s][FT] round 3/6 epoch 2/5 | patches=15 | loss=0.8414 +[12:13:42][ 266s][FT] round 3/6 epoch 3/5 | patches=15 | loss=0.6321 +[12:13:44][ 267s][FT] round 3/6 epoch 4/5 | patches=15 | loss=0.5810 +[12:13:45][ 269s][FT] round 3/6 epoch 5/5 | patches=15 | loss=0.5354 +[12:13:47][ 270s][AL] round 3/6 | labeled=15 | selection_seconds=7.6000 round_seconds=18.4000 pseudo_student_dice=0.2699 pseudo_ensemble_dice=0.3335 consistency_scale=1.0000 val_dice=0.5844 val_vi=0.1675 val_ece=0.0049 +[12:13:54][ 277s][AL] round 4: 64/671 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:13:56][ 280s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.194, uniform=0.017, top1=0.322, top2=0.194, top3=0.194 -> using top1 +[12:13:56][ 280s][WEIGHTS] round 4: teacher Dice on 20 labeled patches + 0 blank (88303 fg px): fly-mito=0.322, kasthuri=0.195, lucchi=0.006, multiclass=0.000 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[12:13:57][ 281s][CONS] round 4: on the 5 new patches student Dice=0.113 vs ensemble 0.276 -> pseudo-labels from the ensemble | consistency x1.00 +[12:13:59][ 283s][FT] round 4/6 epoch 1/5 | patches=20 | loss=0.6223 +[12:14:01][ 285s][FT] round 4/6 epoch 2/5 | patches=20 | loss=0.6012 +[12:14:03][ 286s][FT] round 4/6 epoch 3/5 | patches=20 | loss=0.5775 +[12:14:05][ 288s][FT] round 4/6 epoch 4/5 | patches=20 | loss=0.5477 +[12:14:07][ 290s][FT] round 4/6 epoch 5/5 | patches=20 | loss=0.5058 +[12:14:08][ 292s][AL] round 4/6 | labeled=20 | selection_seconds=7.5000 round_seconds=20.7000 pseudo_student_dice=0.1127 pseudo_ensemble_dice=0.2760 consistency_scale=1.0000 val_dice=0.6994 val_vi=0.1209 val_ece=0.0073 +[12:14:16][ 299s][AL] round 5: 133/666 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:14:19][ 302s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.182, uniform=0.023, top1=0.284, top2=0.182, top3=0.182 -> using top1 +[12:14:19][ 302s][WEIGHTS] round 5: teacher Dice on 25 labeled patches + 0 blank (100786 fg px): fly-mito=0.284, kasthuri=0.184, lucchi=0.018, multiclass=0.000 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[12:14:23][ 306s][CONS] round 5: on the 5 new patches student Dice=0.494 vs ensemble 0.137 -> pseudo-labels from the student | consistency x0.00 +[12:14:24][ 308s][FT] round 5/6 epoch 1/5 | patches=25 | loss=0.4278 +[12:14:25][ 309s][FT] round 5/6 epoch 2/5 | patches=25 | loss=0.3891 +[12:14:27][ 310s][FT] round 5/6 epoch 3/5 | patches=25 | loss=0.3514 +[12:14:28][ 311s][FT] round 5/6 epoch 4/5 | patches=25 | loss=0.3588 +[12:14:29][ 312s][FT] round 5/6 epoch 5/5 | patches=25 | loss=0.3080 +[12:14:30][ 313s][AL] round 5/6 | labeled=25 | selection_seconds=7.4000 round_seconds=20.5000 pseudo_student_dice=0.4939 pseudo_ensemble_dice=0.1373 consistency_scale=0.0000 val_dice=0.6982 val_vi=0.1222 val_ece=0.0099 +[12:14:37][ 321s][AL] round 6: 123/661 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:15:07][ 350s][WEIGHTS] round 6: teacher agreement with the student on 656 unlabeled patches: fly-mito=0.359, kasthuri=0.416, lucchi=0.011, multiclass=0.001 | blended with labeled Dice at 0.50 +[12:15:32][ 376s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.306, uniform=0.035, top1=0.317, top2=0.306, top3=0.306 -> using top1 +[12:15:32][ 376s][WEIGHTS] round 6: teacher Dice on 30 labeled patches + 0 blank (109183 fg px): fly-mito=0.317, kasthuri=0.306, lucchi=0.014, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[12:15:37][ 380s][CONS] round 6: on the 5 new patches student Dice=0.583 vs ensemble 0.190 -> pseudo-labels from the student | consistency x0.28 +[12:15:40][ 383s][FT] round 6/6 epoch 1/5 | patches=30 | loss=0.3750 +[12:15:42][ 385s][FT] round 6/6 epoch 2/5 | patches=30 | loss=0.3679 +[12:15:45][ 388s][FT] round 6/6 epoch 3/5 | patches=30 | loss=0.3280 +[12:15:47][ 390s][FT] round 6/6 epoch 4/5 | patches=30 | loss=0.3097 +[12:15:50][ 393s][FT] round 6/6 epoch 5/5 | patches=30 | loss=0.3035 +[12:15:51][ 394s][AL] round 6/6 | labeled=30 | selection_seconds=7.4000 round_seconds=79.7000 pseudo_student_dice=0.5830 pseudo_ensemble_dice=0.1899 consistency_scale=0.2770 val_dice=0.7796 val_vi=0.1032 val_ece=0.0075 +[12:15:54][ 398s][CONS] final consolidation: pseudo-labels refreshed from the student +[12:15:57][ 400s][FT] final 1/10 epoch 1/1 | patches=30 | loss=0.3236 +[12:15:59][ 403s][FT] final 2/10 epoch 1/1 | patches=30 | loss=0.2798 +[12:16:02][ 405s][FT] final 3/10 epoch 1/1 | patches=30 | loss=0.2968 +[12:16:04][ 408s][FT] final 4/10 epoch 1/1 | patches=30 | loss=0.2693 +[12:16:07][ 410s][FT] final 5/10 epoch 1/1 | patches=30 | loss=0.2312 +[12:16:10][ 413s][FT] final 6/10 epoch 1/1 | patches=30 | loss=0.2003 +[12:16:12][ 415s][FT] final 7/10 epoch 1/1 | patches=30 | loss=0.2203 +[12:16:15][ 418s][FT] final 8/10 epoch 1/1 | patches=30 | loss=0.1931 +[12:16:17][ 420s][FT] final 9/10 epoch 1/1 | patches=30 | loss=0.2420 +[12:16:20][ 423s][FT] final 10/10 epoch 1/1 | patches=30 | loss=0.1780 +[12:16:21][ 424s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=28.9000 val_dice=0.7771 val_vi=0.0989 val_ece=0.0088 +[12:16:21][ 424s][WEIGHTS] final: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[12:16:28][ 431s][TEST] raw student (no EMA, no TTA): patch_dice=0.8138 patch_vi=0.1142 +[12:16:29][ 433s][TEST] final model without TTA: patch_dice=0.8138 patch_vi=0.1142 +[12:16:29][ 433s][RESOURCES] {"phases": {"setup": {"seconds": 0.75, "peak_cpu_rss_mb": 2191.5, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 11.38, "peak_cpu_rss_mb": 4114.7, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 117.78, "peak_cpu_rss_mb": 4292.5, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4168.0, "kd_epochs": 5}, "active_learning": {"seconds": 292.01, "peak_cpu_rss_mb": 4424.3, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4178.0, "rounds": 6}, "inference": {"seconds": 8.63, "peak_cpu_rss_mb": 4424.3, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4178.0, "patches": 147, "images": 3, "patches_per_second": 17.0, "seconds_per_image": 2.877}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 4424.3, "total_seconds": 430.6} +[12:16:29][ 433s][TEST] patch_dice=0.8323 patch_iou=0.7579 patch_precision=0.8281 patch_recall=0.9090 patch_vi=0.1053 image_dice=0.8863 image_iou=0.7972 image_precision=0.8331 image_recall=0.9485 image_vi=0.1257 cal_ece=0.0064 cal_nll=0.0393 cal_brier=0.0111 cal_mean_confidence=0.9915 cal_accuracy=0.9852 cal_predicted_prevalence=0.0663 cal_true_prevalence=0.0594 cal_mean_probability=0.0664 +[12:16:30][ 433s][DONE] drosophila-vnc macs-v11-norestart-acq-laplace b5 s2 | test patch Dice=0.8323 VI=0.1053 | image Dice=0.8863 VI=0.1257 | 433s diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..373f46602dd34437ec684ce6daf1a20707c0bfa7 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json @@ -0,0 +1,652 @@ +{ + "target": "drosophila-vnc", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 2, + "config": { + "task": "mitochondria", + "target": "drosophila-vnc", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "drosophila-vnc", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "multiclass" + ], + "alphas": [ + 0.11690492516893997, + 0.8375897330633602, + 0.0256027087898516, + 0.019902632977848233 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "multiclass" + ], + "mixture": [ + 0.11690492516893997, + 0.8375897330633602, + 0.0256027087898516, + 0.019902632977848233 + ], + "per_source": [ + 0.038599207620062845, + 0.9432861284385984, + 0.013712296406123614, + 0.0044023675352152245 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.1917820351790263, + "mmd2_mixture": 0.19749363071106324, + "mmd2_uniform": 0.3743489659922599, + "mmd2_single_source": [ + 0.7042977370508033, + 0.20706860465273969, + 0.6307956892595754, + 0.6534362949819378 + ], + "median_squared_distance": 0.19645890982127545, + "gammas": [ + 1.272530730356961, + 2.545061460713922, + 5.090122921427844, + 10.180245842855689, + 20.360491685711377 + ], + "max_weight": 0.8375897330633602, + "entropy": 0.5711595532036933, + "effective_sources": 1.7703186403916105, + "min_weight": 0.019902632977848233 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.605076346545893, + 0.31101778520145595, + 0.7002952057460718, + 0.8048259839560096 + ], + "standardized": [ + -0.0012362667606482206, + -1.5993052227157551, + 0.5162331182950395, + 1.0843083711813646 + ], + "tau": 0.5, + "max_weight": 0.9432861284385984, + "entropy": 0.26340029394025066, + "effective_sources": 1.301347536237684, + "min_weight": 0.0044023675352152245 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "kasthuri-pp", + "seconds": 1.4 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.27840715557699913, + 0.18341499813983803, + 0.0577032083417681, + 0.01681466600081078 + ], + "ensemble_val_dice": { + "mixture": 0.18652277313599772, + "mixture@raw": 0.18548230728775292, + "mixture@logit": 0.19296074616265912, + "per_source": 0.18415443581907806, + "per_source@raw": 0.1835051571090219, + "per_source@logit": 0.18905485087217583, + "uniform": 0.0696943056822927, + "uniform@raw": 0.08847159266818436, + "uniform@logit": 0.06549734203813108 + }, + "val_patches": 147 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 3.80090069770813, + 2.597146987915039 + ], + "top_source": "kasthuri-pp", + "weight_stats": { + "max_weight": 0.8375897330633602, + "entropy": 0.5711595532036933, + "effective_sources": 1.7703186403916105, + "min_weight": 0.019902632977848233 + }, + "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": 117.8, + "round0_val": { + "val_dice": 0.2058871697531777, + "val_vi": 0.22072946276036065, + "val_ece": 0.019761043084348236 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 5, + "selection_seconds": 22.1, + "round_seconds": 98.5, + "teacher_dice_labeled": [ + 0.36295908671523197, + 0.5310285360135809, + 0.013736820762368207, + 0.0019869574077847974 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 19579, + "evidence_nonempty_patches": 5, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.5831796951819076, + "restart_teacher_dice": 0.5310285360135809, + "weights": [ + 0.025328879887354556, + 0.9743856190601974, + 0.00016881695486452877, + 0.0001166840975834896 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.531149750640248, + "uniform": 0.0667619235317304, + "top1": 0.5310285360135809, + "top2": 0.5311354323916325, + "top3": 0.5311354323916325 + }, + "val_dice": 0.19819172565480536, + "val_vi": 0.21059685476621876, + "val_ece": 0.030250068283869655 + }, + { + "round": 2, + "n_labeled": 10, + "selection_seconds": 7.6, + "round_seconds": 17.6, + "teacher_dice_labeled": [ + 0.3342600608199596, + 0.36433835993119296, + 0.0058655469460196525, + 0.0008455101244417466 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 46033, + "evidence_nonempty_patches": 10, + "first_correction": false, + "restart": "", + "weights": [ + 0.09351739812091488, + 0.9051474776121796, + 0.0007676199810288104, + 0.0005675042858766415 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.36326152072288853, + "uniform": 0.032810331945778634, + "top1": 0.36433835993119296, + "top2": 0.3632638740962572, + "top3": 0.3632524156254308 + }, + "pseudo_student_dice": 0.15260812899705997, + "pseudo_ensemble_dice": 0.21822732695501648, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.41463101296538946, + "val_vi": 0.19460083461243838, + "val_ece": 0.020840921132885007 + }, + { + "round": 3, + "n_labeled": 15, + "selection_seconds": 7.6, + "round_seconds": 18.4, + "teacher_dice_labeled": [ + 0.3339991879232587, + 0.24314318156375442, + 0.005117478909572614, + 0.0005446547028838769 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 71466, + "evidence_nonempty_patches": 14, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.2424702747842685, + "uniform": 0.021214923601429972, + "top1": 0.3339991879232587, + "top2": 0.24252174949350494, + "top3": 0.2425010931351115 + }, + "pseudo_student_dice": 0.2699233539569674, + "pseudo_ensemble_dice": 0.3335354950696935, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.5843989124516004, + "val_vi": 0.16750684661110102, + "val_ece": 0.004899511730954534 + }, + { + "round": 4, + "n_labeled": 20, + "selection_seconds": 7.5, + "round_seconds": 20.7, + "teacher_dice_labeled": [ + 0.32164944334569073, + 0.19456246320800594, + 0.006338947937308593, + 0.000440294884676609 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 88303, + "evidence_nonempty_patches": 18, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.19353505028082205, + "uniform": 0.017062989853238304, + "top1": 0.32164944334569073, + "top2": 0.19381819105561976, + "top3": 0.19362159275834012 + }, + "pseudo_student_dice": 0.11274676329596497, + "pseudo_ensemble_dice": 0.2759716745795445, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.6993540102840254, + "val_vi": 0.12093193704679728, + "val_ece": 0.007255122349904359 + }, + { + "round": 5, + "n_labeled": 25, + "selection_seconds": 7.4, + "round_seconds": 20.5, + "teacher_dice_labeled": [ + 0.28431461799312685, + 0.1835229273015314, + 0.0176183379740382, + 0.00038568800806977984 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 100786, + "evidence_nonempty_patches": 23, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.18152320914405376, + "uniform": 0.023056570492629457, + "top1": 0.28431461799312685, + "top2": 0.1816722357064336, + "top3": 0.18160171744910858 + }, + "pseudo_student_dice": 0.49388154674498286, + "pseudo_ensemble_dice": 0.13730756264748847, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.6982076177663978, + "val_vi": 0.12219064496731252, + "val_ece": 0.009867805570635917 + }, + { + "round": 6, + "n_labeled": 30, + "selection_seconds": 7.4, + "round_seconds": 79.7, + "teacher_dice_labeled": [ + 0.31740826893484075, + 0.30628203034957413, + 0.013855732783164331, + 0.0007705463176043627 + ], + "teacher_dice_pool": [ + 0.358785858335439, + 0.4158121410750507, + 0.011455642146002096, + 0.0011850973087883954 + ], + "evidence_fg_pixels": 109183, + "evidence_nonempty_patches": 27, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.3063970994102441, + "uniform": 0.03496920739101875, + "top1": 0.31740826893484075, + "top2": 0.30644533343595115, + "top3": 0.3064190484860375 + }, + "pseudo_student_dice": 0.5830105146651909, + "pseudo_ensemble_dice": 0.18987521606326807, + "pseudo_source": "student", + "consistency_scale": 0.277, + "val_dice": 0.7795886737738887, + "val_vi": 0.10321806923354346, + "val_ece": 0.007524647510925679 + }, + { + "round": 7, + "n_labeled": 30, + "selection_seconds": 0.0, + "round_seconds": 28.9, + "final_consolidation": true, + "val_dice": 0.7771382821975249, + "val_vi": 0.09893786775402841, + "val_ece": 0.008758624116729442 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.126648544840516, + "auroc_top20_error": 0.6094229345181423 + }, + "al_seconds": 292.0, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "weight_stats_final": { + "max_weight": 1.0, + "entropy": -0.0, + "effective_sources": 1.0, + "min_weight": 0.0 + }, + "top_source_final": "fly-brain-mito-benchmark", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.8138396684227535, + "patch_iou": 0.7341542006754566, + "patch_precision": 0.8001245475759902, + "patch_recall": 0.9119421120838823, + "patch_vi": 0.11419680921494532, + "n_patches": 147, + "cal_ece": 0.011223098567597907, + "cal_nll": 0.05504557527311981, + "cal_brier": 0.013421014793791255, + "cal_mean_confidence": 0.9949232920450249, + "cal_accuracy": 0.9837239583333334, + "cal_predicted_prevalence": 0.06801392431972789, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.06709936536024964 + }, + "test_notta": { + "patch_dice": 0.8138396684227535, + "patch_iou": 0.7341542006754566, + "patch_precision": 0.8001245475759902, + "patch_recall": 0.9119421120838823, + "patch_vi": 0.11419680921494532, + "n_patches": 147, + "cal_ece": 0.011223098567597907, + "cal_nll": 0.05504557527311981, + "cal_brier": 0.013421014793791255, + "cal_mean_confidence": 0.9949232920450249, + "cal_accuracy": 0.9837239583333334, + "cal_predicted_prevalence": 0.06801392431972789, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.06709936536024964 + }, + "test": { + "patch_dice": 0.8322875309966076, + "patch_iou": 0.7578700879195538, + "patch_precision": 0.8280970395677735, + "patch_recall": 0.9090226856163043, + "patch_vi": 0.10528060887477136, + "n_patches": 147, + "image_dice": 0.8862696494450427, + "image_iou": 0.7971637240024779, + "image_precision": 0.8330888536136852, + "image_recall": 0.9484565471613607, + "image_vi": 0.125726327245971, + "n_images": 3, + "cal_ece": 0.0064357235725075925, + "cal_nll": 0.03925682527344598, + "cal_brier": 0.011056755049628232, + "cal_mean_confidence": 0.9915352959508038, + "cal_accuracy": 0.9852452699829932, + "cal_predicted_prevalence": 0.06629331420068027, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.06635219970296335 + }, + "eval_seconds": 8.6, + "total_seconds": 432.6, + "resources": { + "phases": { + "setup": { + "seconds": 0.75, + "peak_cpu_rss_mb": 2191.5, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 11.38, + "peak_cpu_rss_mb": 4114.7, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 117.78, + "peak_cpu_rss_mb": 4292.5, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4168.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 292.01, + "peak_cpu_rss_mb": 4424.3, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4178.0, + "rounds": 6 + }, + "inference": { + "seconds": 8.63, + "peak_cpu_rss_mb": 4424.3, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4178.0, + "patches": 147, + "images": 3, + "patches_per_second": 17.0, + "seconds_per_image": 2.877 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 4424.3, + "total_seconds": 430.6 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/drosophila-vnc/macs-v11-norestart-acq-laplace_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..9a78da011fed40c550c18442e021a920705cea75 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/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 +1,5,22.1,98.5,"[0.36295908671523197, 0.5310285360135809, 0.013736820762368207, 0.0019869574077847974]",,19579,5,True,kept,0.5831796951819076,0.5310285360135809,"[0.025328879887354556, 0.9743856190601974, 0.00016881695486452877, 0.0001166840975834896]",mixture,"{'mixture': 0.531149750640248, 'uniform': 0.0667619235317304, 'top1': 0.5310285360135809, 'top2': 0.5311354323916325, 'top3': 0.5311354323916325}",0.19819172565480536,0.21059685476621876,0.030250068283869655,,,,, +2,10,7.6,17.6,"[0.3342600608199596, 0.36433835993119296, 0.0058655469460196525, 0.0008455101244417466]",,46033,10,False,,,,"[0.09351739812091488, 0.9051474776121796, 0.0007676199810288104, 0.0005675042858766415]",mixture,"{'mixture': 0.36326152072288853, 'uniform': 0.032810331945778634, 'top1': 0.36433835993119296, 'top2': 0.3632638740962572, 'top3': 0.3632524156254308}",0.41463101296538946,0.19460083461243838,0.020840921132885007,0.15260812899705997,0.21822732695501648,ensemble,1.0, +3,15,7.6,18.4,"[0.3339991879232587, 0.24314318156375442, 0.005117478909572614, 0.0005446547028838769]",,71466,14,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.2424702747842685, 'uniform': 0.021214923601429972, 'top1': 0.3339991879232587, 'top2': 0.24252174949350494, 'top3': 0.2425010931351115}",0.5843989124516004,0.16750684661110102,0.004899511730954534,0.2699233539569674,0.3335354950696935,ensemble,1.0, +4,20,7.5,20.7,"[0.32164944334569073, 0.19456246320800594, 0.006338947937308593, 0.000440294884676609]",,88303,18,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.19353505028082205, 'uniform': 0.017062989853238304, 'top1': 0.32164944334569073, 'top2': 0.19381819105561976, 'top3': 0.19362159275834012}",0.6993540102840254,0.12093193704679728,0.007255122349904359,0.11274676329596497,0.2759716745795445,ensemble,1.0, +5,25,7.4,20.5,"[0.28431461799312685, 0.1835229273015314, 0.0176183379740382, 0.00038568800806977984]",,100786,23,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.18152320914405376, 'uniform': 0.023056570492629457, 'top1': 0.28431461799312685, 'top2': 0.1816722357064336, 'top3': 0.18160171744910858}",0.6982076177663978,0.12219064496731252,0.009867805570635917,0.49388154674498286,0.13730756264748847,student,0.0, +6,30,7.4,79.7,"[0.31740826893484075, 0.30628203034957413, 0.013855732783164331, 0.0007705463176043627]","[0.358785858335439, 0.4158121410750507, 0.011455642146002096, 0.0011850973087883954]",109183,27,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.3063970994102441, 'uniform': 0.03496920739101875, 'top1': 0.31740826893484075, 'top2': 0.30644533343595115, 'top3': 0.3064190484860375}",0.7795886737738887,0.10321806923354346,0.007524647510925679,0.5830105146651909,0.18987521606326807,student,0.277, +7,30,0.0,28.9,,,,,,,,,,,,0.7771382821975249,0.09893786775402841,0.008758624116729442,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/log.txt b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..eeccf2e953595d2ca7543e9db9c8667b1a643ee0 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/log.txt @@ -0,0 +1,107 @@ +[13:30:41][ 0s][START] target=drosophila-vnc variant=macs-v11-norestart-acq-margin budget=5.0% seed=2 | {"task": "mitochondria", "target": "drosophila-vnc", "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": 8, "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} +[13:30:41][ 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. +[13:30:41][ 0s][DATA] drosophila-vnc: train=686 val=147 test=147 patches +[13:30:42][ 2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, multiclass=2.60 +[13:30:42][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[13:30:44][ 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 +[13:30:44][ 3s][WEIGHTS] per-source: fly-mito=0.039, kasthuri=0.943, lucchi=0.014, multiclass=0.004 | min-MMD source (NeuroADDA rule) = kasthuri +[13:30:50][ 10s][TEACHERS] target-val Dice: fly-mito=0.278, kasthuri=0.183, lucchi=0.058, multiclass=0.017 +[13:30:50][ 10s][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 +[13:30:50][ 10s][WEIGHTS] using mixture: fly-mito=0.117, kasthuri=0.838, lucchi=0.026, multiclass=0.020 +[13:30:50][ 10s][INIT] student initialised from teacher kasthuri (weight 0.838) +[13:30:53][ 13s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[13:30:53][ 13s][KD] distilling the mixture ensemble (prob space) into the student +[13:31:08][ 27s][KD] soft labels for 686 patches from 4 teachers in 14s +[13:31:28][ 48s][KD] epoch 1/5 | patches=686 | loss=0.1999 | 21s +[13:31:49][ 68s][KD] epoch 2/5 | patches=686 | loss=0.1335 | 21s +[13:32:10][ 89s][KD] epoch 3/5 | patches=686 | loss=0.1266 | 21s +[13:32:30][ 110s][KD] epoch 4/5 | patches=686 | loss=0.1231 | 21s +[13:32:51][ 131s][KD] epoch 5/5 | patches=686 | loss=0.1203 | 21s +[13:32:52][ 132s][ROUND0] before any target label: val_dice=0.2058 val_vi=0.2207 +[13:32:52][ 132s][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 +[13:32:56][ 136s][AL] round-1 uncertainty quality on 686 patches: {'spearman_uncertainty_error': 0.07993511898383816, 'auroc_top20_error': 0.5728207976303819} +[13:33:11][ 150s][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 +[13:33:12][ 151s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.531, uniform=0.067, top1=0.531, top2=0.531, top3=0.531 -> using mixture +[13:33:12][ 151s][WEIGHTS] round 1: student kept (labeled Dice 0.583 vs best teacher kasthuri 0.531); distilling the corrected ensemble into it +[13:33:12][ 151s][WEIGHTS] round 1: teacher Dice on 5 labeled patches + 0 blank (19579 fg px): fly-mito=0.363, kasthuri=0.531, lucchi=0.014, multiclass=0.002 | weights: fly-mito=0.025, kasthuri=0.974, lucchi=0.000, multiclass=0.000 +[13:33:26][ 166s][KD] soft labels for 681 patches from 4 teachers in 14s +[13:33:47][ 186s][KD] epoch 1/2 | patches=681 | loss=0.0885 | 21s +[13:34:08][ 207s][KD] epoch 2/2 | patches=681 | loss=0.0859 | 21s +[13:34:22][ 221s][CONS] pseudo-labels for 681 unlabeled patches from the current ensemble in 14s +[13:34:23][ 222s][FT] round 1/6 epoch 1/5 | patches=5 | loss=0.9208 +[13:34:24][ 223s][FT] round 1/6 epoch 2/5 | patches=5 | loss=0.8768 +[13:34:25][ 224s][FT] round 1/6 epoch 3/5 | patches=5 | loss=0.8140 +[13:34:26][ 225s][FT] round 1/6 epoch 4/5 | patches=5 | loss=0.8660 +[13:34:27][ 227s][FT] round 1/6 epoch 5/5 | patches=5 | loss=0.7604 +[13:34:28][ 228s][AL] round 1/6 | labeled=5 | selection_seconds=18.3000 round_seconds=94.9000 restart_student_dice=0.5832 restart_teacher_dice=0.5310 val_dice=0.2276 val_vi=0.2140 val_ece=0.0322 +[13:34:32][ 231s][AL] round 2: 210/681 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:34:34][ 233s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.293, uniform=0.025, top1=0.384, top2=0.293, top3=0.293 -> using top1 +[13:34:34][ 233s][WEIGHTS] round 2: teacher Dice on 10 labeled patches + 0 blank (56186 fg px): fly-mito=0.384, kasthuri=0.287, lucchi=0.007, multiclass=0.002 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[13:34:35][ 234s][CONS] round 2: on the 5 new patches student Dice=0.097 vs ensemble 0.398 -> pseudo-labels from the ensemble | consistency x1.00 +[13:34:36][ 236s][FT] round 2/6 epoch 1/5 | patches=10 | loss=1.1322 +[13:34:38][ 237s][FT] round 2/6 epoch 2/5 | patches=10 | loss=1.0126 +[13:34:39][ 239s][FT] round 2/6 epoch 3/5 | patches=10 | loss=0.9161 +[13:34:41][ 240s][FT] round 2/6 epoch 4/5 | patches=10 | loss=0.8452 +[13:34:42][ 241s][FT] round 2/6 epoch 5/5 | patches=10 | loss=0.8654 +[13:34:43][ 243s][AL] round 2/6 | labeled=10 | selection_seconds=3.9000 round_seconds=13.8000 pseudo_student_dice=0.0965 pseudo_ensemble_dice=0.3979 consistency_scale=1.0000 val_dice=0.4151 val_vi=0.1911 val_ece=0.0070 +[13:34:47][ 246s][AL] round 3: 147/676 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:34:49][ 248s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.263, uniform=0.022, top1=0.307, top2=0.263, top3=0.263 -> using top1 +[13:34:49][ 248s][WEIGHTS] round 3: teacher Dice on 15 labeled patches + 0 blank (65212 fg px): fly-mito=0.307, kasthuri=0.261, lucchi=0.014, multiclass=0.001 | weights: fly-mito=1.000, kasthuri=0.000, lucchi=0.000, multiclass=0.000 +[13:34:54][ 253s][CONS] round 3: on the 5 new patches student Dice=0.400 vs ensemble 0.128 -> pseudo-labels from the student | consistency x0.00 +[13:34:54][ 254s][FT] round 3/6 epoch 1/5 | patches=15 | loss=0.8282 +[13:34:55][ 255s][FT] round 3/6 epoch 2/5 | patches=15 | loss=0.6492 +[13:34:56][ 255s][FT] round 3/6 epoch 3/5 | patches=15 | loss=0.5065 +[13:34:57][ 256s][FT] round 3/6 epoch 4/5 | patches=15 | loss=0.5487 +[13:34:58][ 257s][FT] round 3/6 epoch 5/5 | patches=15 | loss=0.5168 +[13:34:59][ 258s][AL] round 3/6 | labeled=15 | selection_seconds=3.9000 round_seconds=14.6000 pseudo_student_dice=0.3998 pseudo_ensemble_dice=0.1283 consistency_scale=0.0000 val_dice=0.5541 val_vi=0.1877 val_ece=0.0034 +[13:35:03][ 262s][AL] round 4: 43/671 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:35:32][ 291s][WEIGHTS] round 4: teacher agreement with the student on 666 unlabeled patches: fly-mito=0.377, kasthuri=0.417, lucchi=0.009, multiclass=0.002 | blended with labeled Dice at 0.50 +[13:35:58][ 317s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.325, uniform=0.031, top1=0.328, top2=0.326, top3=0.325 -> using mixture +[13:35:58][ 317s][WEIGHTS] round 4: teacher Dice on 20 labeled patches + 0 blank (71602 fg px): fly-mito=0.328, kasthuri=0.325, lucchi=0.011, multiclass=0.002 | weights: fly-mito=0.126, kasthuri=0.872, lucchi=0.001, multiclass=0.001 +[13:36:03][ 322s][CONS] round 4: on the 5 new patches student Dice=0.338 vs ensemble 0.000 -> pseudo-labels from the student | consistency x0.00 +[13:36:04][ 323s][FT] round 4/6 epoch 1/5 | patches=20 | loss=0.5445 +[13:36:05][ 324s][FT] round 4/6 epoch 2/5 | patches=20 | loss=0.5171 +[13:36:06][ 325s][FT] round 4/6 epoch 3/5 | patches=20 | loss=0.4741 +[13:36:06][ 326s][FT] round 4/6 epoch 4/5 | patches=20 | loss=0.4724 +[13:36:07][ 327s][FT] round 4/6 epoch 5/5 | patches=20 | loss=0.4460 +[13:36:09][ 328s][AL] round 4/6 | labeled=20 | selection_seconds=3.9000 round_seconds=68.6000 pseudo_student_dice=0.3375 pseudo_ensemble_dice=0.0001 consistency_scale=0.0000 val_dice=0.7049 val_vi=0.1246 val_ece=0.0086 +[13:36:12][ 332s][AL] round 5: 136/666 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:36:42][ 361s][WEIGHTS] round 5: teacher agreement with the student on 661 unlabeled patches: fly-mito=0.343, kasthuri=0.434, lucchi=0.010, multiclass=0.001 | blended with labeled Dice at 0.50 +[13:37:08][ 387s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.312, uniform=0.035, top1=0.312, top2=0.312, top3=0.312 -> using mixture +[13:37:08][ 387s][WEIGHTS] round 5: teacher Dice on 25 labeled patches + 0 blank (93337 fg px): fly-mito=0.312, kasthuri=0.311, lucchi=0.010, multiclass=0.001 | weights: fly-mito=0.123, kasthuri=0.874, lucchi=0.001, multiclass=0.001 +[13:37:12][ 392s][CONS] round 5: on the 5 new patches student Dice=0.425 vs ensemble 0.000 -> pseudo-labels from the student | consistency x0.00 +[13:37:14][ 393s][FT] round 5/6 epoch 1/5 | patches=25 | loss=0.4196 +[13:37:15][ 394s][FT] round 5/6 epoch 2/5 | patches=25 | loss=0.3617 +[13:37:16][ 395s][FT] round 5/6 epoch 3/5 | patches=25 | loss=0.4179 +[13:37:17][ 396s][FT] round 5/6 epoch 4/5 | patches=25 | loss=0.3490 +[13:37:18][ 397s][FT] round 5/6 epoch 5/5 | patches=25 | loss=0.4248 +[13:37:19][ 399s][AL] round 5/6 | labeled=25 | selection_seconds=3.8000 round_seconds=69.5000 pseudo_student_dice=0.4253 pseudo_ensemble_dice=0.0000 consistency_scale=0.0000 val_dice=0.6628 val_vi=0.1320 val_ece=0.0144 +[13:37:23][ 402s][AL] round 6: 141/661 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:37:52][ 432s][WEIGHTS] round 6: teacher agreement with the student on 656 unlabeled patches: fly-mito=0.359, kasthuri=0.426, lucchi=0.007, multiclass=0.001 | blended with labeled Dice at 0.50 +[13:38:18][ 457s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.317, uniform=0.037, top1=0.316, top2=0.317, top3=0.317 -> using mixture +[13:38:18][ 457s][WEIGHTS] round 6: teacher Dice on 30 labeled patches + 0 blank (101937 fg px): fly-mito=0.314, kasthuri=0.316, lucchi=0.008, multiclass=0.001 | weights: fly-mito=0.121, kasthuri=0.877, lucchi=0.001, multiclass=0.001 +[13:38:23][ 462s][CONS] round 6: on the 5 new patches student Dice=0.619 vs ensemble 0.407 -> pseudo-labels from the student | consistency x0.40 +[13:38:25][ 465s][FT] round 6/6 epoch 1/5 | patches=30 | loss=0.4783 +[13:38:28][ 467s][FT] round 6/6 epoch 2/5 | patches=30 | loss=0.4042 +[13:38:30][ 470s][FT] round 6/6 epoch 3/5 | patches=30 | loss=0.3390 +[13:38:33][ 472s][FT] round 6/6 epoch 4/5 | patches=30 | loss=0.3294 +[13:38:35][ 475s][FT] round 6/6 epoch 5/5 | patches=30 | loss=0.3309 +[13:38:37][ 476s][AL] round 6/6 | labeled=30 | selection_seconds=3.8000 round_seconds=76.3000 pseudo_student_dice=0.6186 pseudo_ensemble_dice=0.4073 consistency_scale=0.3950 val_dice=0.6835 val_vi=0.1245 val_ece=0.0127 +[13:38:40][ 480s][CONS] final consolidation: pseudo-labels refreshed from the student +[13:38:43][ 482s][FT] final 1/10 epoch 1/1 | patches=30 | loss=0.3316 +[13:38:45][ 485s][FT] final 2/10 epoch 1/1 | patches=30 | loss=0.2807 +[13:38:48][ 487s][FT] final 3/10 epoch 1/1 | patches=30 | loss=0.3052 +[13:38:50][ 490s][FT] final 4/10 epoch 1/1 | patches=30 | loss=0.2432 +[13:38:53][ 492s][FT] final 5/10 epoch 1/1 | patches=30 | loss=0.2443 +[13:38:55][ 495s][FT] final 6/10 epoch 1/1 | patches=30 | loss=0.2739 +[13:38:58][ 497s][FT] final 7/10 epoch 1/1 | patches=30 | loss=0.2722 +[13:39:00][ 500s][FT] final 8/10 epoch 1/1 | patches=30 | loss=0.2558 +[13:39:03][ 502s][FT] final 9/10 epoch 1/1 | patches=30 | loss=0.2214 +[13:39:05][ 505s][FT] final 10/10 epoch 1/1 | patches=30 | loss=0.2567 +[13:39:06][ 506s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=28.7000 val_dice=0.7389 val_vi=0.1129 val_ece=0.0112 +[13:39:06][ 506s][WEIGHTS] final: fly-mito=0.121, kasthuri=0.877, lucchi=0.001, multiclass=0.001 +[13:39:14][ 513s][TEST] raw student (no EMA, no TTA): patch_dice=0.8171 patch_vi=0.1140 +[13:39:15][ 514s][TEST] final model without TTA: patch_dice=0.8171 patch_vi=0.1140 +[13:39:15][ 514s][RESOURCES] {"phases": {"setup": {"seconds": 0.77, "peak_cpu_rss_mb": 2191.8, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 11.09, "peak_cpu_rss_mb": 4114.3, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 117.87, "peak_cpu_rss_mb": 4292.8, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4168.0, "kd_epochs": 5}, "active_learning": {"seconds": 374.13, "peak_cpu_rss_mb": 4406.5, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4176.0, "rounds": 6}, "inference": {"seconds": 8.56, "peak_cpu_rss_mb": 4406.5, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4176.0, "patches": 147, "images": 3, "patches_per_second": 17.2, "seconds_per_image": 2.853}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 4406.5, "total_seconds": 512.4} +[13:39:15][ 514s][TEST] patch_dice=0.8349 patch_iou=0.7587 patch_precision=0.8561 patch_recall=0.8840 patch_vi=0.1017 image_dice=0.9032 image_iou=0.8241 image_precision=0.8944 image_recall=0.9141 image_vi=0.1090 cal_ece=0.0030 cal_nll=0.0326 cal_brier=0.0094 cal_mean_confidence=0.9897 cal_accuracy=0.9869 cal_predicted_prevalence=0.0606 cal_true_prevalence=0.0594 cal_mean_probability=0.0604 +[13:39:15][ 515s][DONE] drosophila-vnc macs-v11-norestart-acq-margin b5 s2 | test patch Dice=0.8349 VI=0.1017 | image Dice=0.9032 VI=0.1090 | 514s diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/result.json b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..adb571373af8a89204da1cc1b12232e971383cf7 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/result.json @@ -0,0 +1,662 @@ +{ + "target": "drosophila-vnc", + "variant": "macs-v11-norestart-acq-margin", + "budget": 5.0, + "seed": 2, + "config": { + "task": "mitochondria", + "target": "drosophila-vnc", + "variant": "macs-v11-norestart-acq-margin", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "drosophila-vnc", + "--variants", + "macs-v11-norestart-acq-margin", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "multiclass" + ], + "alphas": [ + 0.11690492516893997, + 0.8375897330633602, + 0.0256027087898516, + 0.019902632977848233 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "multiclass" + ], + "mixture": [ + 0.11690492516893997, + 0.8375897330633602, + 0.0256027087898516, + 0.019902632977848233 + ], + "per_source": [ + 0.038599207620062845, + 0.9432861284385984, + 0.013712296406123614, + 0.0044023675352152245 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.1917820351790263, + "mmd2_mixture": 0.19749363071106324, + "mmd2_uniform": 0.3743489659922599, + "mmd2_single_source": [ + 0.7042977370508033, + 0.20706860465273969, + 0.6307956892595754, + 0.6534362949819378 + ], + "median_squared_distance": 0.19645890982127545, + "gammas": [ + 1.272530730356961, + 2.545061460713922, + 5.090122921427844, + 10.180245842855689, + 20.360491685711377 + ], + "max_weight": 0.8375897330633602, + "entropy": 0.5711595532036933, + "effective_sources": 1.7703186403916105, + "min_weight": 0.019902632977848233 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.605076346545893, + 0.31101778520145595, + 0.7002952057460718, + 0.8048259839560096 + ], + "standardized": [ + -0.0012362667606482206, + -1.5993052227157551, + 0.5162331182950395, + 1.0843083711813646 + ], + "tau": 0.5, + "max_weight": 0.9432861284385984, + "entropy": 0.26340029394025066, + "effective_sources": 1.301347536237684, + "min_weight": 0.0044023675352152245 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "kasthuri-pp", + "seconds": 1.7 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.27840715557699913, + 0.18341499813983803, + 0.0577032083417681, + 0.01681466600081078 + ], + "ensemble_val_dice": { + "mixture": 0.18652277313599772, + "mixture@raw": 0.18548230728775292, + "mixture@logit": 0.19296074616265912, + "per_source": 0.18415443581907806, + "per_source@raw": 0.1835051571090219, + "per_source@logit": 0.18905485087217583, + "uniform": 0.0696943056822927, + "uniform@raw": 0.08847159266818436, + "uniform@logit": 0.06549734203813108 + }, + "val_patches": 147 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 3.80090069770813, + 2.597146987915039 + ], + "top_source": "kasthuri-pp", + "weight_stats": { + "max_weight": 0.8375897330633602, + "entropy": 0.5711595532036933, + "effective_sources": 1.7703186403916105, + "min_weight": 0.019902632977848233 + }, + "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": 117.9, + "round0_val": { + "val_dice": 0.20583852352297866, + "val_vi": 0.2206894295217048, + "val_ece": 0.019786654639967937 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 5, + "selection_seconds": 18.3, + "round_seconds": 94.9, + "teacher_dice_labeled": [ + 0.36295908671523197, + 0.5310285360135809, + 0.013736820762368207, + 0.0019869574077847974 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 19579, + "evidence_nonempty_patches": 5, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.583215536844693, + "restart_teacher_dice": 0.5310285360135809, + "weights": [ + 0.025328879887354556, + 0.9743856190601974, + 0.00016881695486452877, + 0.0001166840975834896 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.531149750640248, + "uniform": 0.0667619235317304, + "top1": 0.5310285360135809, + "top2": 0.5311354323916325, + "top3": 0.5311354323916325 + }, + "val_dice": 0.22762768970263805, + "val_vi": 0.2139711243244451, + "val_ece": 0.03221462987886444 + }, + { + "round": 2, + "n_labeled": 10, + "selection_seconds": 3.9, + "round_seconds": 13.8, + "teacher_dice_labeled": [ + 0.38401783708865406, + 0.28656576040595816, + 0.006609724978735469, + 0.001720193654790828 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 56186, + "evidence_nonempty_patches": 10, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.29303699199791533, + "uniform": 0.024812635707781746, + "top1": 0.38401783708865406, + "top2": 0.29324058902043343, + "top3": 0.2931189822302791 + }, + "pseudo_student_dice": 0.09651973250447396, + "pseudo_ensemble_dice": 0.3979095217735406, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.41508196440410405, + "val_vi": 0.1910820541092168, + "val_ece": 0.007020516502335978 + }, + { + "round": 3, + "n_labeled": 15, + "selection_seconds": 3.9, + "round_seconds": 14.6, + "teacher_dice_labeled": [ + 0.3067076899556643, + 0.2610554773924798, + 0.014201309254393293, + 0.0014810969278690527 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 65212, + "evidence_nonempty_patches": 14, + "first_correction": false, + "restart": "", + "weights": [ + 1.0, + 0.0, + 0.0, + 0.0 + ], + "ensemble_choice": "top1", + "candidate_dice": { + "mixture": 0.26331644529645354, + "uniform": 0.022352339093262688, + "top1": 0.3067076899556643, + "top2": 0.2632606335914767, + "top3": 0.26327461896504517 + }, + "pseudo_student_dice": 0.39983894945490583, + "pseudo_ensemble_dice": 0.12834515276777023, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.5541464148558664, + "val_vi": 0.18765088084462966, + "val_ece": 0.0034033482403390374 + }, + { + "round": 4, + "n_labeled": 20, + "selection_seconds": 3.9, + "round_seconds": 68.6, + "teacher_dice_labeled": [ + 0.32843252673042683, + 0.32500226774835783, + 0.011205289601600254, + 0.001607866862380934 + ], + "teacher_dice_pool": [ + 0.37703511707654025, + 0.41697163557551786, + 0.009469044489436806, + 0.001867650226971613 + ], + "evidence_fg_pixels": 71602, + "evidence_nonempty_patches": 15, + "first_correction": false, + "restart": "", + "weights": [ + 0.12596412966410978, + 0.8720632836330818, + 0.0011561112109388508, + 0.0008164754918695407 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.3254468576327058, + "uniform": 0.03129050120454881, + "top1": 0.32843252673042683, + "top2": 0.3255205919573236, + "top3": 0.32547291784264754 + }, + "pseudo_student_dice": 0.33754523002699444, + "pseudo_ensemble_dice": 7.686395080707149e-05, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.7049201214465788, + "val_vi": 0.12461696322525986, + "val_ece": 0.008641253938814855 + }, + { + "round": 5, + "n_labeled": 25, + "selection_seconds": 3.8, + "round_seconds": 69.5, + "teacher_dice_labeled": [ + 0.31201821320261003, + 0.3108151802191612, + 0.009948750155924245, + 0.0009971843286295514 + ], + "teacher_dice_pool": [ + 0.3428589401009912, + 0.43397047373891273, + 0.00959118250712609, + 0.0009592051120907552 + ], + "evidence_fg_pixels": 93337, + "evidence_nonempty_patches": 20, + "first_correction": false, + "restart": "", + "weights": [ + 0.12349790890107877, + 0.8742455597184858, + 0.0013189906070092267, + 0.0009375407734261823 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.3118025581344997, + "uniform": 0.03485867907823771, + "top1": 0.31201821320261003, + "top2": 0.3118660509609209, + "top3": 0.31182082107141296 + }, + "pseudo_student_dice": 0.42525555049156843, + "pseudo_ensemble_dice": 3.389256058295204e-05, + "pseudo_source": "student", + "consistency_scale": 0.0, + "val_dice": 0.6627517799862803, + "val_vi": 0.13200599901825022, + "val_ece": 0.014428097276413115 + }, + { + "round": 6, + "n_labeled": 30, + "selection_seconds": 3.8, + "round_seconds": 76.3, + "teacher_dice_labeled": [ + 0.31439930418264694, + 0.3158112661728686, + 0.008469861850983751, + 0.000731929251973725 + ], + "teacher_dice_pool": [ + 0.35933004834555277, + 0.42554124651196956, + 0.006534919388207661, + 0.0005166021294979816 + ], + "evidence_fg_pixels": 101937, + "evidence_nonempty_patches": 25, + "first_correction": false, + "restart": "", + "weights": [ + 0.12071086575697663, + 0.8771562753946593, + 0.0012404086509053586, + 0.0008924501974586932 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.3167068292410805, + "uniform": 0.03696126287935871, + "top1": 0.3158112661728686, + "top2": 0.3167952146594093, + "top3": 0.31674336464528546 + }, + "pseudo_student_dice": 0.6185659343554081, + "pseudo_ensemble_dice": 0.40730560578661845, + "pseudo_source": "student", + "consistency_scale": 0.395, + "val_dice": 0.6835278718987173, + "val_vi": 0.12453932816493471, + "val_ece": 0.012681996756738407 + }, + { + "round": 7, + "n_labeled": 30, + "selection_seconds": 0.0, + "round_seconds": 28.7, + "final_consolidation": true, + "val_dice": 0.7389110295338063, + "val_vi": 0.11285964003514103, + "val_ece": 0.011176637690902614 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.07993511898383816, + "auroc_top20_error": 0.5728207976303819 + }, + "al_seconds": 374.1, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.12071086575697663, + 0.8771562753946593, + 0.0012404086509053586, + 0.0008924501974586932 + ], + "weight_stats_final": { + "max_weight": 0.8771562753946593, + "entropy": 0.38476242814608885, + "effective_sources": 1.4692652239117807, + "min_weight": 0.0008924501974586932 + }, + "top_source_final": "kasthuri-pp", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.8170945417514479, + "patch_iou": 0.7325863544605868, + "patch_precision": 0.8206006516040203, + "patch_recall": 0.8926114311565548, + "patch_vi": 0.11397135888121473, + "n_patches": 147, + "cal_ece": 0.009251723975275686, + "cal_nll": 0.059870382370229576, + "cal_brier": 0.0121247786045106, + "cal_mean_confidence": 0.9941307853684778, + "cal_accuracy": 0.9849728954081632, + "cal_predicted_prevalence": 0.06171609268707483, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.06030104652931691 + }, + "test_notta": { + "patch_dice": 0.8170945417514479, + "patch_iou": 0.7325863544605868, + "patch_precision": 0.8206006516040203, + "patch_recall": 0.8926114311565548, + "patch_vi": 0.11397135888121473, + "n_patches": 147, + "cal_ece": 0.009251723975275686, + "cal_nll": 0.059870382370229576, + "cal_brier": 0.0121247786045106, + "cal_mean_confidence": 0.9941307853684778, + "cal_accuracy": 0.9849728954081632, + "cal_predicted_prevalence": 0.06171609268707483, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.06030104652931691 + }, + "test": { + "patch_dice": 0.8348826233141908, + "patch_iou": 0.7586723040300558, + "patch_precision": 0.8560631492580926, + "patch_recall": 0.8840063786551214, + "patch_vi": 0.10169670823361057, + "n_patches": 147, + "image_dice": 0.9031688922650402, + "image_iou": 0.8240584503474547, + "image_precision": 0.894351360788363, + "image_recall": 0.914061361626028, + "image_vi": 0.10897988313866609, + "n_images": 3, + "cal_ece": 0.003019011167729832, + "cal_nll": 0.0325651706703299, + "cal_brier": 0.009432980359557166, + "cal_mean_confidence": 0.9897270955516048, + "cal_accuracy": 0.9869393069727891, + "cal_predicted_prevalence": 0.06056016156462585, + "cal_true_prevalence": 0.05939094387755102, + "cal_mean_probability": 0.06044280983044895 + }, + "eval_seconds": 8.6, + "total_seconds": 514.4, + "resources": { + "phases": { + "setup": { + "seconds": 0.77, + "peak_cpu_rss_mb": 2191.8, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 11.09, + "peak_cpu_rss_mb": 4114.3, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 117.87, + "peak_cpu_rss_mb": 4292.8, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4168.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 374.13, + "peak_cpu_rss_mb": 4406.5, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4176.0, + "rounds": 6 + }, + "inference": { + "seconds": 8.56, + "peak_cpu_rss_mb": 4406.5, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4176.0, + "patches": 147, + "images": 3, + "patches_per_second": 17.2, + "seconds_per_image": 2.853 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 4406.5, + "total_seconds": 512.4 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/drosophila-vnc/macs-v11-norestart-acq-margin_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/mitochondria/drosophila-vnc/macs-v11-norestart-acq-margin/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..a4908ad5b6cb7ab531e5000b711bd11f277a1ae9 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/drosophila-vnc/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 +1,5,18.3,94.9,"[0.36295908671523197, 0.5310285360135809, 0.013736820762368207, 0.0019869574077847974]",,19579,5,True,kept,0.583215536844693,0.5310285360135809,"[0.025328879887354556, 0.9743856190601974, 0.00016881695486452877, 0.0001166840975834896]",mixture,"{'mixture': 0.531149750640248, 'uniform': 0.0667619235317304, 'top1': 0.5310285360135809, 'top2': 0.5311354323916325, 'top3': 0.5311354323916325}",0.22762768970263805,0.2139711243244451,0.03221462987886444,,,,, +2,10,3.9,13.8,"[0.38401783708865406, 0.28656576040595816, 0.006609724978735469, 0.001720193654790828]",,56186,10,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.29303699199791533, 'uniform': 0.024812635707781746, 'top1': 0.38401783708865406, 'top2': 0.29324058902043343, 'top3': 0.2931189822302791}",0.41508196440410405,0.1910820541092168,0.007020516502335978,0.09651973250447396,0.3979095217735406,ensemble,1.0, +3,15,3.9,14.6,"[0.3067076899556643, 0.2610554773924798, 0.014201309254393293, 0.0014810969278690527]",,65212,14,False,,,,"[1.0, 0.0, 0.0, 0.0]",top1,"{'mixture': 0.26331644529645354, 'uniform': 0.022352339093262688, 'top1': 0.3067076899556643, 'top2': 0.2632606335914767, 'top3': 0.26327461896504517}",0.5541464148558664,0.18765088084462966,0.0034033482403390374,0.39983894945490583,0.12834515276777023,student,0.0, +4,20,3.9,68.6,"[0.32843252673042683, 0.32500226774835783, 0.011205289601600254, 0.001607866862380934]","[0.37703511707654025, 0.41697163557551786, 0.009469044489436806, 0.001867650226971613]",71602,15,False,,,,"[0.12596412966410978, 0.8720632836330818, 0.0011561112109388508, 0.0008164754918695407]",mixture,"{'mixture': 0.3254468576327058, 'uniform': 0.03129050120454881, 'top1': 0.32843252673042683, 'top2': 0.3255205919573236, 'top3': 0.32547291784264754}",0.7049201214465788,0.12461696322525986,0.008641253938814855,0.33754523002699444,7.686395080707149e-05,student,0.0, +5,25,3.8,69.5,"[0.31201821320261003, 0.3108151802191612, 0.009948750155924245, 0.0009971843286295514]","[0.3428589401009912, 0.43397047373891273, 0.00959118250712609, 0.0009592051120907552]",93337,20,False,,,,"[0.12349790890107877, 0.8742455597184858, 0.0013189906070092267, 0.0009375407734261823]",mixture,"{'mixture': 0.3118025581344997, 'uniform': 0.03485867907823771, 'top1': 0.31201821320261003, 'top2': 0.3118660509609209, 'top3': 0.31182082107141296}",0.6627517799862803,0.13200599901825022,0.014428097276413115,0.42525555049156843,3.389256058295204e-05,student,0.0, +6,30,3.8,76.3,"[0.31439930418264694, 0.3158112661728686, 0.008469861850983751, 0.000731929251973725]","[0.35933004834555277, 0.42554124651196956, 0.006534919388207661, 0.0005166021294979816]",101937,25,False,,,,"[0.12071086575697663, 0.8771562753946593, 0.0012404086509053586, 0.0008924501974586932]",mixture,"{'mixture': 0.3167068292410805, 'uniform': 0.03696126287935871, 'top1': 0.3158112661728686, 'top2': 0.3167952146594093, 'top3': 0.31674336464528546}",0.6835278718987173,0.12453932816493471,0.012681996756738407,0.6185659343554081,0.40730560578661845,student,0.395, +7,30,0.0,28.7,,,,,,,,,,,,0.7389110295338063,0.11285964003514103,0.011176637690902614,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt b/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..0fc5b587e6393eb477222209084f1663f5db92f1 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/log.txt @@ -0,0 +1,108 @@ +[07:50:03][ 0s][START] target=kasthuri-pp variant=macs-v11-norestart-acq-entropy budget=5.0% seed=1 | {"task": "mitochondria", "target": "kasthuri-pp", "variant": "macs-v11-norestart-acq-entropy", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[07:50:03][ 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. +[07:50:04][ 1s][DATA] kasthuri-pp: train=12023 val=2376 test=2376 patches +[07:50:05][ 2s][CALIB] source-val temperatures: fly-mito=1.38, lucchi=3.80, dros-vnc=2.77, multiclass=2.60 +[07:50:05][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[07:50:06][ 4s][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 +[07:50:06][ 4s][WEIGHTS] per-source: fly-mito=0.004, lucchi=0.027, dros-vnc=0.949, multiclass=0.020 | min-MMD source (NeuroADDA rule) = dros-vnc +[07:50:45][ 42s][TEACHERS] target-val Dice: fly-mito=0.194, lucchi=0.137, dros-vnc=0.349, multiclass=0.152 +[07:50:45][ 42s][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 +[07:50:45][ 42s][WEIGHTS] using mixture: fly-mito=0.055, lucchi=0.143, dros-vnc=0.678, multiclass=0.124 +[07:50:45][ 42s][INIT] student initialised from teacher dros-vnc (weight 0.678) +[07:50:54][ 51s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[07:50:54][ 51s][KD] distilling the mixture ensemble (prob space) into the student +[07:52:56][ 173s][KD] soft labels for 6000 patches from 4 teachers in 122s +[07:55:54][ 351s][KD] epoch 1/5 | patches=6000 | loss=0.1770 | 178s +[07:58:51][ 529s][KD] epoch 2/5 | patches=6000 | loss=0.1513 | 178s +[08:01:49][ 706s][KD] epoch 3/5 | patches=6000 | loss=0.1493 | 178s +[08:04:47][ 884s][KD] epoch 4/5 | patches=6000 | loss=0.1449 | 178s +[08:07:45][ 1062s][KD] epoch 5/5 | patches=6000 | loss=0.1438 | 178s +[08:07:51][ 1068s][ROUND0] before any target label: val_dice=0.2223 val_vi=0.1502 +[08:07:51][ 1068s][AL] patches=12023 (pool 10327 after dropping blank; free blank negatives: True) | budget=601 (5.0%) | rounds=6 | per_round=100 | acquisition=entropy | first_round=stratified +[08:08:33][ 1110s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': -0.3458863863149128, 'auroc_top20_error': 0.36265166015625} +[08:09:14][ 1151s][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 +[08:09:30][ 1167s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.203, uniform=0.004, top1=0.204, top2=0.205, top3=0.204 -> using mixture +[08:09:32][ 1169s][WEIGHTS] round 1: student kept (labeled Dice 0.038 vs best teacher dros-vnc 0.204); distilling the corrected ensemble into it +[08:09:32][ 1169s][WEIGHTS] round 1: teacher Dice on 100 labeled patches + 100 blank (277958 fg px): fly-mito=0.038, lucchi=0.009, dros-vnc=0.204, multiclass=0.002 | weights: fly-mito=0.014, lucchi=0.028, dros-vnc=0.935, multiclass=0.023 +[08:11:34][ 1292s][KD] soft labels for 6000 patches from 4 teachers in 123s +[08:14:33][ 1470s][KD] epoch 1/2 | patches=6000 | loss=0.0603 | 178s +[08:17:31][ 1648s][KD] epoch 2/2 | patches=6000 | loss=0.0593 | 178s +[08:18:12][ 1689s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 41s +[08:18:24][ 1702s][FT] round 1/6 epoch 1/5 | patches=200 | loss=0.9295 +[08:18:37][ 1714s][FT] round 1/6 epoch 2/5 | patches=200 | loss=0.8602 +[08:18:49][ 1727s][FT] round 1/6 epoch 3/5 | patches=200 | loss=0.5114 +[08:19:02][ 1739s][FT] round 1/6 epoch 4/5 | patches=200 | loss=0.4687 +[08:19:15][ 1752s][FT] round 1/6 epoch 5/5 | patches=200 | loss=0.3942 +[08:19:20][ 1758s][AL] round 1/6 | labeled=100 | selection_seconds=83.3000 round_seconds=683.6000 restart_student_dice=0.0379 restart_teacher_dice=0.2043 val_dice=0.7305 val_vi=0.0986 val_ece=0.0061 +[08:20:02][ 1799s][AL] round 2: 2781/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[08:20:33][ 1830s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.135, uniform=0.014, top1=0.137, top2=0.137, top3=0.136 -> using mixture +[08:20:33][ 1830s][WEIGHTS] round 2: teacher Dice on 200 labeled patches + 200 blank (1027506 fg px): fly-mito=0.076, lucchi=0.033, dros-vnc=0.137, multiclass=0.008 | weights: fly-mito=0.038, lucchi=0.064, dros-vnc=0.855, multiclass=0.043 +[08:20:54][ 1851s][CONS] round 2: on the 100 new patches student Dice=0.590 vs ensemble 0.109 -> pseudo-labels from the student | consistency x0.30 +[08:21:18][ 1875s][FT] round 2/6 epoch 1/5 | patches=400 | loss=0.2601 +[08:21:42][ 1900s][FT] round 2/6 epoch 2/5 | patches=400 | loss=0.2320 +[08:22:07][ 1924s][FT] round 2/6 epoch 3/5 | patches=400 | loss=0.2258 +[08:22:31][ 1948s][FT] round 2/6 epoch 4/5 | patches=400 | loss=0.1876 +[08:22:55][ 1973s][FT] round 2/6 epoch 5/5 | patches=400 | loss=0.1990 +[08:23:01][ 1978s][AL] round 2/6 | labeled=200 | selection_seconds=42.0000 round_seconds=214.9000 pseudo_student_dice=0.5901 pseudo_ensemble_dice=0.1087 consistency_scale=0.3000 val_dice=0.8368 val_vi=0.0604 val_ece=0.0019 +[08:23:43][ 2020s][AL] round 3: 2240/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[08:25:28][ 2126s][WEIGHTS] round 3: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.116, lucchi=0.014, dros-vnc=0.264, multiclass=0.001 | blended with labeled Dice at 0.50 +[08:27:04][ 2221s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.214, uniform=0.013, top1=0.215, top2=0.215, top3=0.214 -> using mixture +[08:27:04][ 2221s][WEIGHTS] round 3: teacher Dice on 300 labeled patches + 300 blank (1431921 fg px): fly-mito=0.094, lucchi=0.020, dros-vnc=0.215, multiclass=0.004 | weights: fly-mito=0.023, lucchi=0.028, dros-vnc=0.929, multiclass=0.021 +[08:27:25][ 2242s][CONS] round 3: on the 100 new patches student Dice=0.816 vs ensemble 0.232 -> pseudo-labels from the student | consistency x1.00 +[08:28:01][ 2278s][FT] round 3/6 epoch 1/5 | patches=600 | loss=0.2296 +[08:28:37][ 2314s][FT] round 3/6 epoch 2/5 | patches=600 | loss=0.2064 +[08:29:13][ 2350s][FT] round 3/6 epoch 3/5 | patches=600 | loss=0.1887 +[08:29:49][ 2386s][FT] round 3/6 epoch 4/5 | patches=600 | loss=0.1811 +[08:30:25][ 2422s][FT] round 3/6 epoch 5/5 | patches=600 | loss=0.1760 +[08:30:31][ 2428s][AL] round 3/6 | labeled=300 | selection_seconds=41.9000 round_seconds=443.8000 pseudo_student_dice=0.8157 pseudo_ensemble_dice=0.2319 consistency_scale=1.0000 val_dice=0.8475 val_vi=0.0524 val_ece=0.0034 +[08:31:13][ 2470s][AL] round 4: 2399/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[08:33:05][ 2583s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.110, lucchi=0.014, dros-vnc=0.272, multiclass=0.002 | blended with labeled Dice at 0.50 +[08:34:48][ 2685s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.213, uniform=0.012, top1=0.214, top2=0.214, top3=0.213 -> using mixture +[08:34:48][ 2685s][WEIGHTS] round 4: teacher Dice on 400 labeled patches + 400 blank (1887583 fg px): fly-mito=0.091, lucchi=0.018, dros-vnc=0.214, multiclass=0.004 | weights: fly-mito=0.022, lucchi=0.028, dros-vnc=0.930, multiclass=0.021 +[08:35:09][ 2706s][CONS] round 4: on the 100 new patches student Dice=0.820 vs ensemble 0.124 -> pseudo-labels from the student | consistency x1.00 +[08:35:57][ 2754s][FT] round 4/6 epoch 1/5 | patches=800 | loss=0.1679 +[08:36:44][ 2801s][FT] round 4/6 epoch 2/5 | patches=800 | loss=0.1625 +[08:37:32][ 2849s][FT] round 4/6 epoch 3/5 | patches=800 | loss=0.1553 +[08:38:19][ 2897s][FT] round 4/6 epoch 4/5 | patches=800 | loss=0.1472 +[08:39:07][ 2944s][FT] round 4/6 epoch 5/5 | patches=800 | loss=0.1651 +[08:39:13][ 2950s][AL] round 4/6 | labeled=400 | selection_seconds=42.0000 round_seconds=516.3000 pseudo_student_dice=0.8200 pseudo_ensemble_dice=0.1235 consistency_scale=1.0000 val_dice=0.8402 val_vi=0.0520 val_ece=0.0027 +[08:39:55][ 2992s][AL] round 5: 2595/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[08:41:55][ 3112s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.111, lucchi=0.013, dros-vnc=0.290, multiclass=0.001 | blended with labeled Dice at 0.50 +[08:43:45][ 3222s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.220, uniform=0.013, top1=0.221, top2=0.221, top3=0.221 -> using mixture +[08:43:45][ 3222s][WEIGHTS] round 5: teacher Dice on 500 labeled patches + 500 blank (2417674 fg px): fly-mito=0.091, lucchi=0.017, dros-vnc=0.221, multiclass=0.005 | weights: fly-mito=0.021, lucchi=0.026, dros-vnc=0.934, multiclass=0.020 +[08:44:06][ 3243s][CONS] round 5: on the 100 new patches student Dice=0.806 vs ensemble 0.138 -> pseudo-labels from the student | consistency x1.00 +[08:45:05][ 3302s][FT] round 5/6 epoch 1/5 | patches=1000 | loss=0.1638 +[08:46:04][ 3362s][FT] round 5/6 epoch 2/5 | patches=1000 | loss=0.1541 +[08:47:04][ 3421s][FT] round 5/6 epoch 3/5 | patches=1000 | loss=0.1552 +[08:48:03][ 3480s][FT] round 5/6 epoch 4/5 | patches=1000 | loss=0.1423 +[08:49:02][ 3540s][FT] round 5/6 epoch 5/5 | patches=1000 | loss=0.1438 +[08:49:08][ 3546s][AL] round 5/6 | labeled=500 | selection_seconds=42.0000 round_seconds=589.6000 pseudo_student_dice=0.8063 pseudo_ensemble_dice=0.1379 consistency_scale=1.0000 val_dice=0.8604 val_vi=0.0508 val_ece=0.0027 +[08:49:50][ 3587s][AL] round 6: 2336/8000 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[08:51:51][ 3708s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.116, lucchi=0.015, dros-vnc=0.272, multiclass=0.003 | blended with labeled Dice at 0.50 +[08:53:42][ 3819s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.214, uniform=0.013, top1=0.216, top2=0.216, top3=0.215 -> using mixture +[08:53:42][ 3819s][WEIGHTS] round 6: teacher Dice on 512 labeled patches + 512 blank (2411238 fg px): fly-mito=0.093, lucchi=0.018, dros-vnc=0.216, multiclass=0.005 | weights: fly-mito=0.022, lucchi=0.027, dros-vnc=0.930, multiclass=0.021 +[08:54:03][ 3840s][CONS] round 6: on the 100 new patches student Dice=0.830 vs ensemble 0.171 -> pseudo-labels from the student | consistency x1.00 +[08:55:14][ 3911s][FT] round 6/6 epoch 1/5 | patches=1200 | loss=0.1388 +[08:56:25][ 3982s][FT] round 6/6 epoch 2/5 | patches=1200 | loss=0.1365 +[08:57:36][ 4053s][FT] round 6/6 epoch 3/5 | patches=1200 | loss=0.1608 +[08:58:47][ 4124s][FT] round 6/6 epoch 4/5 | patches=1200 | loss=0.1408 +[08:59:58][ 4196s][FT] round 6/6 epoch 5/5 | patches=1200 | loss=0.1164 +[09:00:04][ 4201s][AL] round 6/6 | labeled=600 | selection_seconds=42.0000 round_seconds=650.0000 pseudo_student_dice=0.8301 pseudo_ensemble_dice=0.1706 consistency_scale=1.0000 val_dice=0.8608 val_vi=0.0520 val_ece=0.0029 +[09:00:14][ 4212s][CONS] final consolidation: pseudo-labels refreshed from the student +[09:01:25][ 4282s][FT] final 1/10 epoch 1/1 | patches=1200 | loss=0.1211 +[09:02:36][ 4353s][FT] final 2/10 epoch 1/1 | patches=1200 | loss=0.1199 +[09:03:47][ 4425s][FT] final 3/10 epoch 1/1 | patches=1200 | loss=0.1234 +[09:04:58][ 4496s][FT] final 4/10 epoch 1/1 | patches=1200 | loss=0.1219 +[09:06:09][ 4567s][FT] final 5/10 epoch 1/1 | patches=1200 | loss=0.1027 +[09:07:21][ 4638s][FT] final 6/10 epoch 1/1 | patches=1200 | loss=0.1009 +[09:08:32][ 4709s][FT] final 7/10 epoch 1/1 | patches=1200 | loss=0.0915 +[09:09:43][ 4780s][FT] final 8/10 epoch 1/1 | patches=1200 | loss=0.0918 +[09:10:54][ 4851s][FT] final 9/10 epoch 1/1 | patches=1200 | loss=0.0996 +[09:12:05][ 4923s][FT] final 10/10 epoch 1/1 | patches=1200 | loss=0.0947 +[09:12:11][ 4928s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=721.2000 val_dice=0.8854 val_vi=0.0413 val_ece=0.0029 +[09:12:11][ 4928s][WEIGHTS] final: fly-mito=0.022, lucchi=0.027, dros-vnc=0.930, multiclass=0.021 +[09:14:02][ 5040s][TEST] raw student (no EMA, no TTA): patch_dice=0.8665 patch_vi=0.0276 +[09:14:15][ 5053s][TEST] final model without TTA: patch_dice=0.8665 patch_vi=0.0276 +[09:14:15][ 5053s][RESOURCES] {"phases": {"setup": {"seconds": 0.75, "peak_cpu_rss_mb": 2191.9, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 48.62, "peak_cpu_rss_mb": 4172.5, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 1011.42, "peak_cpu_rss_mb": 5948.1, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4128.0, "kd_epochs": 5}, "active_learning": {"seconds": 3860.3, "peak_cpu_rss_mb": 5948.1, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4170.0, "rounds": 6}, "inference": {"seconds": 124.22, "peak_cpu_rss_mb": 5948.1, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4172.0, "patches": 2376, "images": 24, "patches_per_second": 19.1, "seconds_per_image": 5.176}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 5948.1, "total_seconds": 5045.3} +[09:14:15][ 5053s][TEST] patch_dice=0.8759 patch_iou=0.8475 patch_precision=0.9330 patch_recall=0.9041 patch_vi=0.0263 image_dice=0.9468 image_iou=0.8995 image_precision=0.9461 image_recall=0.9481 image_vi=0.0320 cal_ece=0.0016 cal_nll=0.0114 cal_brier=0.0026 cal_mean_confidence=0.9985 cal_accuracy=0.9969 cal_predicted_prevalence=0.0271 cal_true_prevalence=0.0268 cal_mean_probability=0.0270 +[09:14:16][ 5054s][DONE] kasthuri-pp macs-v11-norestart-acq-entropy b5 s1 | test patch Dice=0.8759 VI=0.0263 | image Dice=0.9468 VI=0.0320 | 5053s diff --git a/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json b/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..74386568b601c7264c94dacf12594063da7ae202 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/result.json @@ -0,0 +1,667 @@ +{ + "target": "kasthuri-pp", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 1, + "config": { + "task": "mitochondria", + "target": "kasthuri-pp", + "variant": "macs-v11-norestart-acq-entropy", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "kasthuri-pp", + "--variants", + "macs-v11-norestart-acq-entropy", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "lucchi-pp", + "drosophila-vnc", + "multiclass" + ], + "alphas": [ + 0.055262541562243715, + 0.1430436080615447, + 0.6777947281235579, + 0.12389912225265376 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "lucchi-pp", + "drosophila-vnc", + "multiclass" + ], + "mixture": [ + 0.055262541562243715, + 0.1430436080615447, + 0.6777947281235579, + 0.12389912225265376 + ], + "per_source": [ + 0.00394341763649415, + 0.027391591802199966, + 0.9486104893290658, + 0.020054501232240087 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.11610906469392875, + "mmd2_mixture": 0.12571430106500386, + "mmd2_uniform": 0.23862555861405582, + "mmd2_single_source": [ + 0.7126149676152926, + 0.43664360111426287, + 0.17981946303704704, + 0.44826934222434894 + ], + "median_squared_distance": 0.17346729616407242, + "gammas": [ + 1.4411938476491837, + 2.8823876952983674, + 5.764775390596735, + 11.52955078119347, + 23.05910156238694 + ], + "max_weight": 0.6777947281235579, + "entropy": 0.9605236371075117, + "effective_sources": 2.6130644127307936, + "min_weight": 0.055262541562243715 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6491305090540279, + 0.5572005449783343, + 0.3890693813540105, + 0.571988659588714 + ], + "standardized": [ + 1.1309430868146413, + 0.16184892180501523, + -1.610532164849744, + 0.31774015623008967 + ], + "tau": 0.5, + "max_weight": 0.9486104893290658, + "entropy": 0.24881632938562975, + "effective_sources": 1.2825064526877779, + "min_weight": 0.00394341763649415 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "drosophila-vnc", + "seconds": 1.4 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.194080938472025, + 0.1369313835617696, + 0.3485636136721979, + 0.15175646072001214 + ], + "ensemble_val_dice": { + "mixture": 0.34088805889447826, + "mixture@raw": 0.34615631820143733, + "mixture@logit": 0.32096912740211536, + "per_source": 0.34760137972472194, + "per_source@raw": 0.3482221938843253, + "per_source@logit": 0.34643796682361483, + "uniform": 0.1882329719806286, + "uniform@raw": 0.18328511906937095, + "uniform@logit": 0.20666689954871842 + }, + "val_patches": 1024 + }, + "temperatures": [ + 1.3822230100631714, + 3.80090069770813, + 2.767641305923462, + 2.597146987915039 + ], + "top_source": "drosophila-vnc", + "weight_stats": { + "max_weight": 0.6777947281235579, + "entropy": 0.9605236371075117, + "effective_sources": 2.6130644127307936, + "min_weight": 0.055262541562243715 + }, + "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": 1011.4, + "round0_val": { + "val_dice": 0.22230521027783137, + "val_vi": 0.15021869142723604, + "val_ece": 0.024154192170000647 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 100, + "selection_seconds": 83.3, + "round_seconds": 683.6, + "teacher_dice_labeled": [ + 0.03781776614152498, + 0.008631654211325996, + 0.20429526102680404, + 0.002462217997419853 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 277958, + "evidence_nonempty_patches": 81, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.037854588964460804, + "restart_teacher_dice": 0.20429526102680404, + "weights": [ + 0.014425496757348876, + 0.027887828907199822, + 0.9349764743833905, + 0.022710199952060868 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.20296856291238885, + "uniform": 0.003791982665222102, + "top1": 0.20429526102680404, + "top2": 0.20459987978743366, + "top3": 0.20364223129632752 + }, + "val_dice": 0.7304906483112401, + "val_vi": 0.09860420800166039, + "val_ece": 0.006122067360769768 + }, + { + "round": 2, + "n_labeled": 200, + "selection_seconds": 42.0, + "round_seconds": 214.9, + "teacher_dice_labeled": [ + 0.07584090652827795, + 0.03299669818917039, + 0.13684430388898855, + 0.0081834642784292 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1027506, + "evidence_nonempty_patches": 178, + "first_correction": false, + "restart": "", + "weights": [ + 0.0378790588809147, + 0.06388027088587322, + 0.8550684864561654, + 0.04317218377704667 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.1345944737548064, + "uniform": 0.01422790922259778, + "top1": 0.13684430388898855, + "top2": 0.13741123518504308, + "top3": 0.13573152285302534 + }, + "pseudo_student_dice": 0.5900695150463006, + "pseudo_ensemble_dice": 0.10871404767259664, + "pseudo_source": "student", + "consistency_scale": 0.3, + "val_dice": 0.8367504873013976, + "val_vi": 0.0604071687246516, + "val_ece": 0.0018813678527083225 + }, + { + "round": 3, + "n_labeled": 300, + "selection_seconds": 41.9, + "round_seconds": 443.8, + "teacher_dice_labeled": [ + 0.09410491162631666, + 0.019753152250683706, + 0.21479288222850765, + 0.004277525336689255 + ], + "teacher_dice_pool": [ + 0.11583249102758529, + 0.014004986028802935, + 0.26411840031088085, + 0.0010927551573748757 + ], + "evidence_fg_pixels": 1431921, + "evidence_nonempty_patches": 269, + "first_correction": false, + "restart": "", + "weights": [ + 0.02265206800706618, + 0.027876604279621704, + 0.928787550559032, + 0.020683777154280256 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.21362074475237053, + "uniform": 0.013180744674238559, + "top1": 0.21479288222850765, + "top2": 0.21514335117603156, + "top3": 0.21427675306672375 + }, + "pseudo_student_dice": 0.8157384611809655, + "pseudo_ensemble_dice": 0.23191085602224795, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8474988516944891, + "val_vi": 0.052352724445911526, + "val_ece": 0.0033775932266113576 + }, + { + "round": 4, + "n_labeled": 400, + "selection_seconds": 42.0, + "round_seconds": 516.3, + "teacher_dice_labeled": [ + 0.0905930387266031, + 0.017877815086345897, + 0.21396067424212706, + 0.004024708778821875 + ], + "teacher_dice_pool": [ + 0.11028078012272939, + 0.013610624233042259, + 0.2721740555641988, + 0.001608773147522873 + ], + "evidence_fg_pixels": 1887583, + "evidence_nonempty_patches": 362, + "first_correction": false, + "restart": "", + "weights": [ + 0.022070079558340224, + 0.027608530821177382, + 0.9295014342696468, + 0.02081995535083564 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.21284120228465636, + "uniform": 0.012371191155121971, + "top1": 0.21396067424212706, + "top2": 0.21427996967549467, + "top3": 0.21346076574775963 + }, + "pseudo_student_dice": 0.820048584988824, + "pseudo_ensemble_dice": 0.12350500020243735, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8401894013194965, + "val_vi": 0.05195238749177214, + "val_ece": 0.002747599314205605 + }, + { + "round": 5, + "n_labeled": 500, + "selection_seconds": 42.0, + "round_seconds": 589.6, + "teacher_dice_labeled": [ + 0.09141055023752703, + 0.017051576100546896, + 0.22105584207629211, + 0.004550119226199254 + ], + "teacher_dice_pool": [ + 0.11144896610606583, + 0.013122411638998625, + 0.2900320636792412, + 0.0014239416017628541 + ], + "evidence_fg_pixels": 2417674, + "evidence_nonempty_patches": 461, + "first_correction": false, + "restart": "", + "weights": [ + 0.020826516930618596, + 0.025628153924054202, + 0.9339558016859595, + 0.019589527459367744 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.21996896223701284, + "uniform": 0.013380655892565459, + "top1": 0.22105584207629211, + "top2": 0.22140687033401255, + "top3": 0.22059841909763112 + }, + "pseudo_student_dice": 0.8062622159868228, + "pseudo_ensemble_dice": 0.13791089170693827, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8603952180050969, + "val_vi": 0.05084424495510591, + "val_ece": 0.002686674098639422 + }, + { + "round": 6, + "n_labeled": 600, + "selection_seconds": 42.0, + "round_seconds": 650.0, + "teacher_dice_labeled": [ + 0.0929883618669614, + 0.0183669711352127, + 0.2156353693180218, + 0.004857743495817465 + ], + "teacher_dice_pool": [ + 0.115820433241756, + 0.015135867659614768, + 0.27208450552603364, + 0.0028672229574325868 + ], + "evidence_fg_pixels": 2411238, + "evidence_nonempty_patches": 471, + "first_correction": false, + "restart": "", + "weights": [ + 0.022237263579107578, + 0.02729243992452504, + 0.9298178119198955, + 0.02065248457647184 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.21446330415652365, + "uniform": 0.013375940114072303, + "top1": 0.2156353693180218, + "top2": 0.2159996588536014, + "top3": 0.21513142769519916 + }, + "pseudo_student_dice": 0.8300632758147568, + "pseudo_ensemble_dice": 0.17062110708034692, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8607512127522224, + "val_vi": 0.05201825374881291, + "val_ece": 0.0029437962306920057 + }, + { + "round": 7, + "n_labeled": 600, + "selection_seconds": 0.0, + "round_seconds": 721.2, + "final_consolidation": true, + "val_dice": 0.8854431193714826, + "val_vi": 0.0413108489859979, + "val_ece": 0.0028508722852614143 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.3458863863149128, + "auroc_top20_error": 0.36265166015625 + }, + "al_seconds": 3860.3, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.022237263579107578, + 0.02729243992452504, + 0.9298178119198955, + 0.02065248457647184 + ], + "weight_stats_final": { + "max_weight": 0.9298178119198955, + "entropy": 0.3307084327822969, + "effective_sources": 1.3919538850150148, + "min_weight": 0.02065248457647184 + }, + "top_source_final": "drosophila-vnc", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.8665314979226896, + "patch_iou": 0.8374277072784165, + "patch_precision": 0.9208533164412269, + "patch_recall": 0.9051563822798545, + "patch_vi": 0.027594797590052678, + "n_patches": 2376, + "cal_ece": 0.0021656661630985764, + "cal_nll": 0.01375716441216002, + "cal_brier": 0.002850755817302684, + "cal_mean_confidence": 0.9987719470711373, + "cal_accuracy": 0.9966062809080387, + "cal_predicted_prevalence": 0.027316623263888888, + "cal_true_prevalence": 0.026826286958122895, + "cal_mean_probability": 0.027120532269459516 + }, + "test_notta": { + "patch_dice": 0.8665314979226896, + "patch_iou": 0.8374277072784165, + "patch_precision": 0.9208533164412269, + "patch_recall": 0.9051563822798545, + "patch_vi": 0.027594797590052678, + "n_patches": 2376, + "cal_ece": 0.0021656661630985764, + "cal_nll": 0.01375716441216002, + "cal_brier": 0.002850755817302684, + "cal_mean_confidence": 0.9987719470711373, + "cal_accuracy": 0.9966062809080387, + "cal_predicted_prevalence": 0.027316623263888888, + "cal_true_prevalence": 0.026826286958122895, + "cal_mean_probability": 0.027120532269459516 + }, + "test": { + "patch_dice": 0.875913909393184, + "patch_iou": 0.8474997327678472, + "patch_precision": 0.9330206901599026, + "patch_recall": 0.904096587521667, + "patch_vi": 0.026343271704068902, + "n_patches": 2376, + "image_dice": 0.9468178241085186, + "image_iou": 0.8995131509390805, + "image_precision": 0.9460566273553056, + "image_recall": 0.9480638794469338, + "image_vi": 0.03204408346277166, + "n_images": 24, + "cal_ece": 0.0016236001779007755, + "cal_nll": 0.011411511158590052, + "cal_brier": 0.002567256205654599, + "cal_mean_confidence": 0.9984630915407051, + "cal_accuracy": 0.9968504215330387, + "cal_predicted_prevalence": 0.027095499263468013, + "cal_true_prevalence": 0.026826286958122895, + "cal_mean_probability": 0.02698464092476411 + }, + "eval_seconds": 124.2, + "total_seconds": 5052.7, + "resources": { + "phases": { + "setup": { + "seconds": 0.75, + "peak_cpu_rss_mb": 2191.9, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 48.62, + "peak_cpu_rss_mb": 4172.5, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 1011.42, + "peak_cpu_rss_mb": 5948.1, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4128.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 3860.3, + "peak_cpu_rss_mb": 5948.1, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4170.0, + "rounds": 6 + }, + "inference": { + "seconds": 124.22, + "peak_cpu_rss_mb": 5948.1, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4172.0, + "patches": 2376, + "images": 24, + "patches_per_second": 19.1, + "seconds_per_image": 5.176 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 5948.1, + "total_seconds": 5045.3 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/kasthuri-pp/macs-v11-norestart-acq-entropy_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..3a54ad3bb0cdf9e4256ea6deafbcf57f2bdce4d3 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/kasthuri-pp/macs-v11-norestart-acq-entropy/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 +1,100,83.3,683.6,"[0.03781776614152498, 0.008631654211325996, 0.20429526102680404, 0.002462217997419853]",,277958,81,True,kept,0.037854588964460804,0.20429526102680404,"[0.014425496757348876, 0.027887828907199822, 0.9349764743833905, 0.022710199952060868]",mixture,"{'mixture': 0.20296856291238885, 'uniform': 0.003791982665222102, 'top1': 0.20429526102680404, 'top2': 0.20459987978743366, 'top3': 0.20364223129632752}",0.7304906483112401,0.09860420800166039,0.006122067360769768,,,,, +2,200,42.0,214.9,"[0.07584090652827795, 0.03299669818917039, 0.13684430388898855, 0.0081834642784292]",,1027506,178,False,,,,"[0.0378790588809147, 0.06388027088587322, 0.8550684864561654, 0.04317218377704667]",mixture,"{'mixture': 0.1345944737548064, 'uniform': 0.01422790922259778, 'top1': 0.13684430388898855, 'top2': 0.13741123518504308, 'top3': 0.13573152285302534}",0.8367504873013976,0.0604071687246516,0.0018813678527083225,0.5900695150463006,0.10871404767259664,student,0.3, +3,300,41.9,443.8,"[0.09410491162631666, 0.019753152250683706, 0.21479288222850765, 0.004277525336689255]","[0.11583249102758529, 0.014004986028802935, 0.26411840031088085, 0.0010927551573748757]",1431921,269,False,,,,"[0.02265206800706618, 0.027876604279621704, 0.928787550559032, 0.020683777154280256]",mixture,"{'mixture': 0.21362074475237053, 'uniform': 0.013180744674238559, 'top1': 0.21479288222850765, 'top2': 0.21514335117603156, 'top3': 0.21427675306672375}",0.8474988516944891,0.052352724445911526,0.0033775932266113576,0.8157384611809655,0.23191085602224795,student,1.0, +4,400,42.0,516.3,"[0.0905930387266031, 0.017877815086345897, 0.21396067424212706, 0.004024708778821875]","[0.11028078012272939, 0.013610624233042259, 0.2721740555641988, 0.001608773147522873]",1887583,362,False,,,,"[0.022070079558340224, 0.027608530821177382, 0.9295014342696468, 0.02081995535083564]",mixture,"{'mixture': 0.21284120228465636, 'uniform': 0.012371191155121971, 'top1': 0.21396067424212706, 'top2': 0.21427996967549467, 'top3': 0.21346076574775963}",0.8401894013194965,0.05195238749177214,0.002747599314205605,0.820048584988824,0.12350500020243735,student,1.0, +5,500,42.0,589.6,"[0.09141055023752703, 0.017051576100546896, 0.22105584207629211, 0.004550119226199254]","[0.11144896610606583, 0.013122411638998625, 0.2900320636792412, 0.0014239416017628541]",2417674,461,False,,,,"[0.020826516930618596, 0.025628153924054202, 0.9339558016859595, 0.019589527459367744]",mixture,"{'mixture': 0.21996896223701284, 'uniform': 0.013380655892565459, 'top1': 0.22105584207629211, 'top2': 0.22140687033401255, 'top3': 0.22059841909763112}",0.8603952180050969,0.05084424495510591,0.002686674098639422,0.8062622159868228,0.13791089170693827,student,1.0, +6,600,42.0,650.0,"[0.0929883618669614, 0.0183669711352127, 0.2156353693180218, 0.004857743495817465]","[0.115820433241756, 0.015135867659614768, 0.27208450552603364, 0.0028672229574325868]",2411238,471,False,,,,"[0.022237263579107578, 0.02729243992452504, 0.9298178119198955, 0.02065248457647184]",mixture,"{'mixture': 0.21446330415652365, 'uniform': 0.013375940114072303, 'top1': 0.2156353693180218, 'top2': 0.2159996588536014, 'top3': 0.21513142769519916}",0.8607512127522224,0.05201825374881291,0.0029437962306920057,0.8300632758147568,0.17062110708034692,student,1.0, +7,600,0.0,721.2,,,,,,,,,,,,0.8854431193714826,0.0413108489859979,0.0028508722852614143,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..8c4576ad876bd8349b793c98e55520137b35a48d --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt @@ -0,0 +1,148 @@ +[23:30:06][ 0s][START] target=lucchi-pp variant=macs-v11-norestart-acq-laplace budget=5.0% seed=1 | {"task": "mitochondria", "target": "lucchi-pp", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[23:30:06][ 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. +[23:30:06][ 0s][DATA] lucchi-pp: train=8050 val=1750 test=1750 patches +[23:30:08][ 2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, dros-vnc=2.77, multiclass=2.60 +[23:30:08][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[23:30:09][ 3s][WEIGHTS] mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 | eff.sources=1.44 | MMD2 mix=0.0050 uniform=0.2092 +[23:30:09][ 3s][WEIGHTS] per-source: fly-mito=0.013, kasthuri=0.022, dros-vnc=0.005, multiclass=0.960 | min-MMD source (NeuroADDA rule) = multiclass +[23:30:47][ 41s][TEACHERS] target-val Dice: fly-mito=0.243, kasthuri=0.161, dros-vnc=0.162, multiclass=0.833 +[23:30:47][ 41s][TEACHERS] ensemble target-val Dice: mixture=0.834, mixture@raw=0.833, mixture@logit=0.835, per_source=0.834, per_source@raw=0.833, per_source@logit=0.834, uniform=0.331, uniform@raw=0.184, uniform@logit=0.220 +[23:30:47][ 41s][WEIGHTS] using mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 +[23:30:47][ 41s][INIT] student initialised from teacher multiclass (weight 0.920) +[23:30:50][ 44s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[23:30:50][ 44s][KD] distilling the mixture ensemble (prob space) into the student +[23:32:54][ 168s][KD] soft labels for 6000 patches from 4 teachers in 123s +[23:35:54][ 348s][KD] epoch 1/5 | patches=6000 | loss=0.1185 | 180s +[23:38:54][ 528s][KD] epoch 2/5 | patches=6000 | loss=0.1064 | 181s +[23:41:55][ 709s][KD] epoch 3/5 | patches=6000 | loss=0.1046 | 181s +[23:44:55][ 889s][KD] epoch 4/5 | patches=6000 | loss=0.1035 | 180s +[23:47:55][ 1069s][KD] epoch 5/5 | patches=6000 | loss=0.1028 | 180s +[23:48:02][ 1075s][ROUND0] before any target label: val_dice=0.8843 val_vi=0.0624 +[23:48:02][ 1075s][AL] patches=8050 (pool 8050 after dropping blank; free blank negatives: True) | budget=402 (5.0%) | rounds=6 | per_round=67 | acquisition=laplace | first_round=stratified +[23:49:29][ 1163s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': 0.0892883911765001, 'auroc_top20_error': 0.473226806640625} +[23:50:11][ 1205s][AL] round 1: stratified cold start: 34 patches from the membrane-rich half (ensemble prevalence >= 0.074) + 33 from the membrane-poor half of 2000 candidates +[23:50:17][ 1211s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.932, uniform=0.354, top1=0.932, top2=0.932, top3=0.932 -> using mixture +[23:50:18][ 1212s][WEIGHTS] round 1: student kept (labeled Dice 0.940 vs best teacher multiclass 0.932); distilling the corrected ensemble into it +[23:50:18][ 1212s][WEIGHTS] round 1: teacher Dice on 67 labeled patches + 0 blank (352197 fg px): fly-mito=0.288, kasthuri=0.000, dros-vnc=0.023, multiclass=0.932 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[23:52:22][ 1336s][KD] soft labels for 6000 patches from 4 teachers in 124s +[23:55:22][ 1516s][KD] epoch 1/2 | patches=6000 | loss=0.0467 | 180s +[23:58:23][ 1697s][KD] epoch 2/2 | patches=6000 | loss=0.0462 | 181s +[23:59:04][ 1738s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 42s +[23:59:09][ 1743s][FT] round 1/6 epoch 1/5 | patches=67 | loss=0.3159 +[23:59:14][ 1748s][FT] round 1/6 epoch 2/5 | patches=67 | loss=0.3304 +[23:59:19][ 1753s][FT] round 1/6 epoch 3/5 | patches=67 | loss=0.3587 +[23:59:24][ 1758s][FT] round 1/6 epoch 4/5 | patches=67 | loss=0.3334 +[23:59:29][ 1763s][FT] round 1/6 epoch 5/5 | patches=67 | loss=0.2772 +[23:59:35][ 1769s][AL] round 1/6 | labeled=67 | selection_seconds=129.3000 round_seconds=687.5000 restart_student_dice=0.9401 restart_teacher_dice=0.9318 val_dice=0.8065 val_vi=0.0844 val_ece=0.0065 +[00:01:02][ 1856s][AL] round 2: 1197/7983 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[00:01:13][ 1867s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.840, uniform=0.218, top1=0.840, top2=0.840, top3=0.840 -> using mixture +[00:01:13][ 1867s][WEIGHTS] round 2: teacher Dice on 134 labeled patches + 0 blank (778080 fg px): fly-mito=0.289, kasthuri=0.000, dros-vnc=0.012, multiclass=0.840 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:01:18][ 1871s][CONS] round 2: on the 67 new patches student Dice=0.770 vs ensemble 0.763 -> pseudo-labels from the ensemble | consistency x1.00 +[00:01:26][ 1880s][FT] round 2/6 epoch 1/5 | patches=134 | loss=0.4674 +[00:01:35][ 1889s][FT] round 2/6 epoch 2/5 | patches=134 | loss=0.3997 +[00:01:44][ 1898s][FT] round 2/6 epoch 3/5 | patches=134 | loss=0.3579 +[00:01:53][ 1907s][FT] round 2/6 epoch 4/5 | patches=134 | loss=0.3644 +[00:02:02][ 1915s][FT] round 2/6 epoch 5/5 | patches=134 | loss=0.3159 +[00:04:10][ 0s][START] target=lucchi-pp variant=macs-v11-norestart-acq-laplace budget=5.0% seed=1 | {"task": "mitochondria", "target": "lucchi-pp", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[00:04:10][ 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. +[00:04:15][ 5s][DATA] lucchi-pp: train=8050 val=1750 test=1750 patches +[00:04:26][ 16s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, dros-vnc=2.77, multiclass=2.60 +[00:04:36][ 26s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 9s +[00:04:37][ 27s][WEIGHTS] mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 | eff.sources=1.44 | MMD2 mix=0.0050 uniform=0.2092 +[00:04:37][ 27s][WEIGHTS] per-source: fly-mito=0.013, kasthuri=0.022, dros-vnc=0.005, multiclass=0.960 | min-MMD source (NeuroADDA rule) = multiclass +[00:05:16][ 66s][TEACHERS] target-val Dice: fly-mito=0.243, kasthuri=0.161, dros-vnc=0.162, multiclass=0.833 +[00:05:16][ 66s][TEACHERS] ensemble target-val Dice: mixture=0.834, mixture@raw=0.833, mixture@logit=0.835, per_source=0.834, per_source@raw=0.833, per_source@logit=0.834, uniform=0.331, uniform@raw=0.184, uniform@logit=0.220 +[00:05:16][ 66s][WEIGHTS] using mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 +[00:05:16][ 66s][INIT] student initialised from teacher multiclass (weight 0.920) +[00:05:27][ 76s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[00:05:27][ 76s][KD] distilling the mixture ensemble (prob space) into the student +[00:07:29][ 199s][KD] soft labels for 6000 patches from 4 teachers in 123s +[00:10:29][ 379s][KD] epoch 1/5 | patches=6000 | loss=0.1185 | 179s +[00:13:31][ 561s][KD] epoch 2/5 | patches=6000 | loss=0.1064 | 182s +[00:16:30][ 740s][KD] epoch 3/5 | patches=6000 | loss=0.1043 | 179s +[00:19:29][ 918s][KD] epoch 4/5 | patches=6000 | loss=0.1035 | 179s +[00:22:28][ 1097s][KD] epoch 5/5 | patches=6000 | loss=0.1031 | 179s +[00:22:33][ 1103s][ROUND0] before any target label: val_dice=0.8827 val_vi=0.0617 +[00:22:33][ 1103s][AL] patches=8050 (pool 8050 after dropping blank; free blank negatives: True) | budget=402 (5.0%) | rounds=6 | per_round=67 | acquisition=laplace | first_round=stratified +[00:24:00][ 1190s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': 0.08900566294125338, 'auroc_top20_error': 0.471344873046875} +[00:24:42][ 1232s][AL] round 1: stratified cold start: 34 patches from the membrane-rich half (ensemble prevalence >= 0.074) + 33 from the membrane-poor half of 2000 candidates +[00:24:47][ 1237s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.932, uniform=0.354, top1=0.932, top2=0.932, top3=0.932 -> using mixture +[00:24:48][ 1238s][WEIGHTS] round 1: student kept (labeled Dice 0.940 vs best teacher multiclass 0.932); distilling the corrected ensemble into it +[00:24:48][ 1238s][WEIGHTS] round 1: teacher Dice on 67 labeled patches + 0 blank (352197 fg px): fly-mito=0.288, kasthuri=0.000, dros-vnc=0.023, multiclass=0.932 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:26:51][ 1361s][KD] soft labels for 6000 patches from 4 teachers in 123s +[00:29:50][ 1540s][KD] epoch 1/2 | patches=6000 | loss=0.0466 | 179s +[00:32:49][ 1719s][KD] epoch 2/2 | patches=6000 | loss=0.0459 | 179s +[00:33:30][ 1760s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 41s +[00:33:35][ 1765s][FT] round 1/6 epoch 1/5 | patches=67 | loss=0.3159 +[00:33:39][ 1769s][FT] round 1/6 epoch 2/5 | patches=67 | loss=0.3248 +[00:33:44][ 1774s][FT] round 1/6 epoch 3/5 | patches=67 | loss=0.3628 +[00:33:48][ 1778s][FT] round 1/6 epoch 4/5 | patches=67 | loss=0.3215 +[00:33:53][ 1783s][FT] round 1/6 epoch 5/5 | patches=67 | loss=0.2480 +[00:33:59][ 1789s][AL] round 1/6 | labeled=67 | selection_seconds=128.5000 round_seconds=679.8000 restart_student_dice=0.9398 restart_teacher_dice=0.9318 val_dice=0.8461 val_vi=0.0856 val_ece=0.0067 +[00:35:25][ 1875s][AL] round 2: 1183/7983 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[00:35:36][ 1886s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.819, uniform=0.253, top1=0.819, top2=0.819, top3=0.819 -> using mixture +[00:35:36][ 1886s][WEIGHTS] round 2: teacher Dice on 134 labeled patches + 0 blank (650845 fg px): fly-mito=0.246, kasthuri=0.000, dros-vnc=0.024, multiclass=0.819 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:35:40][ 1889s][CONS] round 2: on the 67 new patches student Dice=0.680 vs ensemble 0.689 -> pseudo-labels from the ensemble | consistency x1.00 +[00:35:48][ 1898s][FT] round 2/6 epoch 1/5 | patches=134 | loss=0.5413 +[00:35:56][ 1906s][FT] round 2/6 epoch 2/5 | patches=134 | loss=0.3978 +[00:36:05][ 1914s][FT] round 2/6 epoch 3/5 | patches=134 | loss=0.3414 +[00:36:13][ 1923s][FT] round 2/6 epoch 4/5 | patches=134 | loss=0.3627 +[00:36:22][ 1931s][FT] round 2/6 epoch 5/5 | patches=134 | loss=0.3053 +[00:36:27][ 1937s][AL] round 2/6 | labeled=134 | selection_seconds=86.6000 round_seconds=142.8000 pseudo_student_dice=0.6799 pseudo_ensemble_dice=0.6885 consistency_scale=1.0000 val_dice=0.8239 val_vi=0.0783 val_ece=0.0062 +[00:37:53][ 2023s][AL] round 3: 1286/7916 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[00:38:08][ 2038s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.826, uniform=0.199, top1=0.826, top2=0.826, top3=0.826 -> using mixture +[00:38:08][ 2038s][WEIGHTS] round 3: teacher Dice on 201 labeled patches + 0 blank (1077649 fg px): fly-mito=0.271, kasthuri=0.000, dros-vnc=0.016, multiclass=0.826 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:38:12][ 2042s][CONS] round 3: on the 67 new patches student Dice=0.843 vs ensemble 0.836 -> pseudo-labels from the ensemble | consistency x1.00 +[00:38:25][ 2054s][FT] round 3/6 epoch 1/5 | patches=201 | loss=0.3340 +[00:38:37][ 2067s][FT] round 3/6 epoch 2/5 | patches=201 | loss=0.3026 +[00:38:49][ 2079s][FT] round 3/6 epoch 3/5 | patches=201 | loss=0.2820 +[00:39:02][ 2092s][FT] round 3/6 epoch 4/5 | patches=201 | loss=0.2937 +[00:39:14][ 2104s][FT] round 3/6 epoch 5/5 | patches=201 | loss=0.2591 +[00:39:20][ 2110s][AL] round 3/6 | labeled=201 | selection_seconds=85.8000 round_seconds=167.1000 pseudo_student_dice=0.8429 pseudo_ensemble_dice=0.8358 consistency_scale=1.0000 val_dice=0.8594 val_vi=0.0649 val_ece=0.0037 +[00:40:45][ 2195s][AL] round 4: 1380/7849 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[00:41:05][ 2215s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.841, uniform=0.192, top1=0.841, top2=0.841, top3=0.841 -> using mixture +[00:41:05][ 2215s][WEIGHTS] round 4: teacher Dice on 268 labeled patches + 0 blank (1455704 fg px): fly-mito=0.281, kasthuri=0.001, dros-vnc=0.023, multiclass=0.841 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:41:09][ 2219s][CONS] round 4: on the 67 new patches student Dice=0.865 vs ensemble 0.884 -> pseudo-labels from the ensemble | consistency x1.00 +[00:41:25][ 2235s][FT] round 4/6 epoch 1/5 | patches=268 | loss=0.2903 +[00:41:41][ 2251s][FT] round 4/6 epoch 2/5 | patches=268 | loss=0.3076 +[00:41:58][ 2267s][FT] round 4/6 epoch 3/5 | patches=268 | loss=0.2736 +[00:42:14][ 2284s][FT] round 4/6 epoch 4/5 | patches=268 | loss=0.2825 +[00:42:30][ 2300s][FT] round 4/6 epoch 5/5 | patches=268 | loss=0.2638 +[00:42:36][ 2306s][AL] round 4/6 | labeled=268 | selection_seconds=84.9000 round_seconds=190.2000 pseudo_student_dice=0.8653 pseudo_ensemble_dice=0.8842 consistency_scale=1.0000 val_dice=0.8642 val_vi=0.0689 val_ece=0.0044 +[00:44:00][ 2390s][AL] round 5: 1217/7782 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[00:44:25][ 2415s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.849, uniform=0.205, top1=0.849, top2=0.849, top3=0.849 -> using mixture +[00:44:25][ 2415s][WEIGHTS] round 5: teacher Dice on 335 labeled patches + 0 blank (1867412 fg px): fly-mito=0.284, kasthuri=0.001, dros-vnc=0.023, multiclass=0.849 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:44:29][ 2419s][CONS] round 5: on the 67 new patches student Dice=0.875 vs ensemble 0.877 -> pseudo-labels from the ensemble | consistency x1.00 +[00:44:49][ 2439s][FT] round 5/6 epoch 1/5 | patches=335 | loss=0.2433 +[00:45:09][ 2459s][FT] round 5/6 epoch 2/5 | patches=335 | loss=0.2352 +[00:45:29][ 2479s][FT] round 5/6 epoch 3/5 | patches=335 | loss=0.2319 +[00:45:49][ 2499s][FT] round 5/6 epoch 4/5 | patches=335 | loss=0.2302 +[00:46:10][ 2519s][FT] round 5/6 epoch 5/5 | patches=335 | loss=0.2170 +[00:46:15][ 2525s][AL] round 5/6 | labeled=335 | selection_seconds=84.2000 round_seconds=213.6000 pseudo_student_dice=0.8754 pseudo_ensemble_dice=0.8775 consistency_scale=1.0000 val_dice=0.8630 val_vi=0.0682 val_ece=0.0053 +[00:47:39][ 2608s][AL] round 6: 1218/7715 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[00:48:09][ 2638s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.851, uniform=0.205, top1=0.851, top2=0.851, top3=0.851 -> using mixture +[00:48:09][ 2638s][WEIGHTS] round 6: teacher Dice on 402 labeled patches + 0 blank (2210520 fg px): fly-mito=0.279, kasthuri=0.000, dros-vnc=0.020, multiclass=0.851 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:48:13][ 2642s][CONS] round 6: on the 67 new patches student Dice=0.864 vs ensemble 0.862 -> pseudo-labels from the ensemble | consistency x1.00 +[00:48:37][ 2666s][FT] round 6/6 epoch 1/5 | patches=402 | loss=0.2212 +[00:49:01][ 2691s][FT] round 6/6 epoch 2/5 | patches=402 | loss=0.2089 +[00:49:25][ 2715s][FT] round 6/6 epoch 3/5 | patches=402 | loss=0.1954 +[00:49:49][ 2739s][FT] round 6/6 epoch 4/5 | patches=402 | loss=0.2278 +[00:50:14][ 2763s][FT] round 6/6 epoch 5/5 | patches=402 | loss=0.2029 +[00:50:19][ 2769s][AL] round 6/6 | labeled=402 | selection_seconds=83.5000 round_seconds=238.4000 pseudo_student_dice=0.8636 pseudo_ensemble_dice=0.8622 consistency_scale=1.0000 val_dice=0.8705 val_vi=0.0671 val_ece=0.0055 +[00:50:44][ 2793s][FT] final 1/10 epoch 1/1 | patches=402 | loss=0.1935 +[00:51:08][ 2818s][FT] final 2/10 epoch 1/1 | patches=402 | loss=0.1710 +[00:51:32][ 2842s][FT] final 3/10 epoch 1/1 | patches=402 | loss=0.1776 +[00:51:56][ 2866s][FT] final 4/10 epoch 1/1 | patches=402 | loss=0.1640 +[00:52:20][ 2890s][FT] final 5/10 epoch 1/1 | patches=402 | loss=0.1549 +[00:52:45][ 2915s][FT] final 6/10 epoch 1/1 | patches=402 | loss=0.1496 +[00:53:09][ 2939s][FT] final 7/10 epoch 1/1 | patches=402 | loss=0.1495 +[00:53:33][ 2963s][FT] final 8/10 epoch 1/1 | patches=402 | loss=0.1429 +[00:53:58][ 2987s][FT] final 9/10 epoch 1/1 | patches=402 | loss=0.1361 +[00:54:22][ 3012s][FT] final 10/10 epoch 1/1 | patches=402 | loss=0.1372 +[00:54:28][ 3017s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=242.4000 val_dice=0.8826 val_vi=0.0611 val_ece=0.0041 +[00:54:28][ 3017s][WEIGHTS] final: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[00:55:50][ 3099s][TEST] raw student (no EMA, no TTA): patch_dice=0.9136 patch_vi=0.0429 +[00:55:59][ 3109s][TEST] final model without TTA: patch_dice=0.9136 patch_vi=0.0429 +[00:55:59][ 3109s][RESOURCES] {"phases": {"setup": {"seconds": 8.23, "peak_cpu_rss_mb": 2169.5, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 60.15, "peak_cpu_rss_mb": 4165.8, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 1020.9, "peak_cpu_rss_mb": 5938.9, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4128.0, "kd_epochs": 5}, "active_learning": {"seconds": 1914.27, "peak_cpu_rss_mb": 5938.9, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4176.0, "rounds": 6}, "inference": {"seconds": 91.59, "peak_cpu_rss_mb": 5938.9, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4176.0, "patches": 1750, "images": 50, "patches_per_second": 19.1, "seconds_per_image": 1.832}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 5938.9, "total_seconds": 3095.1} +[00:55:59][ 3109s][TEST] patch_dice=0.9262 patch_iou=0.8974 patch_precision=0.9610 patch_recall=0.9315 patch_vi=0.0406 image_dice=0.9555 image_iou=0.9153 image_precision=0.9706 image_recall=0.9415 image_vi=0.0454 cal_ece=0.0018 cal_nll=0.0168 cal_brier=0.0035 cal_mean_confidence=0.9974 cal_accuracy=0.9956 cal_predicted_prevalence=0.0439 cal_true_prevalence=0.0452 cal_mean_probability=0.0442 +[00:56:00][ 3110s][DONE] lucchi-pp macs-v11-norestart-acq-laplace b5 s1 | test patch Dice=0.9262 VI=0.0406 | image Dice=0.9555 VI=0.0454 | 3109s diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..108c2562715df32bea8a6419c11332ad083a64ef --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json @@ -0,0 +1,647 @@ +{ + "target": "lucchi-pp", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "config": { + "task": "mitochondria", + "target": "lucchi-pp", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "fce9b15", + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "lucchi-pp", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "fce9b15", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "drosophila-vnc", + "multiclass" + ], + "alphas": [ + 0.03256907457032885, + 0.031801008835420135, + 0.015561893577512515, + 0.9200680230167385 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "drosophila-vnc", + "multiclass" + ], + "mixture": [ + 0.03256907457032885, + 0.031801008835420135, + 0.015561893577512515, + 0.9200680230167385 + ], + "per_source": [ + 0.013021514318439549, + 0.021594227384204207, + 0.00495364759466876, + 0.9604306107026874 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.001413505929905978, + "mmd2_mixture": 0.005039697348754646, + "mmd2_uniform": 0.2092228336519747, + "mmd2_single_source": [ + 0.5680513520444281, + 0.4644148125694437, + 0.6002511462494531, + 0.004698715465898062 + ], + "median_squared_distance": 0.1892747764462741, + "gammas": [ + 1.3208310409546982, + 2.6416620819093963, + 5.283324163818793, + 10.566648327637585, + 21.13329665527517 + ], + "max_weight": 0.9200680230167385, + "entropy": 0.36261914188486677, + "effective_sources": 1.4370884282110035, + "min_weight": 0.015561893577512515 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6675354526109456, + 0.5899267604637966, + 0.8158227740743191, + 0.007664790896468099 + ], + "standardized": [ + 0.4800153933276398, + 0.22710384776568787, + 0.9632547653742088, + -1.670374006467536 + ], + "tau": 0.5, + "max_weight": 0.9604306107026874, + "entropy": 0.2044174691373809, + "effective_sources": 1.2268102019632934, + "min_weight": 0.00495364759466876 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "multiclass", + "seconds": 10.6 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.24317352702666994, + 0.16070093323689605, + 0.16243146338599246, + 0.8333945099982429 + ], + "ensemble_val_dice": { + "mixture": 0.8341164139370107, + "mixture@raw": 0.8334634452984915, + "mixture@logit": 0.8348661307075531, + "per_source": 0.8335017830355866, + "per_source@raw": 0.8334355010464378, + "per_source@logit": 0.8337681001665089, + "uniform": 0.3312974448451058, + "uniform@raw": 0.18374038775198187, + "uniform@logit": 0.21963766526924644 + }, + "val_patches": 1024 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 2.767641305923462, + 2.597146987915039 + ], + "top_source": "multiclass", + "weight_stats": { + "max_weight": 0.9200680230167385, + "entropy": 0.36261914188486677, + "effective_sources": 1.4370884282110035, + "min_weight": 0.015561893577512515 + }, + "correlations": { + "spearman_weight_teacher_dice": 0.7999999999999999, + "spearman_weight_taxonomy": -0.2581988897471611, + "top_source": "multiclass", + "top_source_is_best_teacher": true + }, + "init_teacher": "multiclass", + "student_init": "top-source", + "blank_fraction": 0.0, + "kd_seconds": 1020.9, + "round0_val": { + "val_dice": 0.8827209277665129, + "val_vi": 0.06173844436601252, + "val_ece": 0.01930872331093795 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 67, + "selection_seconds": 128.5, + "round_seconds": 679.8, + "teacher_dice_labeled": [ + 0.2876389731008006, + 2.8393119779215102e-06, + 0.02286523788150808, + 0.931847537434372 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 352197, + "evidence_nonempty_patches": 62, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.9398350397937785, + "restart_teacher_dice": 0.931847537434372, + "weights": [ + 5.6389282479936555e-05, + 3.1020171205931166e-06, + 1.907901148148134e-06, + 0.9999386007992513 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.931847537434372, + "uniform": 0.3539312329080638, + "top1": 0.931847537434372, + "top2": 0.931847537434372, + "top3": 0.931847537434372 + }, + "val_dice": 0.8461434440443407, + "val_vi": 0.0855945925313224, + "val_ece": 0.006651857957124172 + }, + { + "round": 2, + "n_labeled": 134, + "selection_seconds": 86.6, + "round_seconds": 142.8, + "teacher_dice_labeled": [ + 0.24577679079476183, + 1.5364617743675156e-06, + 0.024486936416884684, + 0.8192993927643852 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 650845, + "evidence_nonempty_patches": 122, + "first_correction": false, + "restart": "", + "weights": [ + 0.00011432785797566688, + 9.558707631533626e-06, + 5.975296532053933e-06, + 0.9998701381378607 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8193014677864318, + "uniform": 0.2526532619810576, + "top1": 0.8192993927643852, + "top2": 0.8193014677864318, + "top3": 0.8193014677864318 + }, + "pseudo_student_dice": 0.6799073601500278, + "pseudo_ensemble_dice": 0.6885056543210284, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.823862031483791, + "val_vi": 0.07833448240855631, + "val_ece": 0.006199167541584895 + }, + { + "round": 3, + "n_labeled": 201, + "selection_seconds": 85.8, + "round_seconds": 167.1, + "teacher_dice_labeled": [ + 0.27051968258106773, + 9.271819140187124e-07, + 0.015514227110664483, + 0.8257373465675559 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1077649, + "evidence_nonempty_patches": 186, + "first_correction": false, + "restart": "", + "weights": [ + 0.00013729036451304996, + 8.962465766820798e-06, + 5.1217981496385855e-06, + 0.9998486253715705 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8257374489433349, + "uniform": 0.19946310223893535, + "top1": 0.8257373465675559, + "top2": 0.8257366798750658, + "top3": 0.8257374489433349 + }, + "pseudo_student_dice": 0.8428575797424998, + "pseudo_ensemble_dice": 0.8358479338209674, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.859391961227498, + "val_vi": 0.06487842490840803, + "val_ece": 0.003686388289243956 + }, + { + "round": 4, + "n_labeled": 268, + "selection_seconds": 84.9, + "round_seconds": 190.2, + "teacher_dice_labeled": [ + 0.2809361991370772, + 0.0006554222291767896, + 0.023371464715406718, + 0.8408604006205874 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1455704, + "evidence_nonempty_patches": 246, + "first_correction": false, + "restart": "", + "weights": [ + 0.0001309794754778175, + 7.755231705957663e-06, + 4.7628931689371864e-06, + 0.9998565023996473 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8408612401217148, + "uniform": 0.19158436581242347, + "top1": 0.8408604006205874, + "top2": 0.8408612401217148, + "top3": 0.8408612401217148 + }, + "pseudo_student_dice": 0.865318273934081, + "pseudo_ensemble_dice": 0.8841677834896886, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.864204841298776, + "val_vi": 0.0689438616669614, + "val_ece": 0.004439818108043122 + }, + { + "round": 5, + "n_labeled": 335, + "selection_seconds": 84.2, + "round_seconds": 213.6, + "teacher_dice_labeled": [ + 0.2835080031881515, + 0.0005110274916737301, + 0.02308174871596876, + 0.8489501264344202 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1867412, + "evidence_nonempty_patches": 310, + "first_correction": false, + "restart": "", + "weights": [ + 0.00012394890757176728, + 7.142297450423627e-06, + 4.380088376626105e-06, + 0.9998645287066013 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8489512174331817, + "uniform": 0.20478316977482933, + "top1": 0.8489501264344202, + "top2": 0.8489512174331817, + "top3": 0.8489512174331817 + }, + "pseudo_student_dice": 0.8753591055736171, + "pseudo_ensemble_dice": 0.8774800102230823, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.862984788458472, + "val_vi": 0.06815227306475158, + "val_ece": 0.005252702164533228 + }, + { + "round": 6, + "n_labeled": 402, + "selection_seconds": 83.5, + "round_seconds": 238.4, + "teacher_dice_labeled": [ + 0.27886655669468124, + 0.0004317570658965266, + 0.019823472474504847, + 0.8510156994214092 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 2210520, + "evidence_nonempty_patches": 377, + "first_correction": false, + "restart": "", + "weights": [ + 0.0001159092629984249, + 6.990796383176612e-06, + 4.153032658780695e-06, + 0.9998729469079597 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.851015583658806, + "uniform": 0.20455471975081585, + "top1": 0.8510156994214092, + "top2": 0.851015583658806, + "top3": 0.851015583658806 + }, + "pseudo_student_dice": 0.8635595641020113, + "pseudo_ensemble_dice": 0.8621746701279371, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8705280669376912, + "val_vi": 0.06705494363318357, + "val_ece": 0.005491697527176027 + }, + { + "round": 7, + "n_labeled": 402, + "selection_seconds": 0.0, + "round_seconds": 242.4, + "final_consolidation": true, + "val_dice": 0.8825650776052844, + "val_vi": 0.06109141014936791, + "val_ece": 0.004077543849137184 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.08900566294125338, + "auroc_top20_error": 0.471344873046875 + }, + "al_seconds": 1914.3, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 0.0001159092629984249, + 6.990796383176612e-06, + 4.153032658780695e-06, + 0.9998729469079597 + ], + "weight_stats_final": { + "max_weight": 0.9998729469079597, + "entropy": 0.0013119464075929815, + "effective_sources": 1.0013128073857591, + "min_weight": 4.153032658780695e-06 + }, + "top_source_final": "multiclass", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.91356207848983, + "patch_iou": 0.8831404748948558, + "patch_precision": 0.946976123955655, + "patch_recall": 0.930699016573354, + "patch_vi": 0.042884940354461186, + "n_patches": 1750, + "cal_ece": 0.0026736035516054433, + "cal_nll": 0.021422703820900835, + "cal_brier": 0.003958336581594274, + "cal_mean_confidence": 0.9978627776587483, + "cal_accuracy": 0.9951891741071428, + "cal_predicted_prevalence": 0.04423939732142857, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.04434355303823744 + }, + "test_notta": { + "patch_dice": 0.91356207848983, + "patch_iou": 0.8831404748948558, + "patch_precision": 0.946976123955655, + "patch_recall": 0.930699016573354, + "patch_vi": 0.042884940354461186, + "n_patches": 1750, + "cal_ece": 0.0026736035516054433, + "cal_nll": 0.021422703820900835, + "cal_brier": 0.003958336581594274, + "cal_mean_confidence": 0.9978627776587483, + "cal_accuracy": 0.9951891741071428, + "cal_predicted_prevalence": 0.04423939732142857, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.04434355303823744 + }, + "test": { + "patch_dice": 0.9261972077942084, + "patch_iou": 0.8974155611349277, + "patch_precision": 0.9609866432654087, + "patch_recall": 0.9315003450236838, + "patch_vi": 0.040560742745314356, + "n_patches": 1750, + "image_dice": 0.9555399724621937, + "image_iou": 0.9153154542048801, + "image_precision": 0.970564719786427, + "image_recall": 0.9415382349126573, + "image_vi": 0.04536973324369179, + "n_images": 50, + "cal_ece": 0.0018369080010505645, + "cal_nll": 0.01683470636319407, + "cal_brier": 0.0034860976003171663, + "cal_mean_confidence": 0.9973964473557894, + "cal_accuracy": 0.9955675223214285, + "cal_predicted_prevalence": 0.043865513392857144, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.04415598747573566 + }, + "eval_seconds": 91.6, + "total_seconds": 3108.8, + "resources": { + "phases": { + "setup": { + "seconds": 8.23, + "peak_cpu_rss_mb": 2169.5, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 60.15, + "peak_cpu_rss_mb": 4165.8, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 1020.9, + "peak_cpu_rss_mb": 5938.9, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4128.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 1914.27, + "peak_cpu_rss_mb": 5938.9, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4176.0, + "rounds": 6 + }, + "inference": { + "seconds": 91.59, + "peak_cpu_rss_mb": 5938.9, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4176.0, + "patches": 1750, + "images": 50, + "patches_per_second": 19.1, + "seconds_per_image": 1.832 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 5938.9, + "total_seconds": 3095.1 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/lucchi-pp/macs-v11-norestart-acq-laplace_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..fe8f7308472e17c64606c1b58a68604142a8c5ac --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/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 +1,67,128.5,679.8,"[0.2876389731008006, 2.8393119779215102e-06, 0.02286523788150808, 0.931847537434372]",,352197,62,True,kept,0.9398350397937785,0.931847537434372,"[5.6389282479936555e-05, 3.1020171205931166e-06, 1.907901148148134e-06, 0.9999386007992513]",mixture,"{'mixture': 0.931847537434372, 'uniform': 0.3539312329080638, 'top1': 0.931847537434372, 'top2': 0.931847537434372, 'top3': 0.931847537434372}",0.8461434440443407,0.0855945925313224,0.006651857957124172,,,,, +2,134,86.6,142.8,"[0.24577679079476183, 1.5364617743675156e-06, 0.024486936416884684, 0.8192993927643852]",,650845,122,False,,,,"[0.00011432785797566688, 9.558707631533626e-06, 5.975296532053933e-06, 0.9998701381378607]",mixture,"{'mixture': 0.8193014677864318, 'uniform': 0.2526532619810576, 'top1': 0.8192993927643852, 'top2': 0.8193014677864318, 'top3': 0.8193014677864318}",0.823862031483791,0.07833448240855631,0.006199167541584895,0.6799073601500278,0.6885056543210284,ensemble,1.0, +3,201,85.8,167.1,"[0.27051968258106773, 9.271819140187124e-07, 0.015514227110664483, 0.8257373465675559]",,1077649,186,False,,,,"[0.00013729036451304996, 8.962465766820798e-06, 5.1217981496385855e-06, 0.9998486253715705]",mixture,"{'mixture': 0.8257374489433349, 'uniform': 0.19946310223893535, 'top1': 0.8257373465675559, 'top2': 0.8257366798750658, 'top3': 0.8257374489433349}",0.859391961227498,0.06487842490840803,0.003686388289243956,0.8428575797424998,0.8358479338209674,ensemble,1.0, +4,268,84.9,190.2,"[0.2809361991370772, 0.0006554222291767896, 0.023371464715406718, 0.8408604006205874]",,1455704,246,False,,,,"[0.0001309794754778175, 7.755231705957663e-06, 4.7628931689371864e-06, 0.9998565023996473]",mixture,"{'mixture': 0.8408612401217148, 'uniform': 0.19158436581242347, 'top1': 0.8408604006205874, 'top2': 0.8408612401217148, 'top3': 0.8408612401217148}",0.864204841298776,0.0689438616669614,0.004439818108043122,0.865318273934081,0.8841677834896886,ensemble,1.0, +5,335,84.2,213.6,"[0.2835080031881515, 0.0005110274916737301, 0.02308174871596876, 0.8489501264344202]",,1867412,310,False,,,,"[0.00012394890757176728, 7.142297450423627e-06, 4.380088376626105e-06, 0.9998645287066013]",mixture,"{'mixture': 0.8489512174331817, 'uniform': 0.20478316977482933, 'top1': 0.8489501264344202, 'top2': 0.8489512174331817, 'top3': 0.8489512174331817}",0.862984788458472,0.06815227306475158,0.005252702164533228,0.8753591055736171,0.8774800102230823,ensemble,1.0, +6,402,83.5,238.4,"[0.27886655669468124, 0.0004317570658965266, 0.019823472474504847, 0.8510156994214092]",,2210520,377,False,,,,"[0.0001159092629984249, 6.990796383176612e-06, 4.153032658780695e-06, 0.9998729469079597]",mixture,"{'mixture': 0.851015583658806, 'uniform': 0.20455471975081585, 'top1': 0.8510156994214092, 'top2': 0.851015583658806, 'top3': 0.851015583658806}",0.8705280669376912,0.06705494363318357,0.005491697527176027,0.8635595641020113,0.8621746701279371,ensemble,1.0, +7,402,0.0,242.4,,,,,,,,,,,,0.8825650776052844,0.06109141014936791,0.004077543849137184,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..59632c24d67e7f774056ce6f03c579a27f534945 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/log.txt @@ -0,0 +1,107 @@ +[12:16:33][ 0s][START] target=lucchi-pp variant=macs-v11-norestart-acq-laplace budget=5.0% seed=2 | {"task": "mitochondria", "target": "lucchi-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": 8, "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} +[12:16:33][ 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. +[12:16:33][ 0s][DATA] lucchi-pp: train=8050 val=1750 test=1750 patches +[12:16:35][ 2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, dros-vnc=2.77, multiclass=2.60 +[12:16:35][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[12:16:36][ 3s][WEIGHTS] mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 | eff.sources=1.44 | MMD2 mix=0.0050 uniform=0.2092 +[12:16:36][ 3s][WEIGHTS] per-source: fly-mito=0.013, kasthuri=0.022, dros-vnc=0.005, multiclass=0.960 | min-MMD source (NeuroADDA rule) = multiclass +[12:17:14][ 41s][TEACHERS] target-val Dice: fly-mito=0.243, kasthuri=0.161, dros-vnc=0.162, multiclass=0.833 +[12:17:14][ 41s][TEACHERS] ensemble target-val Dice: mixture=0.834, mixture@raw=0.833, mixture@logit=0.835, per_source=0.834, per_source@raw=0.833, per_source@logit=0.834, uniform=0.331, uniform@raw=0.184, uniform@logit=0.220 +[12:17:14][ 41s][WEIGHTS] using mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 +[12:17:14][ 41s][INIT] student initialised from teacher multiclass (weight 0.920) +[12:17:17][ 44s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[12:17:17][ 44s][KD] distilling the mixture ensemble (prob space) into the student +[12:19:20][ 167s][KD] soft labels for 6000 patches from 4 teachers in 123s +[12:22:19][ 346s][KD] epoch 1/5 | patches=6000 | loss=0.1187 | 179s +[12:25:18][ 525s][KD] epoch 2/5 | patches=6000 | loss=0.1069 | 179s +[12:28:17][ 704s][KD] epoch 3/5 | patches=6000 | loss=0.1048 | 179s +[12:31:16][ 883s][KD] epoch 4/5 | patches=6000 | loss=0.1037 | 179s +[12:34:15][ 1062s][KD] epoch 5/5 | patches=6000 | loss=0.1029 | 179s +[12:34:21][ 1068s][ROUND0] before any target label: val_dice=0.8804 val_vi=0.0672 +[12:34:21][ 1068s][AL] patches=8050 (pool 8050 after dropping blank; free blank negatives: True) | budget=402 (5.0%) | rounds=6 | per_round=67 | acquisition=laplace | first_round=stratified +[12:35:48][ 1155s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': 0.09212662440628899, 'auroc_top20_error': 0.47728173828125} +[12:36:29][ 1197s][AL] round 1: stratified cold start: 34 patches from the membrane-rich half (ensemble prevalence >= 0.072) + 33 from the membrane-poor half of 2000 candidates +[12:36:35][ 1202s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.946, uniform=0.320, top1=0.946, top2=0.946, top3=0.946 -> using mixture +[12:36:36][ 1203s][WEIGHTS] round 1: student kept (labeled Dice 0.948 vs best teacher multiclass 0.946); distilling the corrected ensemble into it +[12:36:36][ 1203s][WEIGHTS] round 1: teacher Dice on 67 labeled patches + 0 blank (313095 fg px): fly-mito=0.265, kasthuri=0.000, dros-vnc=0.016, multiclass=0.946 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[12:38:39][ 1327s][KD] soft labels for 6000 patches from 4 teachers in 123s +[12:41:38][ 1505s][KD] epoch 1/2 | patches=6000 | loss=0.0470 | 179s +[12:44:37][ 1684s][KD] epoch 2/2 | patches=6000 | loss=0.0482 | 179s +[12:45:19][ 1726s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 41s +[12:45:23][ 1730s][FT] round 1/6 epoch 1/5 | patches=67 | loss=0.3215 +[12:45:28][ 1735s][FT] round 1/6 epoch 2/5 | patches=67 | loss=0.3800 +[12:45:33][ 1740s][FT] round 1/6 epoch 3/5 | patches=67 | loss=0.3388 +[12:45:38][ 1745s][FT] round 1/6 epoch 4/5 | patches=67 | loss=0.3076 +[12:45:43][ 1750s][FT] round 1/6 epoch 5/5 | patches=67 | loss=0.2970 +[12:45:48][ 1755s][AL] round 1/6 | labeled=67 | selection_seconds=128.6000 round_seconds=681.7000 restart_student_dice=0.9485 restart_teacher_dice=0.9463 val_dice=0.8306 val_vi=0.0835 val_ece=0.0068 +[12:47:15][ 1842s][AL] round 2: 1074/7983 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:47:26][ 1853s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.837, uniform=0.226, top1=0.837, top2=0.837, top3=0.837 -> using mixture +[12:47:26][ 1853s][WEIGHTS] round 2: teacher Dice on 134 labeled patches + 0 blank (646357 fg px): fly-mito=0.242, kasthuri=0.000, dros-vnc=0.013, multiclass=0.837 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[12:47:30][ 1857s][CONS] round 2: on the 67 new patches student Dice=0.706 vs ensemble 0.732 -> pseudo-labels from the ensemble | consistency x1.00 +[12:47:38][ 1865s][FT] round 2/6 epoch 1/5 | patches=134 | loss=0.4784 +[12:47:47][ 1874s][FT] round 2/6 epoch 2/5 | patches=134 | loss=0.4344 +[12:47:55][ 1883s][FT] round 2/6 epoch 3/5 | patches=134 | loss=0.4149 +[12:48:04][ 1891s][FT] round 2/6 epoch 4/5 | patches=134 | loss=0.3680 +[12:48:13][ 1900s][FT] round 2/6 epoch 5/5 | patches=134 | loss=0.3755 +[12:48:19][ 1906s][AL] round 2/6 | labeled=134 | selection_seconds=86.7000 round_seconds=144.5000 pseudo_student_dice=0.7064 pseudo_ensemble_dice=0.7318 consistency_scale=1.0000 val_dice=0.8366 val_vi=0.0717 val_ece=0.0038 +[12:49:45][ 1992s][AL] round 3: 1032/7916 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:50:01][ 2008s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.817, uniform=0.202, top1=0.817, top2=0.817, top3=0.817 -> using mixture +[12:50:01][ 2008s][WEIGHTS] round 3: teacher Dice on 201 labeled patches + 0 blank (961409 fg px): fly-mito=0.243, kasthuri=0.000, dros-vnc=0.012, multiclass=0.817 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[12:50:16][ 2023s][CONS] round 3: on the 67 new patches student Dice=0.809 vs ensemble 0.777 -> pseudo-labels from the student | consistency x1.00 +[12:50:29][ 2036s][FT] round 3/6 epoch 1/5 | patches=201 | loss=0.3748 +[12:50:42][ 2049s][FT] round 3/6 epoch 2/5 | patches=201 | loss=0.3606 +[12:50:54][ 2062s][FT] round 3/6 epoch 3/5 | patches=201 | loss=0.3386 +[12:51:07][ 2074s][FT] round 3/6 epoch 4/5 | patches=201 | loss=0.3334 +[12:51:20][ 2087s][FT] round 3/6 epoch 5/5 | patches=201 | loss=0.2969 +[12:51:26][ 2093s][AL] round 3/6 | labeled=201 | selection_seconds=86.9000 round_seconds=181.5000 pseudo_student_dice=0.8086 pseudo_ensemble_dice=0.7771 consistency_scale=1.0000 val_dice=0.8521 val_vi=0.0731 val_ece=0.0066 +[12:52:52][ 2179s][AL] round 4: 1253/7849 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[12:54:27][ 2274s][WEIGHTS] round 4: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.291, kasthuri=0.000, dros-vnc=0.023, multiclass=0.952 | blended with labeled Dice at 0.50 +[12:55:52][ 2359s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.891, uniform=0.264, top1=0.891, top2=0.891, top3=0.891 -> using mixture +[12:55:52][ 2359s][WEIGHTS] round 4: teacher Dice on 268 labeled patches + 0 blank (1406676 fg px): fly-mito=0.277, kasthuri=0.000, dros-vnc=0.018, multiclass=0.891 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[12:56:06][ 2373s][CONS] round 4: on the 67 new patches student Dice=0.854 vs ensemble 0.859 -> pseudo-labels from the student | consistency x1.00 +[12:56:23][ 2390s][FT] round 4/6 epoch 1/5 | patches=268 | loss=0.3130 +[12:56:39][ 2406s][FT] round 4/6 epoch 2/5 | patches=268 | loss=0.2841 +[12:56:56][ 2423s][FT] round 4/6 epoch 3/5 | patches=268 | loss=0.3083 +[12:57:13][ 2440s][FT] round 4/6 epoch 4/5 | patches=268 | loss=0.2707 +[12:57:29][ 2456s][FT] round 4/6 epoch 5/5 | patches=268 | loss=0.2564 +[12:57:35][ 2462s][AL] round 4/6 | labeled=268 | selection_seconds=85.8000 round_seconds=363.0000 pseudo_student_dice=0.8540 pseudo_ensemble_dice=0.8587 consistency_scale=1.0000 val_dice=0.8240 val_vi=0.0822 val_ece=0.0069 +[12:59:00][ 2547s][AL] round 5: 1164/7782 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:00:37][ 2644s][WEIGHTS] round 5: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.283, kasthuri=0.000, dros-vnc=0.022, multiclass=0.948 | blended with labeled Dice at 0.50 +[13:02:05][ 2732s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.893, uniform=0.261, top1=0.893, top2=0.893, top3=0.893 -> using mixture +[13:02:05][ 2732s][WEIGHTS] round 5: teacher Dice on 335 labeled patches + 0 blank (1792685 fg px): fly-mito=0.276, kasthuri=0.000, dros-vnc=0.017, multiclass=0.893 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[13:02:19][ 2746s][CONS] round 5: on the 67 new patches student Dice=0.855 vs ensemble 0.872 -> pseudo-labels from the student | consistency x1.00 +[13:02:40][ 2767s][FT] round 5/6 epoch 1/5 | patches=335 | loss=0.2735 +[13:03:00][ 2787s][FT] round 5/6 epoch 2/5 | patches=335 | loss=0.2802 +[13:03:21][ 2808s][FT] round 5/6 epoch 3/5 | patches=335 | loss=0.2619 +[13:03:41][ 2828s][FT] round 5/6 epoch 4/5 | patches=335 | loss=0.2487 +[13:04:02][ 2849s][FT] round 5/6 epoch 5/5 | patches=335 | loss=0.2306 +[13:04:07][ 2855s][AL] round 5/6 | labeled=335 | selection_seconds=84.9000 round_seconds=386.6000 pseudo_student_dice=0.8549 pseudo_ensemble_dice=0.8723 consistency_scale=1.0000 val_dice=0.8753 val_vi=0.0637 val_ece=0.0034 +[13:05:32][ 2939s][AL] round 6: 1238/7715 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[13:07:11][ 3038s][WEIGHTS] round 6: teacher agreement with the student on 2000 unlabeled patches: fly-mito=0.282, kasthuri=0.001, dros-vnc=0.018, multiclass=0.957 | blended with labeled Dice at 0.50 +[13:08:39][ 3126s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.902, uniform=0.267, top1=0.902, top2=0.902, top3=0.902 -> using mixture +[13:08:39][ 3126s][WEIGHTS] round 6: teacher Dice on 402 labeled patches + 0 blank (2180202 fg px): fly-mito=0.278, kasthuri=0.000, dros-vnc=0.015, multiclass=0.902 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[13:08:54][ 3141s][CONS] round 6: on the 67 new patches student Dice=0.892 vs ensemble 0.889 -> pseudo-labels from the student | consistency x1.00 +[13:09:18][ 3165s][FT] round 6/6 epoch 1/5 | patches=402 | loss=0.2037 +[13:09:43][ 3190s][FT] round 6/6 epoch 2/5 | patches=402 | loss=0.2282 +[13:10:07][ 3214s][FT] round 6/6 epoch 3/5 | patches=402 | loss=0.2195 +[13:10:32][ 3239s][FT] round 6/6 epoch 4/5 | patches=402 | loss=0.2240 +[13:10:56][ 3263s][FT] round 6/6 epoch 5/5 | patches=402 | loss=0.2315 +[13:11:02][ 3269s][AL] round 6/6 | labeled=402 | selection_seconds=84.4000 round_seconds=409.0000 pseudo_student_dice=0.8924 pseudo_ensemble_dice=0.8893 consistency_scale=1.0000 val_dice=0.8780 val_vi=0.0607 val_ece=0.0039 +[13:11:13][ 3280s][CONS] final consolidation: pseudo-labels refreshed from the student +[13:11:37][ 3304s][FT] final 1/10 epoch 1/1 | patches=402 | loss=0.2030 +[13:12:02][ 3329s][FT] final 2/10 epoch 1/1 | patches=402 | loss=0.1907 +[13:12:26][ 3353s][FT] final 3/10 epoch 1/1 | patches=402 | loss=0.1952 +[13:12:51][ 3378s][FT] final 4/10 epoch 1/1 | patches=402 | loss=0.1823 +[13:13:15][ 3402s][FT] final 5/10 epoch 1/1 | patches=402 | loss=0.1797 +[13:13:40][ 3427s][FT] final 6/10 epoch 1/1 | patches=402 | loss=0.1618 +[13:14:04][ 3451s][FT] final 7/10 epoch 1/1 | patches=402 | loss=0.1515 +[13:14:29][ 3476s][FT] final 8/10 epoch 1/1 | patches=402 | loss=0.1343 +[13:14:53][ 3500s][FT] final 9/10 epoch 1/1 | patches=402 | loss=0.1320 +[13:15:18][ 3525s][FT] final 10/10 epoch 1/1 | patches=402 | loss=0.1291 +[13:15:24][ 3531s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=255.7000 val_dice=0.8830 val_vi=0.0588 val_ece=0.0039 +[13:15:24][ 3531s][WEIGHTS] final: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[13:16:47][ 3614s][TEST] raw student (no EMA, no TTA): patch_dice=0.9102 patch_vi=0.0424 +[13:16:56][ 3624s][TEST] final model without TTA: patch_dice=0.9102 patch_vi=0.0424 +[13:16:56][ 3624s][RESOURCES] {"phases": {"setup": {"seconds": 0.8, "peak_cpu_rss_mb": 2160.2, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 42.4, "peak_cpu_rss_mb": 4175.1, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 1017.99, "peak_cpu_rss_mb": 5948.8, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4128.0, "kd_epochs": 5}, "active_learning": {"seconds": 2463.01, "peak_cpu_rss_mb": 5948.8, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4176.0, "rounds": 6}, "inference": {"seconds": 92.62, "peak_cpu_rss_mb": 5948.8, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4176.0, "patches": 1750, "images": 50, "patches_per_second": 18.9, "seconds_per_image": 1.852}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 5948.8, "total_seconds": 3616.8} +[13:16:56][ 3624s][TEST] patch_dice=0.9270 patch_iou=0.8989 patch_precision=0.9584 patch_recall=0.9368 patch_vi=0.0392 image_dice=0.9589 image_iou=0.9214 image_precision=0.9696 image_recall=0.9490 image_vi=0.0428 cal_ece=0.0017 cal_nll=0.0162 cal_brier=0.0033 cal_mean_confidence=0.9974 cal_accuracy=0.9958 cal_predicted_prevalence=0.0443 cal_true_prevalence=0.0452 cal_mean_probability=0.0447 +[13:16:57][ 3624s][DONE] lucchi-pp macs-v11-norestart-acq-laplace b5 s2 | test patch Dice=0.9270 VI=0.0392 | image Dice=0.9589 VI=0.0428 | 3624s diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..800e60f4af64a902cea2b4e4c1c8642f71ddafe4 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/result.json @@ -0,0 +1,662 @@ +{ + "target": "lucchi-pp", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 2, + "config": { + "task": "mitochondria", + "target": "lucchi-pp", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "lucchi-pp", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "drosophila-vnc", + "multiclass" + ], + "alphas": [ + 0.03256907457032885, + 0.031801008835420135, + 0.015561893577512515, + 0.9200680230167385 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "drosophila-vnc", + "multiclass" + ], + "mixture": [ + 0.03256907457032885, + 0.031801008835420135, + 0.015561893577512515, + 0.9200680230167385 + ], + "per_source": [ + 0.013021514318439549, + 0.021594227384204207, + 0.00495364759466876, + 0.9604306107026874 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.001413505929905978, + "mmd2_mixture": 0.005039697348754646, + "mmd2_uniform": 0.2092228336519747, + "mmd2_single_source": [ + 0.5680513520444281, + 0.4644148125694437, + 0.6002511462494531, + 0.004698715465898062 + ], + "median_squared_distance": 0.1892747764462741, + "gammas": [ + 1.3208310409546982, + 2.6416620819093963, + 5.283324163818793, + 10.566648327637585, + 21.13329665527517 + ], + "max_weight": 0.9200680230167385, + "entropy": 0.36261914188486677, + "effective_sources": 1.4370884282110035, + "min_weight": 0.015561893577512515 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6675354526109456, + 0.5899267604637966, + 0.8158227740743191, + 0.007664790896468099 + ], + "standardized": [ + 0.4800153933276398, + 0.22710384776568787, + 0.9632547653742088, + -1.670374006467536 + ], + "tau": 0.5, + "max_weight": 0.9604306107026874, + "entropy": 0.2044174691373809, + "effective_sources": 1.2268102019632934, + "min_weight": 0.00495364759466876 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "multiclass", + "seconds": 1.4 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.24317352702666994, + 0.16070093323689605, + 0.16243146338599246, + 0.8333945099982429 + ], + "ensemble_val_dice": { + "mixture": 0.8341164139370107, + "mixture@raw": 0.8334634452984915, + "mixture@logit": 0.8348661307075531, + "per_source": 0.8335017830355866, + "per_source@raw": 0.8334355010464378, + "per_source@logit": 0.8337681001665089, + "uniform": 0.3312974448451058, + "uniform@raw": 0.18374038775198187, + "uniform@logit": 0.21963766526924644 + }, + "val_patches": 1024 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 2.767641305923462, + 2.597146987915039 + ], + "top_source": "multiclass", + "weight_stats": { + "max_weight": 0.9200680230167385, + "entropy": 0.36261914188486677, + "effective_sources": 1.4370884282110035, + "min_weight": 0.015561893577512515 + }, + "correlations": { + "spearman_weight_teacher_dice": 0.7999999999999999, + "spearman_weight_taxonomy": -0.2581988897471611, + "top_source": "multiclass", + "top_source_is_best_teacher": true + }, + "init_teacher": "multiclass", + "student_init": "top-source", + "blank_fraction": 0.0, + "kd_seconds": 1018.0, + "round0_val": { + "val_dice": 0.8803879820113556, + "val_vi": 0.06716140467912256, + "val_ece": 0.01670684602633669 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 67, + "selection_seconds": 128.6, + "round_seconds": 681.7, + "teacher_dice_labeled": [ + 0.26456266329153644, + 3.1912075848621877e-06, + 0.015528068385600497, + 0.9463065640641752 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 313095, + "evidence_nonempty_patches": 59, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.9484527016764064, + "restart_teacher_dice": 0.9463065640641752, + "weights": [ + 3.8742771162253374e-05, + 2.684472576277311e-06, + 1.5342785029993225e-06, + 0.9999570384777584 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9463065640641752, + "uniform": 0.3201075901937694, + "top1": 0.9463065640641752, + "top2": 0.9463065640641752, + "top3": 0.9463065640641752 + }, + "val_dice": 0.8306197011240201, + "val_vi": 0.08348743124097768, + "val_ece": 0.006788817784942072 + }, + { + "round": 2, + "n_labeled": 134, + "selection_seconds": 86.7, + "round_seconds": 144.5, + "teacher_dice_labeled": [ + 0.24219815146629306, + 1.5464961809276811e-06, + 0.012802129960691319, + 0.8367222328254639 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 646357, + "evidence_nonempty_patches": 119, + "first_correction": false, + "restart": "", + "weights": [ + 9.267327807422955e-05, + 8.03051714291749e-06, + 4.466396604498123e-06, + 0.9998948298081783 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8367218370151065, + "uniform": 0.22632976686687015, + "top1": 0.8367222328254639, + "top2": 0.8367218370151065, + "top3": 0.8367218370151065 + }, + "pseudo_student_dice": 0.7063903969579743, + "pseudo_ensemble_dice": 0.7317769245256655, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8365702659977134, + "val_vi": 0.07170842372260706, + "val_ece": 0.003773437115060449 + }, + { + "round": 3, + "n_labeled": 201, + "selection_seconds": 86.9, + "round_seconds": 181.5, + "teacher_dice_labeled": [ + 0.2432412176643748, + 1.0398523409675826e-06, + 0.011922198046396285, + 0.8167738623605225 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 961409, + "evidence_nonempty_patches": 179, + "first_correction": false, + "restart": "", + "weights": [ + 0.00011431641553317873, + 9.803144569781586e-06, + 5.404554281086763e-06, + 0.9998704758856161 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8167700660471827, + "uniform": 0.2017762310498385, + "top1": 0.8167738623605225, + "top2": 0.8167700660471827, + "top3": 0.8167700660471827 + }, + "pseudo_student_dice": 0.8085989086638157, + "pseudo_ensemble_dice": 0.7770533684262554, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8520933668114993, + "val_vi": 0.07309227793276327, + "val_ece": 0.006567630992236721 + }, + { + "round": 4, + "n_labeled": 268, + "selection_seconds": 85.8, + "round_seconds": 363.0, + "teacher_dice_labeled": [ + 0.2774407873320246, + 0.00021757778155051012, + 0.017681862132876883, + 0.8908474538075571 + ], + "teacher_dice_pool": [ + 0.2914352298181363, + 0.00043444480174767376, + 0.022986299099895986, + 0.9522409389197227 + ], + "evidence_fg_pixels": 1406676, + "evidence_nonempty_patches": 245, + "first_correction": false, + "restart": "", + "weights": [ + 7.6728663456923e-05, + 4.684119979198459e-06, + 2.729580942504572e-06, + 0.9999158576356213 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8908468921531416, + "uniform": 0.26357952088220327, + "top1": 0.8908474538075571, + "top2": 0.8908466768504242, + "top3": 0.8908466745253185 + }, + "pseudo_student_dice": 0.8540059637638082, + "pseudo_ensemble_dice": 0.8586939563953454, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8240250559424885, + "val_vi": 0.08222283438241113, + "val_ece": 0.006853722569206671 + }, + { + "round": 5, + "n_labeled": 335, + "selection_seconds": 84.9, + "round_seconds": 386.6, + "teacher_dice_labeled": [ + 0.27581865697076663, + 0.00021098453816418617, + 0.01700586078043835, + 0.8932543310747327 + ], + "teacher_dice_pool": [ + 0.2830548899311145, + 0.00042141133660207754, + 0.0218039269162802, + 0.9478203037556284 + ], + "evidence_fg_pixels": 1792685, + "evidence_nonempty_patches": 307, + "first_correction": false, + "restart": "", + "weights": [ + 7.369894582484202e-05, + 4.572438172123279e-06, + 2.646723698279208e-06, + 0.9999190818923047 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.8932545092013533, + "uniform": 0.26145255935947603, + "top1": 0.8932543310747327, + "top2": 0.8932540626415855, + "top3": 0.8932542293319554 + }, + "pseudo_student_dice": 0.854906715658334, + "pseudo_ensemble_dice": 0.8722805399151158, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8753142915497427, + "val_vi": 0.06372080616488156, + "val_ece": 0.0033989463018334635 + }, + { + "round": 6, + "n_labeled": 402, + "selection_seconds": 84.4, + "round_seconds": 409.0, + "teacher_dice_labeled": [ + 0.2778182744095054, + 0.00028432021766136274, + 0.01471529608171681, + 0.9022960378346366 + ], + "teacher_dice_pool": [ + 0.2821117277570612, + 0.0005681818181818182, + 0.01807491548267771, + 0.9568074274579239 + ], + "evidence_fg_pixels": 2180202, + "evidence_nonempty_patches": 371, + "first_correction": false, + "restart": "", + "weights": [ + 6.86879160649449e-05, + 4.180239560132354e-06, + 2.3631737087619997e-06, + 0.9999247686706662 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9022957193488081, + "uniform": 0.26732216329589775, + "top1": 0.9022960378346366, + "top2": 0.9022954553953105, + "top3": 0.9022956474829681 + }, + "pseudo_student_dice": 0.8923665379571899, + "pseudo_ensemble_dice": 0.8892548845825352, + "pseudo_source": "student", + "consistency_scale": 1.0, + "val_dice": 0.8780199357605633, + "val_vi": 0.06069790674314873, + "val_ece": 0.0038901085719583125 + }, + { + "round": 7, + "n_labeled": 402, + "selection_seconds": 0.0, + "round_seconds": 255.7, + "final_consolidation": true, + "val_dice": 0.8830426902129573, + "val_vi": 0.058771422313978854, + "val_ece": 0.003914845060847729 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.09212662440628899, + "auroc_top20_error": 0.47728173828125 + }, + "al_seconds": 2463.0, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 6.86879160649449e-05, + 4.180239560132354e-06, + 2.3631737087619997e-06, + 0.9999247686706662 + ], + "weight_stats_final": { + "max_weight": 0.9999247686706662, + "entropy": 0.0008160555233599509, + "effective_sources": 1.0008163885872619, + "min_weight": 2.3631737087619997e-06 + }, + "top_source_final": "multiclass", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.9101973027773687, + "patch_iou": 0.8799479153075589, + "patch_precision": 0.9401215968693091, + "patch_recall": 0.9321027238495903, + "patch_vi": 0.042363949100457855, + "n_patches": 1750, + "cal_ece": 0.0028486549461974936, + "cal_nll": 0.022192469784087544, + "cal_brier": 0.003855580196129815, + "cal_mean_confidence": 0.9981672933390547, + "cal_accuracy": 0.9953186383928572, + "cal_predicted_prevalence": 0.04432533482142857, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.044275469432547426 + }, + "test_notta": { + "patch_dice": 0.9101973027773687, + "patch_iou": 0.8799479153075589, + "patch_precision": 0.9401215968693091, + "patch_recall": 0.9321027238495903, + "patch_vi": 0.042363949100457855, + "n_patches": 1750, + "cal_ece": 0.0028486549461974936, + "cal_nll": 0.022192469784087544, + "cal_brier": 0.003855580196129815, + "cal_mean_confidence": 0.9981672933390547, + "cal_accuracy": 0.9953186383928572, + "cal_predicted_prevalence": 0.04432533482142857, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.044275469432547426 + }, + "test": { + "patch_dice": 0.9270490015694021, + "patch_iou": 0.8989149479160026, + "patch_precision": 0.9583615572217189, + "patch_recall": 0.9367646860294109, + "patch_vi": 0.03922821826799574, + "n_patches": 1750, + "image_dice": 0.9588872499883614, + "image_iou": 0.9213606570440477, + "image_precision": 0.9695966441984472, + "image_recall": 0.9489955116436768, + "image_vi": 0.0427693149732126, + "n_images": 50, + "cal_ece": 0.0016793533075367766, + "cal_nll": 0.01617886206972107, + "cal_brier": 0.0033182671865588133, + "cal_mean_confidence": 0.9974321434861084, + "cal_accuracy": 0.9957527901785714, + "cal_predicted_prevalence": 0.044307477678571426, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.044659475159475076 + }, + "eval_seconds": 92.6, + "total_seconds": 3623.5, + "resources": { + "phases": { + "setup": { + "seconds": 0.8, + "peak_cpu_rss_mb": 2160.2, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 42.4, + "peak_cpu_rss_mb": 4175.1, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 1017.99, + "peak_cpu_rss_mb": 5948.8, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4128.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 2463.01, + "peak_cpu_rss_mb": 5948.8, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4176.0, + "rounds": 6 + }, + "inference": { + "seconds": 92.62, + "peak_cpu_rss_mb": 5948.8, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4176.0, + "patches": 1750, + "images": 50, + "patches_per_second": 18.9, + "seconds_per_image": 1.852 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 5948.8, + "total_seconds": 3616.8 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/lucchi-pp/macs-v11-norestart-acq-laplace_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-laplace/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..8f14b887bc6556c7989429f4dcf916b202d7e6fd --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/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 +1,67,128.6,681.7,"[0.26456266329153644, 3.1912075848621877e-06, 0.015528068385600497, 0.9463065640641752]",,313095,59,True,kept,0.9484527016764064,0.9463065640641752,"[3.8742771162253374e-05, 2.684472576277311e-06, 1.5342785029993225e-06, 0.9999570384777584]",mixture,"{'mixture': 0.9463065640641752, 'uniform': 0.3201075901937694, 'top1': 0.9463065640641752, 'top2': 0.9463065640641752, 'top3': 0.9463065640641752}",0.8306197011240201,0.08348743124097768,0.006788817784942072,,,,, +2,134,86.7,144.5,"[0.24219815146629306, 1.5464961809276811e-06, 0.012802129960691319, 0.8367222328254639]",,646357,119,False,,,,"[9.267327807422955e-05, 8.03051714291749e-06, 4.466396604498123e-06, 0.9998948298081783]",mixture,"{'mixture': 0.8367218370151065, 'uniform': 0.22632976686687015, 'top1': 0.8367222328254639, 'top2': 0.8367218370151065, 'top3': 0.8367218370151065}",0.8365702659977134,0.07170842372260706,0.003773437115060449,0.7063903969579743,0.7317769245256655,ensemble,1.0, +3,201,86.9,181.5,"[0.2432412176643748, 1.0398523409675826e-06, 0.011922198046396285, 0.8167738623605225]",,961409,179,False,,,,"[0.00011431641553317873, 9.803144569781586e-06, 5.404554281086763e-06, 0.9998704758856161]",mixture,"{'mixture': 0.8167700660471827, 'uniform': 0.2017762310498385, 'top1': 0.8167738623605225, 'top2': 0.8167700660471827, 'top3': 0.8167700660471827}",0.8520933668114993,0.07309227793276327,0.006567630992236721,0.8085989086638157,0.7770533684262554,student,1.0, +4,268,85.8,363.0,"[0.2774407873320246, 0.00021757778155051012, 0.017681862132876883, 0.8908474538075571]","[0.2914352298181363, 0.00043444480174767376, 0.022986299099895986, 0.9522409389197227]",1406676,245,False,,,,"[7.6728663456923e-05, 4.684119979198459e-06, 2.729580942504572e-06, 0.9999158576356213]",mixture,"{'mixture': 0.8908468921531416, 'uniform': 0.26357952088220327, 'top1': 0.8908474538075571, 'top2': 0.8908466768504242, 'top3': 0.8908466745253185}",0.8240250559424885,0.08222283438241113,0.006853722569206671,0.8540059637638082,0.8586939563953454,student,1.0, +5,335,84.9,386.6,"[0.27581865697076663, 0.00021098453816418617, 0.01700586078043835, 0.8932543310747327]","[0.2830548899311145, 0.00042141133660207754, 0.0218039269162802, 0.9478203037556284]",1792685,307,False,,,,"[7.369894582484202e-05, 4.572438172123279e-06, 2.646723698279208e-06, 0.9999190818923047]",mixture,"{'mixture': 0.8932545092013533, 'uniform': 0.26145255935947603, 'top1': 0.8932543310747327, 'top2': 0.8932540626415855, 'top3': 0.8932542293319554}",0.8753142915497427,0.06372080616488156,0.0033989463018334635,0.854906715658334,0.8722805399151158,student,1.0, +6,402,84.4,409.0,"[0.2778182744095054, 0.00028432021766136274, 0.01471529608171681, 0.9022960378346366]","[0.2821117277570612, 0.0005681818181818182, 0.01807491548267771, 0.9568074274579239]",2180202,371,False,,,,"[6.86879160649449e-05, 4.180239560132354e-06, 2.3631737087619997e-06, 0.9999247686706662]",mixture,"{'mixture': 0.9022957193488081, 'uniform': 0.26732216329589775, 'top1': 0.9022960378346366, 'top2': 0.9022954553953105, 'top3': 0.9022956474829681}",0.8780199357605633,0.06069790674314873,0.0038901085719583125,0.8923665379571899,0.8892548845825352,student,1.0, +7,402,0.0,255.7,,,,,,,,,,,,0.8830426902129573,0.058771422313978854,0.003914845060847729,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..d5dd9108e20ebf61b2cf81021a226f6eb06a5305 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/log.txt @@ -0,0 +1,98 @@ +[07:06:35][ 0s][START] target=lucchi-pp variant=macs-v11-norestart-acq-random budget=5.0% seed=1 | {"task": "mitochondria", "target": "lucchi-pp", "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": 8, "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} +[07:06:35][ 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. +[07:06:35][ 0s][DATA] lucchi-pp: train=8050 val=1750 test=1750 patches +[07:06:36][ 2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, dros-vnc=2.77, multiclass=2.60 +[07:06:36][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[07:06:38][ 3s][WEIGHTS] mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 | eff.sources=1.44 | MMD2 mix=0.0050 uniform=0.2092 +[07:06:38][ 3s][WEIGHTS] per-source: fly-mito=0.013, kasthuri=0.022, dros-vnc=0.005, multiclass=0.960 | min-MMD source (NeuroADDA rule) = multiclass +[07:07:15][ 41s][TEACHERS] target-val Dice: fly-mito=0.243, kasthuri=0.161, dros-vnc=0.162, multiclass=0.833 +[07:07:15][ 41s][TEACHERS] ensemble target-val Dice: mixture=0.834, mixture@raw=0.833, mixture@logit=0.835, per_source=0.834, per_source@raw=0.833, per_source@logit=0.834, uniform=0.331, uniform@raw=0.184, uniform@logit=0.220 +[07:07:15][ 41s][WEIGHTS] using mixture: fly-mito=0.033, kasthuri=0.032, dros-vnc=0.016, multiclass=0.920 +[07:07:15][ 41s][INIT] student initialised from teacher multiclass (weight 0.920) +[07:07:19][ 44s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[07:07:19][ 44s][KD] distilling the mixture ensemble (prob space) into the student +[07:09:21][ 167s][KD] soft labels for 6000 patches from 4 teachers in 123s +[07:12:19][ 345s][KD] epoch 1/5 | patches=6000 | loss=0.1185 | 178s +[07:15:17][ 523s][KD] epoch 2/5 | patches=6000 | loss=0.1064 | 178s +[07:18:15][ 701s][KD] epoch 3/5 | patches=6000 | loss=0.1043 | 178s +[07:21:13][ 879s][KD] epoch 4/5 | patches=6000 | loss=0.1035 | 178s +[07:24:11][ 1057s][KD] epoch 5/5 | patches=6000 | loss=0.1029 | 178s +[07:24:17][ 1063s][ROUND0] before any target label: val_dice=0.8825 val_vi=0.0631 +[07:24:17][ 1063s][AL] patches=8050 (pool 8050 after dropping blank; free blank negatives: True) | budget=402 (5.0%) | rounds=6 | per_round=67 | acquisition=random | first_round=stratified +[07:24:59][ 1104s][AL] round-1 uncertainty quality on 8000 patches: {'spearman_uncertainty_error': -0.0022443099490930996, 'auroc_top20_error': 0.49312246093750006} +[07:25:40][ 1145s][AL] round 1: stratified cold start: 34 patches from the membrane-rich half (ensemble prevalence >= 0.074) + 33 from the membrane-poor half of 2000 candidates +[07:25:45][ 1151s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.932, uniform=0.354, top1=0.932, top2=0.932, top3=0.932 -> using mixture +[07:25:46][ 1152s][WEIGHTS] round 1: student kept (labeled Dice 0.940 vs best teacher multiclass 0.932); distilling the corrected ensemble into it +[07:25:46][ 1152s][WEIGHTS] round 1: teacher Dice on 67 labeled patches + 0 blank (352197 fg px): fly-mito=0.288, kasthuri=0.000, dros-vnc=0.023, multiclass=0.932 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:27:49][ 1274s][KD] soft labels for 6000 patches from 4 teachers in 123s +[07:30:47][ 1452s][KD] epoch 1/2 | patches=6000 | loss=0.0471 | 178s +[07:33:45][ 1630s][KD] epoch 2/2 | patches=6000 | loss=0.0459 | 178s +[07:34:26][ 1672s][CONS] pseudo-labels for 2000 unlabeled patches from the current ensemble in 41s +[07:34:31][ 1676s][FT] round 1/6 epoch 1/5 | patches=67 | loss=0.3095 +[07:34:36][ 1681s][FT] round 1/6 epoch 2/5 | patches=67 | loss=0.3610 +[07:34:40][ 1686s][FT] round 1/6 epoch 3/5 | patches=67 | loss=0.3744 +[07:34:45][ 1690s][FT] round 1/6 epoch 4/5 | patches=67 | loss=0.2857 +[07:34:50][ 1695s][FT] round 1/6 epoch 5/5 | patches=67 | loss=0.3504 +[07:34:56][ 1701s][AL] round 1/6 | labeled=67 | selection_seconds=82.7000 round_seconds=632.7000 restart_student_dice=0.9398 restart_teacher_dice=0.9318 val_dice=0.7964 val_vi=0.0972 val_ece=0.0086 +[07:35:06][ 1712s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.931, uniform=0.316, top1=0.931, top2=0.931, top3=0.931 -> using mixture +[07:35:06][ 1712s][WEIGHTS] round 2: teacher Dice on 134 labeled patches + 0 blank (686915 fg px): fly-mito=0.280, kasthuri=0.000, dros-vnc=0.031, multiclass=0.931 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:35:10][ 1716s][CONS] round 2: on the 67 new patches student Dice=0.922 vs ensemble 0.930 -> pseudo-labels from the ensemble | consistency x1.00 +[07:35:19][ 1724s][FT] round 2/6 epoch 1/5 | patches=134 | loss=0.3237 +[07:35:28][ 1733s][FT] round 2/6 epoch 2/5 | patches=134 | loss=0.2841 +[07:35:36][ 1741s][FT] round 2/6 epoch 3/5 | patches=134 | loss=0.2896 +[07:35:45][ 1750s][FT] round 2/6 epoch 4/5 | patches=134 | loss=0.2877 +[07:35:53][ 1759s][FT] round 2/6 epoch 5/5 | patches=134 | loss=0.2835 +[07:35:59][ 1764s][AL] round 2/6 | labeled=134 | selection_seconds=0.2000 round_seconds=57.6000 pseudo_student_dice=0.9218 pseudo_ensemble_dice=0.9298 consistency_scale=1.0000 val_dice=0.8360 val_vi=0.0763 val_ece=0.0050 +[07:36:15][ 1780s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.936, uniform=0.309, top1=0.936, top2=0.936, top3=0.936 -> using mixture +[07:36:15][ 1780s][WEIGHTS] round 3: teacher Dice on 201 labeled patches + 0 blank (1044085 fg px): fly-mito=0.285, kasthuri=0.000, dros-vnc=0.028, multiclass=0.936 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:36:19][ 1784s][CONS] round 3: on the 67 new patches student Dice=0.949 vs ensemble 0.946 -> pseudo-labels from the ensemble | consistency x1.00 +[07:36:31][ 1797s][FT] round 3/6 epoch 1/5 | patches=201 | loss=0.2540 +[07:36:44][ 1809s][FT] round 3/6 epoch 2/5 | patches=201 | loss=0.3283 +[07:36:56][ 1822s][FT] round 3/6 epoch 3/5 | patches=201 | loss=0.2696 +[07:37:09][ 1834s][FT] round 3/6 epoch 4/5 | patches=201 | loss=0.2373 +[07:37:22][ 1847s][FT] round 3/6 epoch 5/5 | patches=201 | loss=0.2316 +[07:37:27][ 1853s][AL] round 3/6 | labeled=201 | selection_seconds=0.2000 round_seconds=82.7000 pseudo_student_dice=0.9488 pseudo_ensemble_dice=0.9456 consistency_scale=1.0000 val_dice=0.8718 val_vi=0.0733 val_ece=0.0063 +[07:37:48][ 1873s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.934, uniform=0.297, top1=0.934, top2=0.934, top3=0.934 -> using mixture +[07:37:48][ 1873s][WEIGHTS] round 4: teacher Dice on 268 labeled patches + 0 blank (1398805 fg px): fly-mito=0.289, kasthuri=0.000, dros-vnc=0.027, multiclass=0.934 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:37:52][ 1877s][CONS] round 4: on the 67 new patches student Dice=0.938 vs ensemble 0.929 -> pseudo-labels from the ensemble | consistency x1.00 +[07:38:08][ 1894s][FT] round 4/6 epoch 1/5 | patches=268 | loss=0.2372 +[07:38:25][ 1910s][FT] round 4/6 epoch 2/5 | patches=268 | loss=0.2139 +[07:38:41][ 1927s][FT] round 4/6 epoch 3/5 | patches=268 | loss=0.3157 +[07:38:58][ 1943s][FT] round 4/6 epoch 4/5 | patches=268 | loss=0.2718 +[07:39:14][ 1959s][FT] round 4/6 epoch 5/5 | patches=268 | loss=0.2351 +[07:39:20][ 1965s][AL] round 4/6 | labeled=268 | selection_seconds=0.2000 round_seconds=106.7000 pseudo_student_dice=0.9379 pseudo_ensemble_dice=0.9288 consistency_scale=1.0000 val_dice=0.8609 val_vi=0.0737 val_ece=0.0064 +[07:39:45][ 1991s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.935, uniform=0.314, top1=0.935, top2=0.935, top3=0.935 -> using mixture +[07:39:45][ 1991s][WEIGHTS] round 5: teacher Dice on 335 labeled patches + 0 blank (1751052 fg px): fly-mito=0.285, kasthuri=0.000, dros-vnc=0.024, multiclass=0.935 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:39:49][ 1995s][CONS] round 5: on the 67 new patches student Dice=0.944 vs ensemble 0.937 -> pseudo-labels from the ensemble | consistency x1.00 +[07:40:09][ 2015s][FT] round 5/6 epoch 1/5 | patches=335 | loss=0.2320 +[07:40:30][ 2035s][FT] round 5/6 epoch 2/5 | patches=335 | loss=0.2251 +[07:40:50][ 2055s][FT] round 5/6 epoch 3/5 | patches=335 | loss=0.2079 +[07:41:10][ 2076s][FT] round 5/6 epoch 4/5 | patches=335 | loss=0.2125 +[07:41:30][ 2096s][FT] round 5/6 epoch 5/5 | patches=335 | loss=0.2060 +[07:41:36][ 2102s][AL] round 5/6 | labeled=335 | selection_seconds=0.2000 round_seconds=130.5000 pseudo_student_dice=0.9436 pseudo_ensemble_dice=0.9366 consistency_scale=1.0000 val_dice=0.8758 val_vi=0.0692 val_ece=0.0052 +[07:42:06][ 2132s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.934, uniform=0.324, top1=0.934, top2=0.934, top3=0.934 -> using mixture +[07:42:06][ 2132s][WEIGHTS] round 6: teacher Dice on 402 labeled patches + 0 blank (2057337 fg px): fly-mito=0.282, kasthuri=0.000, dros-vnc=0.023, multiclass=0.934 | weights: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:42:11][ 2136s][CONS] round 6: on the 67 new patches student Dice=0.934 vs ensemble 0.929 -> pseudo-labels from the ensemble | consistency x1.00 +[07:42:35][ 2160s][FT] round 6/6 epoch 1/5 | patches=402 | loss=0.2464 +[07:42:59][ 2184s][FT] round 6/6 epoch 2/5 | patches=402 | loss=0.2427 +[07:43:23][ 2209s][FT] round 6/6 epoch 3/5 | patches=402 | loss=0.2298 +[07:43:48][ 2233s][FT] round 6/6 epoch 4/5 | patches=402 | loss=0.2270 +[07:44:12][ 2257s][FT] round 6/6 epoch 5/5 | patches=402 | loss=0.2242 +[07:44:18][ 2263s][AL] round 6/6 | labeled=402 | selection_seconds=0.2000 round_seconds=155.8000 pseudo_student_dice=0.9343 pseudo_ensemble_dice=0.9295 consistency_scale=1.0000 val_dice=0.8795 val_vi=0.0632 val_ece=0.0040 +[07:44:42][ 2288s][FT] final 1/10 epoch 1/1 | patches=402 | loss=0.2052 +[07:45:07][ 2312s][FT] final 2/10 epoch 1/1 | patches=402 | loss=0.2055 +[07:45:31][ 2336s][FT] final 3/10 epoch 1/1 | patches=402 | loss=0.2056 +[07:45:55][ 2361s][FT] final 4/10 epoch 1/1 | patches=402 | loss=0.2401 +[07:46:20][ 2385s][FT] final 5/10 epoch 1/1 | patches=402 | loss=0.2020 +[07:46:44][ 2410s][FT] final 6/10 epoch 1/1 | patches=402 | loss=0.1953 +[07:47:09][ 2434s][FT] final 7/10 epoch 1/1 | patches=402 | loss=0.1838 +[07:47:33][ 2458s][FT] final 8/10 epoch 1/1 | patches=402 | loss=0.1807 +[07:47:57][ 2483s][FT] final 9/10 epoch 1/1 | patches=402 | loss=0.1836 +[07:48:22][ 2507s][FT] final 10/10 epoch 1/1 | patches=402 | loss=0.1761 +[07:48:28][ 2513s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=243.8000 val_dice=0.8844 val_vi=0.0651 val_ece=0.0045 +[07:48:28][ 2513s][WEIGHTS] final: fly-mito=0.000, kasthuri=0.000, dros-vnc=0.000, multiclass=1.000 +[07:49:50][ 2595s][TEST] raw student (no EMA, no TTA): patch_dice=0.9146 patch_vi=0.0435 +[07:49:59][ 2605s][TEST] final model without TTA: patch_dice=0.9146 patch_vi=0.0435 +[07:49:59][ 2605s][RESOURCES] {"phases": {"setup": {"seconds": 0.75, "peak_cpu_rss_mb": 2174.8, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 42.33, "peak_cpu_rss_mb": 4176.2, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 1012.81, "peak_cpu_rss_mb": 5951.0, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4128.0, "kd_epochs": 5}, "active_learning": {"seconds": 1450.45, "peak_cpu_rss_mb": 5951.0, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4176.0, "rounds": 6}, "inference": {"seconds": 91.64, "peak_cpu_rss_mb": 5951.0, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4176.0, "patches": 1750, "images": 50, "patches_per_second": 19.1, "seconds_per_image": 1.833}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 5951.0, "total_seconds": 2598.0} +[07:49:59][ 2605s][TEST] patch_dice=0.9217 patch_iou=0.8940 patch_precision=0.9546 patch_recall=0.9358 patch_vi=0.0413 image_dice=0.9519 image_iou=0.9087 image_precision=0.9614 image_recall=0.9437 image_vi=0.0484 cal_ece=0.0017 cal_nll=0.0169 cal_brier=0.0036 cal_mean_confidence=0.9970 cal_accuracy=0.9953 cal_predicted_prevalence=0.0446 cal_true_prevalence=0.0452 cal_mean_probability=0.0448 +[07:50:00][ 2605s][DONE] lucchi-pp macs-v11-norestart-acq-random b5 s1 | test patch Dice=0.9217 VI=0.0413 | image Dice=0.9519 VI=0.0484 | 2605s diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/result.json b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..1718d0a39a82b9acdf8134ea8af59e5599998ec2 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/result.json @@ -0,0 +1,647 @@ +{ + "target": "lucchi-pp", + "variant": "macs-v11-norestart-acq-random", + "budget": 5.0, + "seed": 1, + "config": { + "task": "mitochondria", + "target": "lucchi-pp", + "variant": "macs-v11-norestart-acq-random", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "lucchi-pp", + "--variants", + "macs-v11-norestart-acq-random", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "drosophila-vnc", + "multiclass" + ], + "alphas": [ + 0.03256907457032885, + 0.031801008835420135, + 0.015561893577512515, + 0.9200680230167385 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "drosophila-vnc", + "multiclass" + ], + "mixture": [ + 0.03256907457032885, + 0.031801008835420135, + 0.015561893577512515, + 0.9200680230167385 + ], + "per_source": [ + 0.013021514318439549, + 0.021594227384204207, + 0.00495364759466876, + 0.9604306107026874 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.001413505929905978, + "mmd2_mixture": 0.005039697348754646, + "mmd2_uniform": 0.2092228336519747, + "mmd2_single_source": [ + 0.5680513520444281, + 0.4644148125694437, + 0.6002511462494531, + 0.004698715465898062 + ], + "median_squared_distance": 0.1892747764462741, + "gammas": [ + 1.3208310409546982, + 2.6416620819093963, + 5.283324163818793, + 10.566648327637585, + 21.13329665527517 + ], + "max_weight": 0.9200680230167385, + "entropy": 0.36261914188486677, + "effective_sources": 1.4370884282110035, + "min_weight": 0.015561893577512515 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6675354526109456, + 0.5899267604637966, + 0.8158227740743191, + 0.007664790896468099 + ], + "standardized": [ + 0.4800153933276398, + 0.22710384776568787, + 0.9632547653742088, + -1.670374006467536 + ], + "tau": 0.5, + "max_weight": 0.9604306107026874, + "entropy": 0.2044174691373809, + "effective_sources": 1.2268102019632934, + "min_weight": 0.00495364759466876 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "multiclass", + "seconds": 1.3 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.24317352702666994, + 0.16070093323689605, + 0.16243146338599246, + 0.8333945099982429 + ], + "ensemble_val_dice": { + "mixture": 0.8341164139370107, + "mixture@raw": 0.8334634452984915, + "mixture@logit": 0.8348661307075531, + "per_source": 0.8335017830355866, + "per_source@raw": 0.8334355010464378, + "per_source@logit": 0.8337681001665089, + "uniform": 0.3312974448451058, + "uniform@raw": 0.18374038775198187, + "uniform@logit": 0.21963766526924644 + }, + "val_patches": 1024 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 2.767641305923462, + 2.597146987915039 + ], + "top_source": "multiclass", + "weight_stats": { + "max_weight": 0.9200680230167385, + "entropy": 0.36261914188486677, + "effective_sources": 1.4370884282110035, + "min_weight": 0.015561893577512515 + }, + "correlations": { + "spearman_weight_teacher_dice": 0.7999999999999999, + "spearman_weight_taxonomy": -0.2581988897471611, + "top_source": "multiclass", + "top_source_is_best_teacher": true + }, + "init_teacher": "multiclass", + "student_init": "top-source", + "blank_fraction": 0.0, + "kd_seconds": 1012.8, + "round0_val": { + "val_dice": 0.8824999818736274, + "val_vi": 0.06309425262480922, + "val_ece": 0.01823718431407162 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 67, + "selection_seconds": 82.7, + "round_seconds": 632.7, + "teacher_dice_labeled": [ + 0.2876389731008006, + 2.8393119779215102e-06, + 0.02286523788150808, + 0.931847537434372 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 352197, + "evidence_nonempty_patches": 62, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.939764804154994, + "restart_teacher_dice": 0.931847537434372, + "weights": [ + 5.6389282479936555e-05, + 3.1020171205931166e-06, + 1.907901148148134e-06, + 0.9999386007992513 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.931847537434372, + "uniform": 0.3539312329080638, + "top1": 0.931847537434372, + "top2": 0.931847537434372, + "top3": 0.931847537434372 + }, + "val_dice": 0.7963996417830616, + "val_vi": 0.09719983582607203, + "val_ece": 0.008608905565021063 + }, + { + "round": 2, + "n_labeled": 134, + "selection_seconds": 0.2, + "round_seconds": 57.6, + "teacher_dice_labeled": [ + 0.2797322410747025, + 1.455782075246464e-06, + 0.03129283038951154, + 0.9308365557494995 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 686915, + "evidence_nonempty_patches": 121, + "first_correction": false, + "restart": "", + "weights": [ + 5.2632044600482846e-05, + 3.1335047376057656e-06, + 2.0967594661903174e-06, + 0.9999421376911957 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9308365557494995, + "uniform": 0.3157631136821172, + "top1": 0.9308365557494995, + "top2": 0.9308365557494995, + "top3": 0.9308365557494995 + }, + "pseudo_student_dice": 0.9217522272049676, + "pseudo_ensemble_dice": 0.9297682266068903, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8360258919249423, + "val_vi": 0.07626112770829221, + "val_ece": 0.005044878277267725 + }, + { + "round": 3, + "n_labeled": 201, + "selection_seconds": 0.2, + "round_seconds": 82.7, + "teacher_dice_labeled": [ + 0.285088893636434, + 9.573124792382856e-07, + 0.027690813420591064, + 0.9358991276509686 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1044085, + "evidence_nonempty_patches": 179, + "first_correction": false, + "restart": "", + "weights": [ + 5.278706224823984e-05, + 2.978803091277857e-06, + 1.922732452354128e-06, + 0.9999423114022081 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9358986801938234, + "uniform": 0.3092815616297828, + "top1": 0.9358991276509686, + "top2": 0.9358986801938234, + "top3": 0.9358986801938234 + }, + "pseudo_student_dice": 0.9488361135853891, + "pseudo_ensemble_dice": 0.9455641796027374, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8717971151028363, + "val_vi": 0.07333970560102024, + "val_ece": 0.006264443101983306 + }, + { + "round": 4, + "n_labeled": 268, + "selection_seconds": 0.2, + "round_seconds": 106.7, + "teacher_dice_labeled": [ + 0.2887129617740495, + 7.146374179864233e-07, + 0.02675218913568767, + 0.9341140653396488 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1398805, + "evidence_nonempty_patches": 243, + "first_correction": false, + "restart": "", + "weights": [ + 5.572085136061948e-05, + 3.0324375408485426e-06, + 1.9390704386544167e-06, + 0.99993930764066 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9341141124811376, + "uniform": 0.2974809266161287, + "top1": 0.9341140653396488, + "top2": 0.9341141124811376, + "top3": 0.9341141124811376 + }, + "pseudo_student_dice": 0.9379381073761351, + "pseudo_ensemble_dice": 0.9288095752421666, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8609243697760304, + "val_vi": 0.0737337951427876, + "val_ece": 0.006416487514190574 + }, + { + "round": 5, + "n_labeled": 335, + "selection_seconds": 0.2, + "round_seconds": 130.5, + "teacher_dice_labeled": [ + 0.2851837942410718, + 5.709202892510553e-07, + 0.02400631314910587, + 0.9346082435626396 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 1751052, + "evidence_nonempty_patches": 304, + "first_correction": false, + "restart": "", + "weights": [ + 5.352363032039484e-05, + 3.017491387355586e-06, + 1.8772546836238203e-06, + 0.9999415816236087 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.934608280971694, + "uniform": 0.3136637816728857, + "top1": 0.9346082435626396, + "top2": 0.934608280971694, + "top3": 0.934608280971694 + }, + "pseudo_student_dice": 0.9435884035290092, + "pseudo_ensemble_dice": 0.9365794422929289, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8758075852599518, + "val_vi": 0.06919363862118208, + "val_ece": 0.005191076904368195 + }, + { + "round": 6, + "n_labeled": 402, + "selection_seconds": 0.2, + "round_seconds": 155.8, + "teacher_dice_labeled": [ + 0.28191093592370076, + 4.857738693491748e-07, + 0.022959482419344, + 0.9338568269067726 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 2057337, + "evidence_nonempty_patches": 357, + "first_correction": false, + "restart": "", + "weights": [ + 5.219100403170781e-05, + 3.040252132571167e-06, + 1.8717196545786043e-06, + 0.9999428970241812 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9338566313420587, + "uniform": 0.32391720725548095, + "top1": 0.9338568269067726, + "top2": 0.9338566313420587, + "top3": 0.9338566313420587 + }, + "pseudo_student_dice": 0.9342727392589453, + "pseudo_ensemble_dice": 0.9294967729090275, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.8795111878237105, + "val_vi": 0.063227044280279, + "val_ece": 0.00398630717600138 + }, + { + "round": 7, + "n_labeled": 402, + "selection_seconds": 0.0, + "round_seconds": 243.8, + "final_consolidation": true, + "val_dice": 0.884373415141364, + "val_vi": 0.06514450746984593, + "val_ece": 0.004541055743000292 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.0022443099490930996, + "auroc_top20_error": 0.49312246093750006 + }, + "al_seconds": 1450.4, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 5.219100403170781e-05, + 3.040252132571167e-06, + 1.8717196545786043e-06, + 0.9999428970241812 + ], + "weight_stats_final": { + "max_weight": 0.9999428970241812, + "entropy": 0.0006350434984520711, + "effective_sources": 1.0006352451812648, + "min_weight": 1.8717196545786043e-06 + }, + "top_source_final": "multiclass", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.9146401557969808, + "patch_iou": 0.8843900144666924, + "patch_precision": 0.9444387414189249, + "patch_recall": 0.9370968354478443, + "patch_vi": 0.0434695779629698, + "n_patches": 1750, + "cal_ece": 0.002688002960785389, + "cal_nll": 0.021318272008663725, + "cal_brier": 0.004021202493275474, + "cal_mean_confidence": 0.997700837782214, + "cal_accuracy": 0.9950128348214285, + "cal_predicted_prevalence": 0.04469475446428572, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.04464832570282386 + }, + "test_notta": { + "patch_dice": 0.9146401557969808, + "patch_iou": 0.8843900144666924, + "patch_precision": 0.9444387414189249, + "patch_recall": 0.9370968354478443, + "patch_vi": 0.0434695779629698, + "n_patches": 1750, + "cal_ece": 0.002688002960785389, + "cal_nll": 0.021318272008663725, + "cal_brier": 0.004021202493275474, + "cal_mean_confidence": 0.997700837782214, + "cal_accuracy": 0.9950128348214285, + "cal_predicted_prevalence": 0.04469475446428572, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.04464832570282386 + }, + "test": { + "patch_dice": 0.9216756956926461, + "patch_iou": 0.8940452932960153, + "patch_precision": 0.9545525749986802, + "patch_recall": 0.9358326668206619, + "patch_vi": 0.04133018567270993, + "n_patches": 1750, + "image_dice": 0.9518754282319007, + "image_iou": 0.908654900675551, + "image_precision": 0.9613938476386131, + "image_recall": 0.9436737891974423, + "image_vi": 0.04840384507907384, + "n_images": 50, + "cal_ece": 0.0017047703042317487, + "cal_nll": 0.016949773001781545, + "cal_brier": 0.0036152859965237925, + "cal_mean_confidence": 0.9969826720899462, + "cal_accuracy": 0.9952779017857143, + "cal_predicted_prevalence": 0.044606026785714285, + "cal_true_prevalence": 0.04515513392857143, + "cal_mean_probability": 0.04480511960019529 + }, + "eval_seconds": 91.6, + "total_seconds": 2604.6, + "resources": { + "phases": { + "setup": { + "seconds": 0.75, + "peak_cpu_rss_mb": 2174.8, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 42.33, + "peak_cpu_rss_mb": 4176.2, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 1012.81, + "peak_cpu_rss_mb": 5951.0, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4128.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 1450.45, + "peak_cpu_rss_mb": 5951.0, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4176.0, + "rounds": 6 + }, + "inference": { + "seconds": 91.64, + "peak_cpu_rss_mb": 5951.0, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4176.0, + "patches": 1750, + "images": 50, + "patches_per_second": 19.1, + "seconds_per_image": 1.833 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 5951.0, + "total_seconds": 2598.0 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/lucchi-pp/macs-v11-norestart-acq-random_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/mitochondria/lucchi-pp/macs-v11-norestart-acq-random/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..f2791e11f2f8290e88aafc9a9eb3c7b760112363 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/lucchi-pp/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 +1,67,82.7,632.7,"[0.2876389731008006, 2.8393119779215102e-06, 0.02286523788150808, 0.931847537434372]",,352197,62,True,kept,0.939764804154994,0.931847537434372,"[5.6389282479936555e-05, 3.1020171205931166e-06, 1.907901148148134e-06, 0.9999386007992513]",mixture,"{'mixture': 0.931847537434372, 'uniform': 0.3539312329080638, 'top1': 0.931847537434372, 'top2': 0.931847537434372, 'top3': 0.931847537434372}",0.7963996417830616,0.09719983582607203,0.008608905565021063,,,,, +2,134,0.2,57.6,"[0.2797322410747025, 1.455782075246464e-06, 0.03129283038951154, 0.9308365557494995]",,686915,121,False,,,,"[5.2632044600482846e-05, 3.1335047376057656e-06, 2.0967594661903174e-06, 0.9999421376911957]",mixture,"{'mixture': 0.9308365557494995, 'uniform': 0.3157631136821172, 'top1': 0.9308365557494995, 'top2': 0.9308365557494995, 'top3': 0.9308365557494995}",0.8360258919249423,0.07626112770829221,0.005044878277267725,0.9217522272049676,0.9297682266068903,ensemble,1.0, +3,201,0.2,82.7,"[0.285088893636434, 9.573124792382856e-07, 0.027690813420591064, 0.9358991276509686]",,1044085,179,False,,,,"[5.278706224823984e-05, 2.978803091277857e-06, 1.922732452354128e-06, 0.9999423114022081]",mixture,"{'mixture': 0.9358986801938234, 'uniform': 0.3092815616297828, 'top1': 0.9358991276509686, 'top2': 0.9358986801938234, 'top3': 0.9358986801938234}",0.8717971151028363,0.07333970560102024,0.006264443101983306,0.9488361135853891,0.9455641796027374,ensemble,1.0, +4,268,0.2,106.7,"[0.2887129617740495, 7.146374179864233e-07, 0.02675218913568767, 0.9341140653396488]",,1398805,243,False,,,,"[5.572085136061948e-05, 3.0324375408485426e-06, 1.9390704386544167e-06, 0.99993930764066]",mixture,"{'mixture': 0.9341141124811376, 'uniform': 0.2974809266161287, 'top1': 0.9341140653396488, 'top2': 0.9341141124811376, 'top3': 0.9341141124811376}",0.8609243697760304,0.0737337951427876,0.006416487514190574,0.9379381073761351,0.9288095752421666,ensemble,1.0, +5,335,0.2,130.5,"[0.2851837942410718, 5.709202892510553e-07, 0.02400631314910587, 0.9346082435626396]",,1751052,304,False,,,,"[5.352363032039484e-05, 3.017491387355586e-06, 1.8772546836238203e-06, 0.9999415816236087]",mixture,"{'mixture': 0.934608280971694, 'uniform': 0.3136637816728857, 'top1': 0.9346082435626396, 'top2': 0.934608280971694, 'top3': 0.934608280971694}",0.8758075852599518,0.06919363862118208,0.005191076904368195,0.9435884035290092,0.9365794422929289,ensemble,1.0, +6,402,0.2,155.8,"[0.28191093592370076, 4.857738693491748e-07, 0.022959482419344, 0.9338568269067726]",,2057337,357,False,,,,"[5.219100403170781e-05, 3.040252132571167e-06, 1.8717196545786043e-06, 0.9999428970241812]",mixture,"{'mixture': 0.9338566313420587, 'uniform': 0.32391720725548095, 'top1': 0.9338568269067726, 'top2': 0.9338566313420587, 'top3': 0.9338566313420587}",0.8795111878237105,0.063227044280279,0.00398630717600138,0.9342727392589453,0.9294967729090275,ensemble,1.0, +7,402,0.0,243.8,,,,,,,,,,,,0.884373415141364,0.06514450746984593,0.004541055743000292,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..9c9f551fb69aaf6948b1fb4558793e54d4f46f47 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/log.txt @@ -0,0 +1,103 @@ +[23:14:51][ 0s][START] target=multiclass variant=macs-v11-norestart-acq-laplace budget=5.0% seed=1 | {"task": "mitochondria", "target": "multiclass", "variant": "macs-v11-norestart-acq-laplace", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[23:14:51][ 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. +[23:14:51][ 0s][DATA] multiclass: train=1855 val=350 test=175 patches +[23:14:55][ 4s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, dros-vnc=2.77 +[23:14:55][ 4s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 1s +[23:14:57][ 6s][WEIGHTS] mixture: fly-mito=0.012, kasthuri=0.012, lucchi=0.968, dros-vnc=0.008 | eff.sources=1.19 | MMD2 mix=0.0059 uniform=0.2484 +[23:14:57][ 6s][WEIGHTS] per-source: fly-mito=0.014, kasthuri=0.019, lucchi=0.962, dros-vnc=0.005 | min-MMD source (NeuroADDA rule) = lucchi +[23:15:10][ 19s][TEACHERS] target-val Dice: fly-mito=0.258, kasthuri=0.175, lucchi=0.968, dros-vnc=0.176 +[23:15:10][ 19s][TEACHERS] ensemble target-val Dice: mixture=0.967, mixture@raw=0.968, mixture@logit=0.967, per_source=0.967, per_source@raw=0.967, per_source@logit=0.966, uniform=0.248, uniform@raw=0.208, uniform@logit=0.195 +[23:15:10][ 19s][WEIGHTS] using mixture: fly-mito=0.012, kasthuri=0.012, lucchi=0.968, dros-vnc=0.008 +[23:15:10][ 19s][INIT] student initialised from teacher lucchi (weight 0.968) +[23:15:13][ 22s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[23:15:13][ 22s][KD] distilling the mixture ensemble (prob space) into the student +[23:15:50][ 60s][KD] soft labels for 1855 patches from 4 teachers in 37s +[23:16:45][ 114s][KD] epoch 1/5 | patches=1855 | loss=0.1031 | 55s +[23:17:40][ 169s][KD] epoch 2/5 | patches=1855 | loss=0.0736 | 55s +[23:18:35][ 224s][KD] epoch 3/5 | patches=1855 | loss=0.0710 | 55s +[23:19:30][ 279s][KD] epoch 4/5 | patches=1855 | loss=0.0699 | 55s +[23:20:25][ 334s][KD] epoch 5/5 | patches=1855 | loss=0.0687 | 55s +[23:20:27][ 336s][ROUND0] before any target label: val_dice=0.9624 val_vi=0.0434 +[23:20:27][ 336s][AL] patches=1855 (pool 1855 after dropping blank; free blank negatives: True) | budget=92 (5.0%) | rounds=6 | per_round=15 | acquisition=laplace | first_round=stratified +[23:20:47][ 356s][AL] round-1 uncertainty quality on 1855 patches: {'spearman_uncertainty_error': -0.14785451464746693, 'auroc_top20_error': 0.2738500882731163} +[23:21:26][ 395s][AL] round 1: stratified cold start: 8 patches from the membrane-rich half (ensemble prevalence >= 0.071) + 7 from the membrane-poor half of 1855 candidates +[23:21:27][ 396s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.968, uniform=0.178, top1=0.968, top2=0.968, top3=0.968 -> using mixture +[23:21:28][ 397s][WEIGHTS] round 1: student kept (labeled Dice 0.966 vs best teacher lucchi 0.968); distilling the corrected ensemble into it +[23:21:28][ 397s][WEIGHTS] round 1: teacher Dice on 15 labeled patches + 0 blank (73632 fg px): fly-mito=0.294, kasthuri=0.000, lucchi=0.968, dros-vnc=0.018 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:22:05][ 434s][KD] soft labels for 1840 patches from 4 teachers in 38s +[23:23:00][ 489s][KD] epoch 1/2 | patches=1840 | loss=0.0453 | 55s +[23:23:55][ 544s][KD] epoch 2/2 | patches=1840 | loss=0.0448 | 55s +[23:24:33][ 582s][CONS] pseudo-labels for 1840 unlabeled patches from the current ensemble in 38s +[23:24:34][ 583s][FT] round 1/6 epoch 1/5 | patches=15 | loss=0.2333 +[23:24:36][ 585s][FT] round 1/6 epoch 2/5 | patches=15 | loss=0.1945 +[23:24:37][ 586s][FT] round 1/6 epoch 3/5 | patches=15 | loss=0.1615 +[23:24:38][ 588s][FT] round 1/6 epoch 4/5 | patches=15 | loss=0.1698 +[23:24:40][ 589s][FT] round 1/6 epoch 5/5 | patches=15 | loss=0.1832 +[23:24:42][ 591s][AL] round 1/6 | labeled=15 | selection_seconds=58.3000 round_seconds=252.7000 restart_student_dice=0.9665 restart_teacher_dice=0.9683 val_dice=0.9368 val_vi=0.0490 val_ece=0.0022 +[23:25:02][ 611s][AL] round 2: 272/1840 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:25:05][ 614s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.968, uniform=0.130, top1=0.968, top2=0.968, top3=0.968 -> using mixture +[23:25:05][ 614s][WEIGHTS] round 2: teacher Dice on 30 labeled patches + 0 blank (155821 fg px): fly-mito=0.300, kasthuri=0.000, lucchi=0.968, dros-vnc=0.011 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:25:07][ 616s][CONS] round 2: on the 15 new patches student Dice=0.959 vs ensemble 0.968 -> pseudo-labels from the ensemble | consistency x1.00 +[23:25:09][ 618s][FT] round 2/6 epoch 1/5 | patches=30 | loss=0.1966 +[23:25:11][ 621s][FT] round 2/6 epoch 2/5 | patches=30 | loss=0.1494 +[23:25:14][ 623s][FT] round 2/6 epoch 3/5 | patches=30 | loss=0.1582 +[23:25:16][ 625s][FT] round 2/6 epoch 4/5 | patches=30 | loss=0.1536 +[23:25:19][ 628s][FT] round 2/6 epoch 5/5 | patches=30 | loss=0.1538 +[23:25:21][ 630s][AL] round 2/6 | labeled=30 | selection_seconds=20.2000 round_seconds=36.4000 pseudo_student_dice=0.9591 pseudo_ensemble_dice=0.9683 consistency_scale=1.0000 val_dice=0.9536 val_vi=0.0476 val_ece=0.0014 +[23:25:40][ 650s][AL] round 3: 270/1825 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:25:45][ 654s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.966, uniform=0.120, top1=0.966, top2=0.966, top3=0.966 -> using mixture +[23:25:45][ 654s][WEIGHTS] round 3: teacher Dice on 45 labeled patches + 0 blank (240801 fg px): fly-mito=0.311, kasthuri=0.000, lucchi=0.966, dros-vnc=0.019 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:25:46][ 655s][CONS] round 3: on the 15 new patches student Dice=0.966 vs ensemble 0.962 -> pseudo-labels from the ensemble | consistency x1.00 +[23:25:49][ 658s][FT] round 3/6 epoch 1/5 | patches=45 | loss=0.1997 +[23:25:53][ 662s][FT] round 3/6 epoch 2/5 | patches=45 | loss=0.1722 +[23:25:56][ 665s][FT] round 3/6 epoch 3/5 | patches=45 | loss=0.1707 +[23:25:59][ 668s][FT] round 3/6 epoch 4/5 | patches=45 | loss=0.1597 +[23:26:02][ 671s][FT] round 3/6 epoch 5/5 | patches=45 | loss=0.1385 +[23:26:05][ 674s][AL] round 3/6 | labeled=45 | selection_seconds=20.0000 round_seconds=41.7000 pseudo_student_dice=0.9658 pseudo_ensemble_dice=0.9617 consistency_scale=1.0000 val_dice=0.9543 val_vi=0.0476 val_ece=0.0014 +[23:26:24][ 693s][AL] round 4: 282/1810 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:26:29][ 699s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.966, uniform=0.148, top1=0.966, top2=0.966, top3=0.966 -> using mixture +[23:26:29][ 699s][WEIGHTS] round 4: teacher Dice on 60 labeled patches + 0 blank (328081 fg px): fly-mito=0.318, kasthuri=0.000, lucchi=0.966, dros-vnc=0.033 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:26:31][ 700s][CONS] round 4: on the 15 new patches student Dice=0.971 vs ensemble 0.966 -> pseudo-labels from the ensemble | consistency x1.00 +[23:26:35][ 704s][FT] round 4/6 epoch 1/5 | patches=60 | loss=0.1675 +[23:26:39][ 708s][FT] round 4/6 epoch 2/5 | patches=60 | loss=0.1548 +[23:26:43][ 713s][FT] round 4/6 epoch 3/5 | patches=60 | loss=0.1342 +[23:26:48][ 717s][FT] round 4/6 epoch 4/5 | patches=60 | loss=0.1252 +[23:26:52][ 721s][FT] round 4/6 epoch 5/5 | patches=60 | loss=0.1480 +[23:26:54][ 723s][AL] round 4/6 | labeled=60 | selection_seconds=19.8000 round_seconds=47.3000 pseudo_student_dice=0.9707 pseudo_ensemble_dice=0.9663 consistency_scale=1.0000 val_dice=0.9217 val_vi=0.0488 val_ece=0.0021 +[23:27:13][ 743s][AL] round 5: 255/1795 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:27:20][ 749s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.964, uniform=0.157, top1=0.964, top2=0.964, top3=0.964 -> using mixture +[23:27:20][ 749s][WEIGHTS] round 5: teacher Dice on 75 labeled patches + 0 blank (412459 fg px): fly-mito=0.323, kasthuri=0.000, lucchi=0.964, dros-vnc=0.033 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:27:21][ 750s][CONS] round 5: on the 15 new patches student Dice=0.965 vs ensemble 0.955 -> pseudo-labels from the ensemble | consistency x1.00 +[23:27:26][ 755s][FT] round 5/6 epoch 1/5 | patches=75 | loss=0.1604 +[23:27:32][ 761s][FT] round 5/6 epoch 2/5 | patches=75 | loss=0.1526 +[23:27:37][ 766s][FT] round 5/6 epoch 3/5 | patches=75 | loss=0.1478 +[23:27:42][ 771s][FT] round 5/6 epoch 4/5 | patches=75 | loss=0.1327 +[23:27:47][ 776s][FT] round 5/6 epoch 5/5 | patches=75 | loss=0.1382 +[23:27:49][ 778s][AL] round 5/6 | labeled=75 | selection_seconds=19.7000 round_seconds=53.0000 pseudo_student_dice=0.9648 pseudo_ensemble_dice=0.9555 consistency_scale=1.0000 val_dice=0.9241 val_vi=0.0486 val_ece=0.0021 +[23:28:09][ 798s][AL] round 6: 270/1780 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[23:28:16][ 805s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.963, uniform=0.160, top1=0.963, top2=0.963, top3=0.963 -> using mixture +[23:28:16][ 805s][WEIGHTS] round 6: teacher Dice on 90 labeled patches + 0 blank (495033 fg px): fly-mito=0.313, kasthuri=0.000, lucchi=0.963, dros-vnc=0.033 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:28:18][ 807s][CONS] round 6: on the 15 new patches student Dice=0.965 vs ensemble 0.961 -> pseudo-labels from the ensemble | consistency x1.00 +[23:28:24][ 813s][FT] round 6/6 epoch 1/5 | patches=90 | loss=0.1348 +[23:28:30][ 819s][FT] round 6/6 epoch 2/5 | patches=90 | loss=0.1244 +[23:28:36][ 825s][FT] round 6/6 epoch 3/5 | patches=90 | loss=0.1252 +[23:28:42][ 831s][FT] round 6/6 epoch 4/5 | patches=90 | loss=0.1216 +[23:28:48][ 837s][FT] round 6/6 epoch 5/5 | patches=90 | loss=0.1199 +[23:28:50][ 839s][AL] round 6/6 | labeled=90 | selection_seconds=19.7000 round_seconds=58.5000 pseudo_student_dice=0.9648 pseudo_ensemble_dice=0.9607 consistency_scale=1.0000 val_dice=0.9269 val_vi=0.0477 val_ece=0.0019 +[23:28:56][ 845s][FT] final 1/10 epoch 1/1 | patches=90 | loss=0.1197 +[23:29:02][ 851s][FT] final 2/10 epoch 1/1 | patches=90 | loss=0.1198 +[23:29:08][ 857s][FT] final 3/10 epoch 1/1 | patches=90 | loss=0.1106 +[23:29:14][ 863s][FT] final 4/10 epoch 1/1 | patches=90 | loss=0.1230 +[23:29:20][ 869s][FT] final 5/10 epoch 1/1 | patches=90 | loss=0.1152 +[23:29:26][ 875s][FT] final 6/10 epoch 1/1 | patches=90 | loss=0.1010 +[23:29:32][ 881s][FT] final 7/10 epoch 1/1 | patches=90 | loss=0.1047 +[23:29:38][ 887s][FT] final 8/10 epoch 1/1 | patches=90 | loss=0.1098 +[23:29:44][ 893s][FT] final 9/10 epoch 1/1 | patches=90 | loss=0.1180 +[23:29:50][ 899s][FT] final 10/10 epoch 1/1 | patches=90 | loss=0.0981 +[23:29:52][ 901s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=60.1000 val_dice=0.9394 val_vi=0.0461 val_ece=0.0019 +[23:29:52][ 901s][WEIGHTS] final: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[23:30:01][ 910s][TEST] raw student (no EMA, no TTA): patch_dice=0.9013 patch_vi=0.0835 +[23:30:02][ 911s][TEST] final model without TTA: patch_dice=0.9013 patch_vi=0.0835 +[23:30:02][ 911s][RESOURCES] {"phases": {"setup": {"seconds": 2.66, "peak_cpu_rss_mb": 2160.1, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 18.41, "peak_cpu_rss_mb": 4176.4, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 312.05, "peak_cpu_rss_mb": 4656.7, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4148.0, "kd_epochs": 5}, "active_learning": {"seconds": 564.86, "peak_cpu_rss_mb": 4793.3, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4156.0, "rounds": 6}, "inference": {"seconds": 10.07, "peak_cpu_rss_mb": 4793.3, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4156.0, "patches": 175, "images": 5, "patches_per_second": 17.4, "seconds_per_image": 2.014}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 4793.3, "total_seconds": 908.1} +[23:30:02][ 911s][TEST] patch_dice=0.8905 patch_iou=0.8441 patch_precision=0.9614 patch_recall=0.8780 patch_vi=0.0924 image_dice=0.9038 image_iou=0.8275 image_precision=0.9575 image_recall=0.8651 image_vi=0.1166 cal_ece=0.0076 cal_nll=0.0696 cal_brier=0.0107 cal_mean_confidence=0.9948 cal_accuracy=0.9872 cal_predicted_prevalence=0.0829 cal_true_prevalence=0.0886 cal_mean_probability=0.0827 +[23:30:03][ 912s][DONE] multiclass macs-v11-norestart-acq-laplace b5 s1 | test patch Dice=0.8905 VI=0.0924 | image Dice=0.9038 VI=0.1166 | 911s diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..4221ab30381e5d89729ae13b8d839627e8035d4e --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/result.json @@ -0,0 +1,647 @@ +{ + "target": "multiclass", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "config": { + "task": "mitochondria", + "target": "multiclass", + "variant": "macs-v11-norestart-acq-laplace", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "fce9b15", + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "multiclass", + "--variants", + "macs-v11-norestart-acq-laplace", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "fce9b15", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "drosophila-vnc" + ], + "alphas": [ + 0.01152117817089438, + 0.011843492157429523, + 0.9683960236342841, + 0.008239306037391987 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "drosophila-vnc" + ], + "mixture": [ + 0.01152117817089438, + 0.011843492157429523, + 0.9683960236342841, + 0.008239306037391987 + ], + "per_source": [ + 0.014321460439305007, + 0.018762755437246332, + 0.9619826418368663, + 0.0049331422865822525 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.004128238346556883, + "mmd2_mixture": 0.005874251861260116, + "mmd2_uniform": 0.2484069500788153, + "mmd2_single_source": [ + 0.6121996102822926, + 0.5667987013145535, + 0.00484548842698207, + 0.6689345615903355 + ], + "median_squared_distance": 0.20474058073505946, + "gammas": [ + 1.2210573942031921, + 2.4421147884063843, + 4.884229576812769, + 9.768459153625537, + 19.536918307251074 + ], + "max_weight": 0.9683960236342841, + "entropy": 0.17460135147032324, + "effective_sources": 1.1907714227038573, + "min_weight": 0.008239306037391987 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6529838206918517, + 0.6112545706593259, + 0.0030196534926445473, + 0.8176336137299548 + ], + "standardized": [ + 0.426446119561544, + 0.2913887851649038, + -1.6771725126169683, + 0.9593376078905211 + ], + "tau": 0.5, + "max_weight": 0.9619826418368663, + "entropy": 0.19889648186407677, + "effective_sources": 1.2200556614747875, + "min_weight": 0.0049331422865822525 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "lucchi-pp", + "seconds": 1.9 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.25842427994689, + 0.17499716869513837, + 0.9675528820923397, + 0.17625340082547544 + ], + "ensemble_val_dice": { + "mixture": 0.9673725087302072, + "mixture@raw": 0.9675092802728925, + "mixture@logit": 0.9667406060014452, + "per_source": 0.9673224803379604, + "per_source@raw": 0.9674927929469518, + "per_source@logit": 0.9664724666731698, + "uniform": 0.2475083000162392, + "uniform@raw": 0.20821212102935532, + "uniform@logit": 0.19508060271550287 + }, + "val_patches": 350 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 3.80090069770813, + 2.767641305923462 + ], + "top_source": "lucchi-pp", + "weight_stats": { + "max_weight": 0.9683960236342841, + "entropy": 0.17460135147032324, + "effective_sources": 1.1907714227038573, + "min_weight": 0.008239306037391987 + }, + "correlations": { + "spearman_weight_teacher_dice": 0.39999999999999997, + "spearman_weight_taxonomy": -0.8944271909999159, + "top_source": "lucchi-pp", + "top_source_is_best_teacher": true + }, + "init_teacher": "lucchi-pp", + "student_init": "top-source", + "blank_fraction": 0.0, + "kd_seconds": 312.0, + "round0_val": { + "val_dice": 0.9623507293600193, + "val_vi": 0.043359066853027754, + "val_ece": 0.009826597372024078 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 15, + "selection_seconds": 58.3, + "round_seconds": 252.7, + "teacher_dice_labeled": [ + 0.2938186560944717, + 1.358086727418413e-05, + 0.9683060184264144, + 0.01782471245039349 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 73632, + "evidence_nonempty_patches": 14, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.9664782881707965, + "restart_teacher_dice": 0.9683060184264144, + "weights": [ + 1.4001540225714672e-05, + 7.623939953124306e-07, + 0.9999846022791299, + 6.337866489948326e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9683060184264144, + "uniform": 0.1778269851676831, + "top1": 0.9683060184264144, + "top2": 0.9683060184264144, + "top3": 0.9683060184264144 + }, + "val_dice": 0.9367825440230815, + "val_vi": 0.04902821843243505, + "val_ece": 0.002242055302164851 + }, + { + "round": 2, + "n_labeled": 30, + "selection_seconds": 20.2, + "round_seconds": 36.4, + "teacher_dice_labeled": [ + 0.30025932706211406, + 6.417579032485785e-06, + 0.9682767700219854, + 0.011387269224389809 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 155821, + "evidence_nonempty_patches": 29, + "first_correction": false, + "restart": "", + "weights": [ + 1.493736268096847e-05, + 7.625617057448709e-07, + 0.9999837056298168, + 5.944457965627549e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9682767700219854, + "uniform": 0.12951527045112196, + "top1": 0.9682767700219854, + "top2": 0.9682767700219854, + "top3": 0.9682767700219854 + }, + "pseudo_student_dice": 0.9590947745807612, + "pseudo_ensemble_dice": 0.968250747149762, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9536202317207382, + "val_vi": 0.04758055893308673, + "val_ece": 0.0013883460739741694 + }, + { + "round": 3, + "n_labeled": 45, + "selection_seconds": 20.0, + "round_seconds": 41.7, + "teacher_dice_labeled": [ + 0.31087552052835477, + 4.15278942865923e-06, + 0.965947077844261, + 0.018605052939869852 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 240801, + "evidence_nonempty_patches": 43, + "first_correction": false, + "restart": "", + "weights": [ + 1.7001853221602358e-05, + 7.805162522607558e-07, + 0.9999815636339252, + 6.539966008991305e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.965947077844261, + "uniform": 0.11951632802861115, + "top1": 0.965947077844261, + "top2": 0.965947077844261, + "top3": 0.965947077844261 + }, + "pseudo_student_dice": 0.9657906130425116, + "pseudo_ensemble_dice": 0.9616518292208995, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.954270906544748, + "val_vi": 0.047638874112841144, + "val_ece": 0.0013604008995015291 + }, + { + "round": 4, + "n_labeled": 60, + "selection_seconds": 19.8, + "round_seconds": 47.3, + "teacher_dice_labeled": [ + 0.317738364665974, + 3.048018483184082e-06, + 0.9660351508741329, + 0.033080268430627946 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 328081, + "evidence_nonempty_patches": 58, + "first_correction": false, + "restart": "", + "weights": [ + 1.819357973396529e-05, + 7.798195075771067e-07, + 0.9999802714085584, + 7.551922000094123e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9660351508741329, + "uniform": 0.14769137834183466, + "top1": 0.9660351508741329, + "top2": 0.9660351508741329, + "top3": 0.9660351508741329 + }, + "pseudo_student_dice": 0.9706596671377888, + "pseudo_ensemble_dice": 0.9662802457660312, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9216831759004065, + "val_vi": 0.04882421978576393, + "val_ece": 0.0020669694208124826 + }, + { + "round": 5, + "n_labeled": 75, + "selection_seconds": 19.7, + "round_seconds": 53.0, + "teacher_dice_labeled": [ + 0.32291000827099386, + 2.4244775250933425e-06, + 0.9639021682139222, + 0.03308238053915913 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 412459, + "evidence_nonempty_patches": 73, + "first_correction": false, + "restart": "", + "weights": [ + 1.957226710843179e-05, + 7.96625492397049e-07, + 0.9999788596188555, + 7.714885436697631e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9639021682139222, + "uniform": 0.156949700850309, + "top1": 0.9639021682139222, + "top2": 0.9639021682139222, + "top3": 0.9639021682139222 + }, + "pseudo_student_dice": 0.964805861328671, + "pseudo_ensemble_dice": 0.955485177998267, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9241157053196762, + "val_vi": 0.04862066420187986, + "val_ece": 0.0020597238850812863 + }, + { + "round": 6, + "n_labeled": 90, + "selection_seconds": 19.7, + "round_seconds": 58.5, + "teacher_dice_labeled": [ + 0.31313687709842153, + 2.0200632683815656e-06, + 0.9633628632939749, + 0.03338318269293788 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 495033, + "evidence_nonempty_patches": 88, + "first_correction": false, + "restart": "", + "weights": [ + 1.7845958603297553e-05, + 8.009314735799591e-07, + 0.9999805751114135, + 7.779985096301431e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9633628632939749, + "uniform": 0.15983339962159931, + "top1": 0.9633628632939749, + "top2": 0.9633628632939749, + "top3": 0.9633628632939749 + }, + "pseudo_student_dice": 0.9647581094503258, + "pseudo_ensemble_dice": 0.9606947095384917, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9269402183550002, + "val_vi": 0.04773342905549056, + "val_ece": 0.0019010978224997363 + }, + { + "round": 7, + "n_labeled": 90, + "selection_seconds": 0.0, + "round_seconds": 60.1, + "final_consolidation": true, + "val_dice": 0.939400585836005, + "val_vi": 0.04612817190652949, + "val_ece": 0.001865632105289674 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.14785451464746693, + "auroc_top20_error": 0.2738500882731163 + }, + "al_seconds": 564.9, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 1.7845958603297553e-05, + 8.009314735799591e-07, + 0.9999805751114135, + 7.779985096301431e-07 + ], + "weight_stats_final": { + "max_weight": 0.9999805751114135, + "entropy": 0.00023673447107835246, + "effective_sources": 1.0002367624948947, + "min_weight": 7.779985096301431e-07 + }, + "top_source_final": "lucchi-pp", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.9013441843656629, + "patch_iou": 0.8612735447008448, + "patch_precision": 0.9643553459614856, + "patch_recall": 0.8939486028232443, + "patch_vi": 0.08351672149892893, + "n_patches": 175, + "cal_ece": 0.008238096063656893, + "cal_nll": 0.07850860294000966, + "cal_brier": 0.01023246901159236, + "cal_mean_confidence": 0.9966647752568322, + "cal_accuracy": 0.9884933035714286, + "cal_predicted_prevalence": 0.0837890625, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08348922600161458 + }, + "test_notta": { + "patch_dice": 0.9013441843656629, + "patch_iou": 0.8612735447008448, + "patch_precision": 0.9643553459614856, + "patch_recall": 0.8939486028232443, + "patch_vi": 0.08351672149892893, + "n_patches": 175, + "cal_ece": 0.008238096063656893, + "cal_nll": 0.07850860294000966, + "cal_brier": 0.01023246901159236, + "cal_mean_confidence": 0.9966647752568322, + "cal_accuracy": 0.9884933035714286, + "cal_predicted_prevalence": 0.0837890625, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08348922600161458 + }, + "test": { + "patch_dice": 0.8904517688434722, + "patch_iou": 0.8440724660048682, + "patch_precision": 0.9614157225591229, + "patch_recall": 0.8779824872614105, + "patch_vi": 0.09238620068438104, + "n_patches": 175, + "image_dice": 0.9037671893357683, + "image_iou": 0.827489132989274, + "image_precision": 0.9575323651068441, + "image_recall": 0.8650652426532572, + "image_vi": 0.11661794111904884, + "n_images": 5, + "cal_ece": 0.0076183893408824175, + "cal_nll": 0.06957390569200572, + "cal_brier": 0.01065475153991808, + "cal_mean_confidence": 0.9947722827666771, + "cal_accuracy": 0.9871819196428572, + "cal_predicted_prevalence": 0.08293526785714286, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08272788942355813 + }, + "eval_seconds": 10.1, + "total_seconds": 911.3, + "resources": { + "phases": { + "setup": { + "seconds": 2.66, + "peak_cpu_rss_mb": 2160.1, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 18.41, + "peak_cpu_rss_mb": 4176.4, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 312.05, + "peak_cpu_rss_mb": 4656.7, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4148.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 564.86, + "peak_cpu_rss_mb": 4793.3, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4156.0, + "rounds": 6 + }, + "inference": { + "seconds": 10.07, + "peak_cpu_rss_mb": 4793.3, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4156.0, + "patches": 175, + "images": 5, + "patches_per_second": 17.4, + "seconds_per_image": 2.014 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 4793.3, + "total_seconds": 908.1 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/multiclass/macs-v11-norestart-acq-laplace_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..60e346fb9a0324f56a945b49259350f2673508d6 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-laplace/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 +1,15,58.3,252.7,"[0.2938186560944717, 1.358086727418413e-05, 0.9683060184264144, 0.01782471245039349]",,73632,14,True,kept,0.9664782881707965,0.9683060184264144,"[1.4001540225714672e-05, 7.623939953124306e-07, 0.9999846022791299, 6.337866489948326e-07]",mixture,"{'mixture': 0.9683060184264144, 'uniform': 0.1778269851676831, 'top1': 0.9683060184264144, 'top2': 0.9683060184264144, 'top3': 0.9683060184264144}",0.9367825440230815,0.04902821843243505,0.002242055302164851,,,,, +2,30,20.2,36.4,"[0.30025932706211406, 6.417579032485785e-06, 0.9682767700219854, 0.011387269224389809]",,155821,29,False,,,,"[1.493736268096847e-05, 7.625617057448709e-07, 0.9999837056298168, 5.944457965627549e-07]",mixture,"{'mixture': 0.9682767700219854, 'uniform': 0.12951527045112196, 'top1': 0.9682767700219854, 'top2': 0.9682767700219854, 'top3': 0.9682767700219854}",0.9536202317207382,0.04758055893308673,0.0013883460739741694,0.9590947745807612,0.968250747149762,ensemble,1.0, +3,45,20.0,41.7,"[0.31087552052835477, 4.15278942865923e-06, 0.965947077844261, 0.018605052939869852]",,240801,43,False,,,,"[1.7001853221602358e-05, 7.805162522607558e-07, 0.9999815636339252, 6.539966008991305e-07]",mixture,"{'mixture': 0.965947077844261, 'uniform': 0.11951632802861115, 'top1': 0.965947077844261, 'top2': 0.965947077844261, 'top3': 0.965947077844261}",0.954270906544748,0.047638874112841144,0.0013604008995015291,0.9657906130425116,0.9616518292208995,ensemble,1.0, +4,60,19.8,47.3,"[0.317738364665974, 3.048018483184082e-06, 0.9660351508741329, 0.033080268430627946]",,328081,58,False,,,,"[1.819357973396529e-05, 7.798195075771067e-07, 0.9999802714085584, 7.551922000094123e-07]",mixture,"{'mixture': 0.9660351508741329, 'uniform': 0.14769137834183466, 'top1': 0.9660351508741329, 'top2': 0.9660351508741329, 'top3': 0.9660351508741329}",0.9216831759004065,0.04882421978576393,0.0020669694208124826,0.9706596671377888,0.9662802457660312,ensemble,1.0, +5,75,19.7,53.0,"[0.32291000827099386, 2.4244775250933425e-06, 0.9639021682139222, 0.03308238053915913]",,412459,73,False,,,,"[1.957226710843179e-05, 7.96625492397049e-07, 0.9999788596188555, 7.714885436697631e-07]",mixture,"{'mixture': 0.9639021682139222, 'uniform': 0.156949700850309, 'top1': 0.9639021682139222, 'top2': 0.9639021682139222, 'top3': 0.9639021682139222}",0.9241157053196762,0.04862066420187986,0.0020597238850812863,0.964805861328671,0.955485177998267,ensemble,1.0, +6,90,19.7,58.5,"[0.31313687709842153, 2.0200632683815656e-06, 0.9633628632939749, 0.03338318269293788]",,495033,88,False,,,,"[1.7845958603297553e-05, 8.009314735799591e-07, 0.9999805751114135, 7.779985096301431e-07]",mixture,"{'mixture': 0.9633628632939749, 'uniform': 0.15983339962159931, 'top1': 0.9633628632939749, 'top2': 0.9633628632939749, 'top3': 0.9633628632939749}",0.9269402183550002,0.04773342905549056,0.0019010978224997363,0.9647581094503258,0.9606947095384917,ensemble,1.0, +7,90,0.0,60.1,,,,,,,,,,,,0.939400585836005,0.04612817190652949,0.001865632105289674,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/log.txt b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..0031b3b24d92fa57c1b5fa9a2360a820eae8d67c --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/log.txt @@ -0,0 +1,103 @@ +[06:52:05][ 0s][START] target=multiclass variant=macs-v11-norestart-acq-mcdropout budget=5.0% seed=1 | {"task": "mitochondria", "target": "multiclass", "variant": "macs-v11-norestart-acq-mcdropout", "budget": 5.0, "seed": 1, "sources": "", "run_tag": "", "arch": "unet", "teacher_ckpt": "model_best.pth", "device": "cuda:0", "num_workers": 8, "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} +[06:52:05][ 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. +[06:52:06][ 1s][DATA] multiclass: train=1855 val=350 test=175 patches +[06:52:09][ 4s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, dros-vnc=2.77 +[06:52:10][ 5s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 1s +[06:52:11][ 6s][WEIGHTS] mixture: fly-mito=0.012, kasthuri=0.012, lucchi=0.968, dros-vnc=0.008 | eff.sources=1.19 | MMD2 mix=0.0059 uniform=0.2484 +[06:52:11][ 6s][WEIGHTS] per-source: fly-mito=0.014, kasthuri=0.019, lucchi=0.962, dros-vnc=0.005 | min-MMD source (NeuroADDA rule) = lucchi +[06:52:25][ 20s][TEACHERS] target-val Dice: fly-mito=0.258, kasthuri=0.175, lucchi=0.968, dros-vnc=0.176 +[06:52:25][ 20s][TEACHERS] ensemble target-val Dice: mixture=0.967, mixture@raw=0.968, mixture@logit=0.967, per_source=0.967, per_source@raw=0.967, per_source@logit=0.966, uniform=0.248, uniform@raw=0.208, uniform@logit=0.195 +[06:52:25][ 20s][WEIGHTS] using mixture: fly-mito=0.012, kasthuri=0.012, lucchi=0.968, dros-vnc=0.008 +[06:52:25][ 20s][INIT] student initialised from teacher lucchi (weight 0.968) +[06:52:29][ 24s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[06:52:29][ 24s][KD] distilling the mixture ensemble (prob space) into the student +[06:53:07][ 62s][KD] soft labels for 1855 patches from 4 teachers in 38s +[06:54:02][ 117s][KD] epoch 1/5 | patches=1855 | loss=0.1031 | 55s +[06:54:58][ 172s][KD] epoch 2/5 | patches=1855 | loss=0.0736 | 55s +[06:55:53][ 228s][KD] epoch 3/5 | patches=1855 | loss=0.0710 | 55s +[06:56:48][ 283s][KD] epoch 4/5 | patches=1855 | loss=0.0699 | 55s +[06:57:43][ 338s][KD] epoch 5/5 | patches=1855 | loss=0.0687 | 55s +[06:57:45][ 340s][ROUND0] before any target label: val_dice=0.9623 val_vi=0.0434 +[06:57:45][ 340s][AL] patches=1855 (pool 1855 after dropping blank; free blank negatives: True) | budget=92 (5.0%) | rounds=6 | per_round=15 | acquisition=mcdropout | first_round=stratified +[06:57:57][ 351s][AL] round-1 uncertainty quality on 1855 patches: {'spearman_uncertainty_error': -0.17366018293250704, 'auroc_top20_error': 0.2635751701891152} +[06:58:35][ 390s][AL] round 1: stratified cold start: 8 patches from the membrane-rich half (ensemble prevalence >= 0.071) + 7 from the membrane-poor half of 1855 candidates +[06:58:37][ 392s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.968, uniform=0.178, top1=0.968, top2=0.968, top3=0.968 -> using mixture +[06:58:37][ 392s][WEIGHTS] round 1: student kept (labeled Dice 0.966 vs best teacher lucchi 0.968); distilling the corrected ensemble into it +[06:58:37][ 392s][WEIGHTS] round 1: teacher Dice on 15 labeled patches + 0 blank (73632 fg px): fly-mito=0.294, kasthuri=0.000, lucchi=0.968, dros-vnc=0.018 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[06:59:15][ 430s][KD] soft labels for 1840 patches from 4 teachers in 38s +[07:00:10][ 485s][KD] epoch 1/2 | patches=1840 | loss=0.0458 | 55s +[07:01:04][ 539s][KD] epoch 2/2 | patches=1840 | loss=0.0444 | 55s +[07:01:42][ 577s][CONS] pseudo-labels for 1840 unlabeled patches from the current ensemble in 38s +[07:01:44][ 579s][FT] round 1/6 epoch 1/5 | patches=15 | loss=0.2247 +[07:01:45][ 580s][FT] round 1/6 epoch 2/5 | patches=15 | loss=0.1806 +[07:01:47][ 582s][FT] round 1/6 epoch 3/5 | patches=15 | loss=0.1608 +[07:01:48][ 583s][FT] round 1/6 epoch 4/5 | patches=15 | loss=0.1641 +[07:01:50][ 585s][FT] round 1/6 epoch 5/5 | patches=15 | loss=0.1570 +[07:01:52][ 587s][AL] round 1/6 | labeled=15 | selection_seconds=49.5000 round_seconds=244.5000 restart_student_dice=0.9665 restart_teacher_dice=0.9683 val_dice=0.9173 val_vi=0.0488 val_ece=0.0020 +[07:02:03][ 598s][AL] round 2: 275/1840 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[07:02:06][ 601s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.960, uniform=0.110, top1=0.960, top2=0.960, top3=0.960 -> using mixture +[07:02:06][ 601s][WEIGHTS] round 2: teacher Dice on 30 labeled patches + 0 blank (158826 fg px): fly-mito=0.317, kasthuri=0.000, lucchi=0.960, dros-vnc=0.012 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[07:02:08][ 603s][CONS] round 2: on the 15 new patches student Dice=0.915 vs ensemble 0.952 -> pseudo-labels from the ensemble | consistency x1.00 +[07:02:10][ 605s][FT] round 2/6 epoch 1/5 | patches=30 | loss=0.2202 +[07:02:13][ 607s][FT] round 2/6 epoch 2/5 | patches=30 | loss=0.2085 +[07:02:15][ 610s][FT] round 2/6 epoch 3/5 | patches=30 | loss=0.2256 +[07:02:18][ 612s][FT] round 2/6 epoch 4/5 | patches=30 | loss=0.2274 +[07:02:20][ 615s][FT] round 2/6 epoch 5/5 | patches=30 | loss=0.2305 +[07:02:22][ 617s][AL] round 2/6 | labeled=30 | selection_seconds=11.2000 round_seconds=27.9000 pseudo_student_dice=0.9154 pseudo_ensemble_dice=0.9524 consistency_scale=1.0000 val_dice=0.9199 val_vi=0.0470 val_ece=0.0014 +[07:02:33][ 628s][AL] round 3: 267/1825 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[07:02:37][ 632s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.956, uniform=0.085, top1=0.956, top2=0.956, top3=0.956 -> using mixture +[07:02:37][ 632s][WEIGHTS] round 3: teacher Dice on 45 labeled patches + 0 blank (249309 fg px): fly-mito=0.326, kasthuri=0.000, lucchi=0.956, dros-vnc=0.008 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[07:02:39][ 634s][CONS] round 3: on the 15 new patches student Dice=0.935 vs ensemble 0.949 -> pseudo-labels from the ensemble | consistency x1.00 +[07:02:42][ 637s][FT] round 3/6 epoch 1/5 | patches=45 | loss=0.2193 +[07:02:46][ 640s][FT] round 3/6 epoch 2/5 | patches=45 | loss=0.2044 +[07:02:49][ 644s][FT] round 3/6 epoch 3/5 | patches=45 | loss=0.2324 +[07:02:52][ 647s][FT] round 3/6 epoch 4/5 | patches=45 | loss=0.1865 +[07:02:56][ 650s][FT] round 3/6 epoch 5/5 | patches=45 | loss=0.2087 +[07:02:58][ 653s][AL] round 3/6 | labeled=45 | selection_seconds=11.2000 round_seconds=33.5000 pseudo_student_dice=0.9350 pseudo_ensemble_dice=0.9486 consistency_scale=1.0000 val_dice=0.9569 val_vi=0.0439 val_ece=0.0014 +[07:03:09][ 664s][AL] round 4: 284/1810 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[07:03:14][ 669s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.953, uniform=0.093, top1=0.953, top2=0.953, top3=0.953 -> using mixture +[07:03:14][ 669s][WEIGHTS] round 4: teacher Dice on 60 labeled patches + 0 blank (345308 fg px): fly-mito=0.331, kasthuri=0.000, lucchi=0.953, dros-vnc=0.012 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[07:03:16][ 670s][CONS] round 4: on the 15 new patches student Dice=0.938 vs ensemble 0.947 -> pseudo-labels from the ensemble | consistency x1.00 +[07:03:20][ 675s][FT] round 4/6 epoch 1/5 | patches=60 | loss=0.2134 +[07:03:24][ 679s][FT] round 4/6 epoch 2/5 | patches=60 | loss=0.2314 +[07:03:28][ 683s][FT] round 4/6 epoch 3/5 | patches=60 | loss=0.1660 +[07:03:33][ 687s][FT] round 4/6 epoch 4/5 | patches=60 | loss=0.1597 +[07:03:37][ 692s][FT] round 4/6 epoch 5/5 | patches=60 | loss=0.1548 +[07:03:39][ 694s][AL] round 4/6 | labeled=60 | selection_seconds=11.1000 round_seconds=39.1000 pseudo_student_dice=0.9376 pseudo_ensemble_dice=0.9466 consistency_scale=1.0000 val_dice=0.9516 val_vi=0.0466 val_ece=0.0016 +[07:03:50][ 705s][AL] round 5: 282/1795 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[07:03:56][ 711s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.951, uniform=0.082, top1=0.951, top2=0.951, top3=0.951 -> using mixture +[07:03:56][ 711s][WEIGHTS] round 5: teacher Dice on 75 labeled patches + 0 blank (434460 fg px): fly-mito=0.337, kasthuri=0.000, lucchi=0.951, dros-vnc=0.009 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[07:03:58][ 713s][CONS] round 5: on the 15 new patches student Dice=0.939 vs ensemble 0.943 -> pseudo-labels from the ensemble | consistency x1.00 +[07:04:03][ 718s][FT] round 5/6 epoch 1/5 | patches=75 | loss=0.1598 +[07:04:08][ 723s][FT] round 5/6 epoch 2/5 | patches=75 | loss=0.1692 +[07:04:13][ 728s][FT] round 5/6 epoch 3/5 | patches=75 | loss=0.1489 +[07:04:18][ 733s][FT] round 5/6 epoch 4/5 | patches=75 | loss=0.1852 +[07:04:24][ 738s][FT] round 5/6 epoch 5/5 | patches=75 | loss=0.1564 +[07:04:26][ 741s][AL] round 5/6 | labeled=75 | selection_seconds=11.0000 round_seconds=44.5000 pseudo_student_dice=0.9388 pseudo_ensemble_dice=0.9427 consistency_scale=1.0000 val_dice=0.9669 val_vi=0.0421 val_ece=0.0010 +[07:04:37][ 751s][AL] round 6: 281/1780 candidates skipped as agreed-empty (predicted prevalence < 0.005) +[07:04:44][ 759s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.948, uniform=0.078, top1=0.948, top2=0.948, top3=0.948 -> using mixture +[07:04:44][ 759s][WEIGHTS] round 6: teacher Dice on 90 labeled patches + 0 blank (524223 fg px): fly-mito=0.337, kasthuri=0.000, lucchi=0.948, dros-vnc=0.008 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[07:04:46][ 760s][CONS] round 6: on the 15 new patches student Dice=0.940 vs ensemble 0.935 -> pseudo-labels from the ensemble | consistency x1.00 +[07:04:52][ 766s][FT] round 6/6 epoch 1/5 | patches=90 | loss=0.1420 +[07:04:58][ 772s][FT] round 6/6 epoch 2/5 | patches=90 | loss=0.1490 +[07:05:04][ 778s][FT] round 6/6 epoch 3/5 | patches=90 | loss=0.1484 +[07:05:10][ 784s][FT] round 6/6 epoch 4/5 | patches=90 | loss=0.1381 +[07:05:16][ 791s][FT] round 6/6 epoch 5/5 | patches=90 | loss=0.1317 +[07:05:18][ 793s][AL] round 6/6 | labeled=90 | selection_seconds=10.9000 round_seconds=50.0000 pseudo_student_dice=0.9402 pseudo_ensemble_dice=0.9349 consistency_scale=1.0000 val_dice=0.9590 val_vi=0.0425 val_ece=0.0015 +[07:05:24][ 799s][FT] final 1/10 epoch 1/1 | patches=90 | loss=0.1287 +[07:05:30][ 805s][FT] final 2/10 epoch 1/1 | patches=90 | loss=0.1296 +[07:05:36][ 811s][FT] final 3/10 epoch 1/1 | patches=90 | loss=0.1549 +[07:05:42][ 817s][FT] final 4/10 epoch 1/1 | patches=90 | loss=0.1405 +[07:05:48][ 823s][FT] final 5/10 epoch 1/1 | patches=90 | loss=0.1553 +[07:05:54][ 829s][FT] final 6/10 epoch 1/1 | patches=90 | loss=0.1374 +[07:06:00][ 835s][FT] final 7/10 epoch 1/1 | patches=90 | loss=0.1280 +[07:06:06][ 841s][FT] final 8/10 epoch 1/1 | patches=90 | loss=0.1297 +[07:06:13][ 847s][FT] final 9/10 epoch 1/1 | patches=90 | loss=0.1212 +[07:06:19][ 853s][FT] final 10/10 epoch 1/1 | patches=90 | loss=0.1287 +[07:06:21][ 856s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=60.7000 val_dice=0.9601 val_vi=0.0425 val_ece=0.0016 +[07:06:21][ 856s][WEIGHTS] final: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[07:06:30][ 865s][TEST] raw student (no EMA, no TTA): patch_dice=0.9012 patch_vi=0.0781 +[07:06:31][ 866s][TEST] final model without TTA: patch_dice=0.9012 patch_vi=0.0781 +[07:06:31][ 866s][RESOURCES] {"phases": {"setup": {"seconds": 2.6, "peak_cpu_rss_mb": 2174.1, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 19.91, "peak_cpu_rss_mb": 4175.9, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 313.94, "peak_cpu_rss_mb": 4654.3, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4148.0, "kd_epochs": 5}, "active_learning": {"seconds": 515.51, "peak_cpu_rss_mb": 4779.4, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4156.0, "rounds": 6}, "inference": {"seconds": 10.18, "peak_cpu_rss_mb": 4779.4, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4156.0, "patches": 175, "images": 5, "patches_per_second": 17.2, "seconds_per_image": 2.036}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 4779.4, "total_seconds": 862.1} +[07:06:31][ 866s][TEST] patch_dice=0.8906 patch_iou=0.8443 patch_precision=0.9634 patch_recall=0.8710 patch_vi=0.0889 image_dice=0.9068 image_iou=0.8336 image_precision=0.9667 image_recall=0.8625 image_vi=0.1087 cal_ece=0.0074 cal_nll=0.0643 cal_brier=0.0104 cal_mean_confidence=0.9949 cal_accuracy=0.9875 cal_predicted_prevalence=0.0819 cal_true_prevalence=0.0886 cal_mean_probability=0.0817 +[07:06:32][ 866s][DONE] multiclass macs-v11-norestart-acq-mcdropout b5 s1 | test patch Dice=0.8906 VI=0.0889 | image Dice=0.9068 VI=0.1087 | 866s diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/result.json b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/result.json new file mode 100644 index 0000000000000000000000000000000000000000..3996d22c279ff056bd95d7435d45f77114df098f --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/result.json @@ -0,0 +1,647 @@ +{ + "target": "multiclass", + "variant": "macs-v11-norestart-acq-mcdropout", + "budget": 5.0, + "seed": 1, + "config": { + "task": "mitochondria", + "target": "multiclass", + "variant": "macs-v11-norestart-acq-mcdropout", + "budget": 5.0, + "seed": 1, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "multiclass", + "--variants", + "macs-v11-norestart-acq-mcdropout", + "--budgets", + "5", + "--seeds", + "1", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "drosophila-vnc" + ], + "alphas": [ + 0.01152117817089438, + 0.011843492157429523, + 0.9683960236342841, + 0.008239306037391987 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "drosophila-vnc" + ], + "mixture": [ + 0.01152117817089438, + 0.011843492157429523, + 0.9683960236342841, + 0.008239306037391987 + ], + "per_source": [ + 0.014321460439305007, + 0.018762755437246332, + 0.9619826418368663, + 0.0049331422865822525 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.004128238346556883, + "mmd2_mixture": 0.005874251861260116, + "mmd2_uniform": 0.2484069500788153, + "mmd2_single_source": [ + 0.6121996102822926, + 0.5667987013145535, + 0.00484548842698207, + 0.6689345615903355 + ], + "median_squared_distance": 0.20474058073505946, + "gammas": [ + 1.2210573942031921, + 2.4421147884063843, + 4.884229576812769, + 9.768459153625537, + 19.536918307251074 + ], + "max_weight": 0.9683960236342841, + "entropy": 0.17460135147032324, + "effective_sources": 1.1907714227038573, + "min_weight": 0.008239306037391987 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6529838206918517, + 0.6112545706593259, + 0.0030196534926445473, + 0.8176336137299548 + ], + "standardized": [ + 0.426446119561544, + 0.2913887851649038, + -1.6771725126169683, + 0.9593376078905211 + ], + "tau": 0.5, + "max_weight": 0.9619826418368663, + "entropy": 0.19889648186407677, + "effective_sources": 1.2200556614747875, + "min_weight": 0.0049331422865822525 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "lucchi-pp", + "seconds": 1.8 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.25842427994689, + 0.17499716869513837, + 0.9675528820923397, + 0.17625340082547544 + ], + "ensemble_val_dice": { + "mixture": 0.9673725087302072, + "mixture@raw": 0.9675092802728925, + "mixture@logit": 0.9667406060014452, + "per_source": 0.9673224803379604, + "per_source@raw": 0.9674927929469518, + "per_source@logit": 0.9664724666731698, + "uniform": 0.2475083000162392, + "uniform@raw": 0.20821212102935532, + "uniform@logit": 0.19508060271550287 + }, + "val_patches": 350 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 3.80090069770813, + 2.767641305923462 + ], + "top_source": "lucchi-pp", + "weight_stats": { + "max_weight": 0.9683960236342841, + "entropy": 0.17460135147032324, + "effective_sources": 1.1907714227038573, + "min_weight": 0.008239306037391987 + }, + "correlations": { + "spearman_weight_teacher_dice": 0.39999999999999997, + "spearman_weight_taxonomy": -0.8944271909999159, + "top_source": "lucchi-pp", + "top_source_is_best_teacher": true + }, + "init_teacher": "lucchi-pp", + "student_init": "top-source", + "blank_fraction": 0.0, + "kd_seconds": 313.9, + "round0_val": { + "val_dice": 0.9623334305682075, + "val_vi": 0.04335391593733666, + "val_ece": 0.009815706574123641 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 15, + "selection_seconds": 49.5, + "round_seconds": 244.5, + "teacher_dice_labeled": [ + 0.2938186560944717, + 1.358086727418413e-05, + 0.9683060184264144, + 0.01782471245039349 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 73632, + "evidence_nonempty_patches": 14, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.9664592073305366, + "restart_teacher_dice": 0.9683060184264144, + "weights": [ + 1.4001540225714672e-05, + 7.623939953124306e-07, + 0.9999846022791299, + 6.337866489948326e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9683060184264144, + "uniform": 0.1778269851676831, + "top1": 0.9683060184264144, + "top2": 0.9683060184264144, + "top3": 0.9683060184264144 + }, + "val_dice": 0.9173456805000978, + "val_vi": 0.04880938196453225, + "val_ece": 0.0019907262106910054 + }, + { + "round": 2, + "n_labeled": 30, + "selection_seconds": 11.2, + "round_seconds": 27.9, + "teacher_dice_labeled": [ + 0.3173465856486683, + 6.296158713568852e-06, + 0.9596961177622643, + 0.012167428825233183 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 158826, + "evidence_nonempty_patches": 29, + "first_correction": false, + "restart": "", + "weights": [ + 1.930840025204062e-05, + 8.308790538811181e-07, + 0.9999792079452686, + 6.527754253695774e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9596961177622643, + "uniform": 0.10953483251709344, + "top1": 0.9596961177622643, + "top2": 0.9596961177622643, + "top3": 0.9596961177622643 + }, + "pseudo_student_dice": 0.9154265970395382, + "pseudo_ensemble_dice": 0.9523539089349594, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9199457034997295, + "val_vi": 0.04696459491845864, + "val_ece": 0.0014337311012806643 + }, + { + "round": 3, + "n_labeled": 45, + "selection_seconds": 11.2, + "round_seconds": 33.5, + "teacher_dice_labeled": [ + 0.3257213471599878, + 3.9883380993973625e-06, + 0.9557455766913427, + 0.0077667654317043755 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 249309, + "evidence_nonempty_patches": 44, + "first_correction": false, + "restart": "", + "weights": [ + 2.184103871682799e-05, + 8.643380991181191e-07, + 0.9999766447811093, + 6.498420748066025e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9557455766913427, + "uniform": 0.08522600799797364, + "top1": 0.9557455766913427, + "top2": 0.9557455766913427, + "top3": 0.9557455766913427 + }, + "pseudo_student_dice": 0.935025086731597, + "pseudo_ensemble_dice": 0.9486213968326307, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9568760978602615, + "val_vi": 0.04392632051659756, + "val_ece": 0.0014087834690751826 + }, + { + "round": 4, + "n_labeled": 60, + "selection_seconds": 11.1, + "round_seconds": 39.1, + "teacher_dice_labeled": [ + 0.3306782409870612, + 2.876687896623344e-06, + 0.953241522364928, + 0.011722199688750427 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 345308, + "evidence_nonempty_patches": 58, + "first_correction": false, + "restart": "", + "weights": [ + 2.353287628743771e-05, + 8.862434439476414e-07, + 0.9999748876775357, + 6.932027328636734e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.953241522364928, + "uniform": 0.09327275335451977, + "top1": 0.953241522364928, + "top2": 0.9532401180340814, + "top3": 0.9532401180340814 + }, + "pseudo_student_dice": 0.9375722972757856, + "pseudo_ensemble_dice": 0.9466407396693772, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9515922041033916, + "val_vi": 0.046617966582002165, + "val_ece": 0.0015584984441673006 + }, + { + "round": 5, + "n_labeled": 75, + "selection_seconds": 11.0, + "round_seconds": 44.5, + "teacher_dice_labeled": [ + 0.3370702985095187, + 2.2869950029159185e-06, + 0.9510865173473869, + 0.009315243853653518 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 434460, + "evidence_nonempty_patches": 73, + "first_correction": false, + "restart": "", + "weights": [ + 2.563265033209032e-05, + 9.055420513454941e-07, + 0.9999727703506793, + 6.914569371822995e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.951085403501211, + "uniform": 0.08158109569092956, + "top1": 0.9510865173473869, + "top2": 0.951085403501211, + "top3": 0.951085403501211 + }, + "pseudo_student_dice": 0.9387855831674218, + "pseudo_ensemble_dice": 0.9427215406757591, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9669416920542827, + "val_vi": 0.04208485860095205, + "val_ece": 0.0009632317730545605 + }, + { + "round": 6, + "n_labeled": 90, + "selection_seconds": 10.9, + "round_seconds": 50.0, + "teacher_dice_labeled": [ + 0.3367360970217586, + 1.897468397663837e-06, + 0.9482914644922257, + 0.007808602929601186 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 524223, + "evidence_nonempty_patches": 88, + "first_correction": false, + "restart": "", + "weights": [ + 2.6271239426460906e-05, + 9.312052140217418e-07, + 0.9999720971324418, + 7.004229176918938e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9482915646914972, + "uniform": 0.07816421722195963, + "top1": 0.9482914644922257, + "top2": 0.9482915646914972, + "top3": 0.9482915646914972 + }, + "pseudo_student_dice": 0.9402430808498519, + "pseudo_ensemble_dice": 0.9349307959448163, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9590052377199897, + "val_vi": 0.042470053662345776, + "val_ece": 0.0014609700344433907 + }, + { + "round": 7, + "n_labeled": 90, + "selection_seconds": 0.0, + "round_seconds": 60.7, + "final_consolidation": true, + "val_dice": 0.9601041626574933, + "val_vi": 0.0425130483770737, + "val_ece": 0.0015501542647085793 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": -0.17366018293250704, + "auroc_top20_error": 0.2635751701891152 + }, + "al_seconds": 515.5, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 2.6271239426460906e-05, + 9.312052140217418e-07, + 0.9999720971324418, + 7.004229176918938e-07 + ], + "weight_stats_final": { + "max_weight": 0.9999720971324418, + "entropy": 0.00032784372850889245, + "effective_sources": 1.0003278974751373, + "min_weight": 7.004229176918938e-07 + }, + "top_source_final": "lucchi-pp", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.9012048686045504, + "patch_iou": 0.8625053772871586, + "patch_precision": 0.9667976181399929, + "patch_recall": 0.8877684900731933, + "patch_vi": 0.0780674285127883, + "n_patches": 175, + "cal_ece": 0.0079915288534808, + "cal_nll": 0.07751496472345935, + "cal_brier": 0.009727566682766411, + "cal_mean_confidence": 0.9968754574249095, + "cal_accuracy": 0.9888839285714286, + "cal_predicted_prevalence": 0.08275111607142857, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08251441156042 + }, + "test_notta": { + "patch_dice": 0.9012048686045504, + "patch_iou": 0.8625053772871586, + "patch_precision": 0.9667976181399929, + "patch_recall": 0.8877684900731933, + "patch_vi": 0.0780674285127883, + "n_patches": 175, + "cal_ece": 0.0079915288534808, + "cal_nll": 0.07751496472345935, + "cal_brier": 0.009727566682766411, + "cal_mean_confidence": 0.9968754574249095, + "cal_accuracy": 0.9888839285714286, + "cal_predicted_prevalence": 0.08275111607142857, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08251441156042 + }, + "test": { + "patch_dice": 0.890612216595407, + "patch_iou": 0.8443423811654212, + "patch_precision": 0.9633658807751189, + "patch_recall": 0.8710363203680562, + "patch_vi": 0.08886569048031275, + "n_patches": 175, + "image_dice": 0.9067927395589909, + "image_iou": 0.8335595058787699, + "image_precision": 0.9667482658150515, + "image_recall": 0.8625173404770262, + "image_vi": 0.10868839487033517, + "n_images": 5, + "cal_ece": 0.007365289146293246, + "cal_nll": 0.06428089257028802, + "cal_brier": 0.010359269602132274, + "cal_mean_confidence": 0.9949014089486825, + "cal_accuracy": 0.9875390625, + "cal_predicted_prevalence": 0.08190848214285715, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08171287846710304 + }, + "eval_seconds": 10.2, + "total_seconds": 865.9, + "resources": { + "phases": { + "setup": { + "seconds": 2.6, + "peak_cpu_rss_mb": 2174.1, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 19.91, + "peak_cpu_rss_mb": 4175.9, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 313.94, + "peak_cpu_rss_mb": 4654.3, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4148.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 515.51, + "peak_cpu_rss_mb": 4779.4, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4156.0, + "rounds": 6 + }, + "inference": { + "seconds": 10.18, + "peak_cpu_rss_mb": 4779.4, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4156.0, + "patches": 175, + "images": 5, + "patches_per_second": 17.2, + "seconds_per_image": 2.036 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 4779.4, + "total_seconds": 862.1 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/multiclass/macs-v11-norestart-acq-mcdropout_b5_s1.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/rounds.csv b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/budget-5/seed-1/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..55d96a22afd8c488bd61e5fa45715875316b8ee7 --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-mcdropout/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 +1,15,49.5,244.5,"[0.2938186560944717, 1.358086727418413e-05, 0.9683060184264144, 0.01782471245039349]",,73632,14,True,kept,0.9664592073305366,0.9683060184264144,"[1.4001540225714672e-05, 7.623939953124306e-07, 0.9999846022791299, 6.337866489948326e-07]",mixture,"{'mixture': 0.9683060184264144, 'uniform': 0.1778269851676831, 'top1': 0.9683060184264144, 'top2': 0.9683060184264144, 'top3': 0.9683060184264144}",0.9173456805000978,0.04880938196453225,0.0019907262106910054,,,,, +2,30,11.2,27.9,"[0.3173465856486683, 6.296158713568852e-06, 0.9596961177622643, 0.012167428825233183]",,158826,29,False,,,,"[1.930840025204062e-05, 8.308790538811181e-07, 0.9999792079452686, 6.527754253695774e-07]",mixture,"{'mixture': 0.9596961177622643, 'uniform': 0.10953483251709344, 'top1': 0.9596961177622643, 'top2': 0.9596961177622643, 'top3': 0.9596961177622643}",0.9199457034997295,0.04696459491845864,0.0014337311012806643,0.9154265970395382,0.9523539089349594,ensemble,1.0, +3,45,11.2,33.5,"[0.3257213471599878, 3.9883380993973625e-06, 0.9557455766913427, 0.0077667654317043755]",,249309,44,False,,,,"[2.184103871682799e-05, 8.643380991181191e-07, 0.9999766447811093, 6.498420748066025e-07]",mixture,"{'mixture': 0.9557455766913427, 'uniform': 0.08522600799797364, 'top1': 0.9557455766913427, 'top2': 0.9557455766913427, 'top3': 0.9557455766913427}",0.9568760978602615,0.04392632051659756,0.0014087834690751826,0.935025086731597,0.9486213968326307,ensemble,1.0, +4,60,11.1,39.1,"[0.3306782409870612, 2.876687896623344e-06, 0.953241522364928, 0.011722199688750427]",,345308,58,False,,,,"[2.353287628743771e-05, 8.862434439476414e-07, 0.9999748876775357, 6.932027328636734e-07]",mixture,"{'mixture': 0.953241522364928, 'uniform': 0.09327275335451977, 'top1': 0.953241522364928, 'top2': 0.9532401180340814, 'top3': 0.9532401180340814}",0.9515922041033916,0.046617966582002165,0.0015584984441673006,0.9375722972757856,0.9466407396693772,ensemble,1.0, +5,75,11.0,44.5,"[0.3370702985095187, 2.2869950029159185e-06, 0.9510865173473869, 0.009315243853653518]",,434460,73,False,,,,"[2.563265033209032e-05, 9.055420513454941e-07, 0.9999727703506793, 6.914569371822995e-07]",mixture,"{'mixture': 0.951085403501211, 'uniform': 0.08158109569092956, 'top1': 0.9510865173473869, 'top2': 0.951085403501211, 'top3': 0.951085403501211}",0.9669416920542827,0.04208485860095205,0.0009632317730545605,0.9387855831674218,0.9427215406757591,ensemble,1.0, +6,90,10.9,50.0,"[0.3367360970217586, 1.897468397663837e-06, 0.9482914644922257, 0.007808602929601186]",,524223,88,False,,,,"[2.6271239426460906e-05, 9.312052140217418e-07, 0.9999720971324418, 7.004229176918938e-07]",mixture,"{'mixture': 0.9482915646914972, 'uniform': 0.07816421722195963, 'top1': 0.9482914644922257, 'top2': 0.9482915646914972, 'top3': 0.9482915646914972}",0.9590052377199897,0.042470053662345776,0.0014609700344433907,0.9402430808498519,0.9349307959448163,ensemble,1.0, +7,90,0.0,60.7,,,,,,,,,,,,0.9601041626574933,0.0425130483770737,0.0015501542647085793,,,,,True diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/log.txt b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/log.txt new file mode 100644 index 0000000000000000000000000000000000000000..fc9ffbe737f71658d2df532a5e2356f44408baac --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/log.txt @@ -0,0 +1,98 @@ +[13:17:00][ 0s][START] target=multiclass variant=macs-v11-norestart-acq-random budget=5.0% seed=2 | {"task": "mitochondria", "target": "multiclass", "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": 8, "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} +[13:17:00][ 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. +[13:17:00][ 0s][DATA] multiclass: train=1855 val=350 test=175 patches +[13:17:02][ 2s][CALIB] source-val temperatures: fly-mito=1.38, kasthuri=2.51, lucchi=3.80, dros-vnc=2.77 +[13:17:02][ 2s][FEATURES] 20 domain x encoder feature sets (20 cached) | n=512/domain | 0s +[13:17:03][ 3s][WEIGHTS] mixture: fly-mito=0.012, kasthuri=0.012, lucchi=0.968, dros-vnc=0.008 | eff.sources=1.19 | MMD2 mix=0.0059 uniform=0.2484 +[13:17:03][ 3s][WEIGHTS] per-source: fly-mito=0.014, kasthuri=0.019, lucchi=0.962, dros-vnc=0.005 | min-MMD source (NeuroADDA rule) = lucchi +[13:17:17][ 17s][TEACHERS] target-val Dice: fly-mito=0.258, kasthuri=0.175, lucchi=0.968, dros-vnc=0.176 +[13:17:17][ 17s][TEACHERS] ensemble target-val Dice: mixture=0.967, mixture@raw=0.968, mixture@logit=0.967, per_source=0.967, per_source@raw=0.967, per_source@logit=0.966, uniform=0.248, uniform@raw=0.208, uniform@logit=0.195 +[13:17:17][ 17s][WEIGHTS] using mixture: fly-mito=0.012, kasthuri=0.012, lucchi=0.968, dros-vnc=0.008 +[13:17:17][ 17s][INIT] student initialised from teacher lucchi (weight 0.968) +[13:17:20][ 20s][ADABN] init: BatchNorm statistics re-estimated on 512 unlabeled target patches +[13:17:20][ 20s][KD] distilling the mixture ensemble (prob space) into the student +[13:17:58][ 58s][KD] soft labels for 1855 patches from 4 teachers in 38s +[13:18:54][ 114s][KD] epoch 1/5 | patches=1855 | loss=0.1029 | 56s +[13:19:50][ 170s][KD] epoch 2/5 | patches=1855 | loss=0.0736 | 56s +[13:20:45][ 225s][KD] epoch 3/5 | patches=1855 | loss=0.0711 | 56s +[13:21:41][ 281s][KD] epoch 4/5 | patches=1855 | loss=0.0696 | 56s +[13:22:37][ 337s][KD] epoch 5/5 | patches=1855 | loss=0.0688 | 56s +[13:22:39][ 339s][ROUND0] before any target label: val_dice=0.9665 val_vi=0.0426 +[13:22:39][ 339s][AL] patches=1855 (pool 1855 after dropping blank; free blank negatives: True) | budget=92 (5.0%) | rounds=6 | per_round=15 | acquisition=random | first_round=stratified +[13:22:49][ 349s][AL] round-1 uncertainty quality on 1855 patches: {'spearman_uncertainty_error': 0.0054727604455829205, 'auroc_top20_error': 0.488566270224715} +[13:23:28][ 388s][AL] round 1: stratified cold start: 8 patches from the membrane-rich half (ensemble prevalence >= 0.071) + 7 from the membrane-poor half of 1855 candidates +[13:23:30][ 390s][WEIGHTS] round 1: labeled Dice of candidate ensembles: mixture=0.970, uniform=0.134, top1=0.970, top2=0.970, top3=0.970 -> using mixture +[13:23:31][ 390s][WEIGHTS] round 1: student kept (labeled Dice 0.972 vs best teacher lucchi 0.970); distilling the corrected ensemble into it +[13:23:31][ 390s][WEIGHTS] round 1: teacher Dice on 15 labeled patches + 0 blank (92564 fg px): fly-mito=0.364, kasthuri=0.000, lucchi=0.970, dros-vnc=0.006 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:24:09][ 429s][KD] soft labels for 1840 patches from 4 teachers in 39s +[13:25:04][ 484s][KD] epoch 1/2 | patches=1840 | loss=0.0452 | 55s +[13:26:00][ 540s][KD] epoch 2/2 | patches=1840 | loss=0.0442 | 55s +[13:26:39][ 579s][CONS] pseudo-labels for 1840 unlabeled patches from the current ensemble in 39s +[13:26:41][ 580s][FT] round 1/6 epoch 1/5 | patches=15 | loss=0.1575 +[13:26:42][ 582s][FT] round 1/6 epoch 2/5 | patches=15 | loss=0.1287 +[13:26:44][ 584s][FT] round 1/6 epoch 3/5 | patches=15 | loss=0.1285 +[13:26:45][ 585s][FT] round 1/6 epoch 4/5 | patches=15 | loss=0.1143 +[13:26:47][ 587s][FT] round 1/6 epoch 5/5 | patches=15 | loss=0.0946 +[13:26:49][ 589s][AL] round 1/6 | labeled=15 | selection_seconds=49.2000 round_seconds=248.1000 restart_student_dice=0.9715 restart_teacher_dice=0.9695 val_dice=0.9575 val_vi=0.0475 val_ece=0.0024 +[13:26:53][ 592s][WEIGHTS] round 2: labeled Dice of candidate ensembles: mixture=0.967, uniform=0.119, top1=0.967, top2=0.967, top3=0.967 -> using mixture +[13:26:53][ 592s][WEIGHTS] round 2: teacher Dice on 30 labeled patches + 0 blank (175833 fg px): fly-mito=0.349, kasthuri=0.000, lucchi=0.967, dros-vnc=0.004 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:26:54][ 594s][CONS] round 2: on the 15 new patches student Dice=0.966 vs ensemble 0.963 -> pseudo-labels from the ensemble | consistency x1.00 +[13:26:57][ 597s][FT] round 2/6 epoch 1/5 | patches=30 | loss=0.1229 +[13:26:59][ 599s][FT] round 2/6 epoch 2/5 | patches=30 | loss=0.1334 +[13:27:02][ 602s][FT] round 2/6 epoch 3/5 | patches=30 | loss=0.1382 +[13:27:04][ 604s][FT] round 2/6 epoch 4/5 | patches=30 | loss=0.1593 +[13:27:07][ 607s][FT] round 2/6 epoch 5/5 | patches=30 | loss=0.1273 +[13:27:09][ 609s][AL] round 2/6 | labeled=30 | selection_seconds=0.2000 round_seconds=17.3000 pseudo_student_dice=0.9664 pseudo_ensemble_dice=0.9634 consistency_scale=1.0000 val_dice=0.9241 val_vi=0.0501 val_ece=0.0023 +[13:27:13][ 613s][WEIGHTS] round 3: labeled Dice of candidate ensembles: mixture=0.967, uniform=0.122, top1=0.967, top2=0.967, top3=0.967 -> using mixture +[13:27:13][ 613s][WEIGHTS] round 3: teacher Dice on 45 labeled patches + 0 blank (316358 fg px): fly-mito=0.403, kasthuri=0.000, lucchi=0.967, dros-vnc=0.004 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:27:15][ 615s][CONS] round 3: on the 15 new patches student Dice=0.971 vs ensemble 0.967 -> pseudo-labels from the ensemble | consistency x1.00 +[13:27:18][ 618s][FT] round 3/6 epoch 1/5 | patches=45 | loss=0.1137 +[13:27:21][ 621s][FT] round 3/6 epoch 2/5 | patches=45 | loss=0.1220 +[13:27:25][ 625s][FT] round 3/6 epoch 3/5 | patches=45 | loss=0.1169 +[13:27:28][ 628s][FT] round 3/6 epoch 4/5 | patches=45 | loss=0.1136 +[13:27:32][ 632s][FT] round 3/6 epoch 5/5 | patches=45 | loss=0.1145 +[13:27:34][ 634s][AL] round 3/6 | labeled=45 | selection_seconds=0.2000 round_seconds=22.8000 pseudo_student_dice=0.9714 pseudo_ensemble_dice=0.9669 consistency_scale=1.0000 val_dice=0.9488 val_vi=0.0456 val_ece=0.0017 +[13:27:39][ 639s][WEIGHTS] round 4: labeled Dice of candidate ensembles: mixture=0.968, uniform=0.097, top1=0.968, top2=0.968, top3=0.968 -> using mixture +[13:27:39][ 639s][WEIGHTS] round 4: teacher Dice on 60 labeled patches + 0 blank (426649 fg px): fly-mito=0.400, kasthuri=0.000, lucchi=0.968, dros-vnc=0.004 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:27:41][ 641s][CONS] round 4: on the 15 new patches student Dice=0.975 vs ensemble 0.972 -> pseudo-labels from the ensemble | consistency x1.00 +[13:27:45][ 645s][FT] round 4/6 epoch 1/5 | patches=60 | loss=0.1508 +[13:27:49][ 649s][FT] round 4/6 epoch 2/5 | patches=60 | loss=0.1239 +[13:27:54][ 653s][FT] round 4/6 epoch 3/5 | patches=60 | loss=0.1123 +[13:27:58][ 658s][FT] round 4/6 epoch 4/5 | patches=60 | loss=0.1291 +[13:28:02][ 662s][FT] round 4/6 epoch 5/5 | patches=60 | loss=0.1231 +[13:28:05][ 664s][AL] round 4/6 | labeled=60 | selection_seconds=0.2000 round_seconds=28.4000 pseudo_student_dice=0.9755 pseudo_ensemble_dice=0.9717 consistency_scale=1.0000 val_dice=0.9585 val_vi=0.0430 val_ece=0.0017 +[13:28:11][ 671s][WEIGHTS] round 5: labeled Dice of candidate ensembles: mixture=0.968, uniform=0.099, top1=0.968, top2=0.968, top3=0.968 -> using mixture +[13:28:11][ 671s][WEIGHTS] round 5: teacher Dice on 75 labeled patches + 0 blank (490179 fg px): fly-mito=0.373, kasthuri=0.000, lucchi=0.968, dros-vnc=0.004 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:28:13][ 672s][CONS] round 5: on the 15 new patches student Dice=0.968 vs ensemble 0.964 -> pseudo-labels from the ensemble | consistency x1.00 +[13:28:18][ 678s][FT] round 5/6 epoch 1/5 | patches=75 | loss=0.1554 +[13:28:23][ 683s][FT] round 5/6 epoch 2/5 | patches=75 | loss=0.1468 +[13:28:28][ 688s][FT] round 5/6 epoch 3/5 | patches=75 | loss=0.1420 +[13:28:33][ 693s][FT] round 5/6 epoch 4/5 | patches=75 | loss=0.1303 +[13:28:39][ 699s][FT] round 5/6 epoch 5/5 | patches=75 | loss=0.1279 +[13:28:41][ 701s][AL] round 5/6 | labeled=75 | selection_seconds=0.2000 round_seconds=34.2000 pseudo_student_dice=0.9684 pseudo_ensemble_dice=0.9643 consistency_scale=1.0000 val_dice=0.9295 val_vi=0.0457 val_ece=0.0016 +[13:28:48][ 708s][WEIGHTS] round 6: labeled Dice of candidate ensembles: mixture=0.969, uniform=0.116, top1=0.969, top2=0.969, top3=0.969 -> using mixture +[13:28:48][ 708s][WEIGHTS] round 6: teacher Dice on 90 labeled patches + 0 blank (591202 fg px): fly-mito=0.373, kasthuri=0.000, lucchi=0.969, dros-vnc=0.011 | weights: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:28:50][ 710s][CONS] round 6: on the 15 new patches student Dice=0.981 vs ensemble 0.974 -> pseudo-labels from the ensemble | consistency x1.00 +[13:28:56][ 716s][FT] round 6/6 epoch 1/5 | patches=90 | loss=0.1371 +[13:29:02][ 722s][FT] round 6/6 epoch 2/5 | patches=90 | loss=0.1548 +[13:29:08][ 728s][FT] round 6/6 epoch 3/5 | patches=90 | loss=0.1484 +[13:29:15][ 734s][FT] round 6/6 epoch 4/5 | patches=90 | loss=0.1398 +[13:29:21][ 741s][FT] round 6/6 epoch 5/5 | patches=90 | loss=0.1456 +[13:29:23][ 743s][AL] round 6/6 | labeled=90 | selection_seconds=0.2000 round_seconds=39.8000 pseudo_student_dice=0.9809 pseudo_ensemble_dice=0.9744 consistency_scale=1.0000 val_dice=0.9057 val_vi=0.0609 val_ece=0.0043 +[13:29:29][ 749s][FT] final 1/10 epoch 1/1 | patches=90 | loss=0.1327 +[13:29:35][ 755s][FT] final 2/10 epoch 1/1 | patches=90 | loss=0.1318 +[13:29:42][ 761s][FT] final 3/10 epoch 1/1 | patches=90 | loss=0.1340 +[13:29:48][ 768s][FT] final 4/10 epoch 1/1 | patches=90 | loss=0.1238 +[13:29:54][ 774s][FT] final 5/10 epoch 1/1 | patches=90 | loss=0.1072 +[13:30:00][ 780s][FT] final 6/10 epoch 1/1 | patches=90 | loss=0.1099 +[13:30:06][ 786s][FT] final 7/10 epoch 1/1 | patches=90 | loss=0.1132 +[13:30:12][ 792s][FT] final 8/10 epoch 1/1 | patches=90 | loss=0.1116 +[13:30:18][ 798s][FT] final 9/10 epoch 1/1 | patches=90 | loss=0.1039 +[13:30:25][ 805s][FT] final 10/10 epoch 1/1 | patches=90 | loss=0.1019 +[13:30:27][ 807s][AL] final consolidation (10 epochs, cosine lr) | selection_seconds=0.0000 round_seconds=61.8000 val_dice=0.9343 val_vi=0.0450 val_ece=0.0019 +[13:30:27][ 807s][WEIGHTS] final: fly-mito=0.000, kasthuri=0.000, lucchi=1.000, dros-vnc=0.000 +[13:30:36][ 816s][TEST] raw student (no EMA, no TTA): patch_dice=0.9026 patch_vi=0.0807 +[13:30:37][ 817s][TEST] final model without TTA: patch_dice=0.9026 patch_vi=0.0807 +[13:30:37][ 817s][RESOURCES] {"phases": {"setup": {"seconds": 0.76, "peak_cpu_rss_mb": 2173.3, "peak_gpu_alloc_mb": 592.4, "peak_gpu_reserved_mb": 594.0}, "features_and_weights": {"seconds": 18.09, "peak_cpu_rss_mb": 4175.1, "peak_gpu_alloc_mb": 2298.2, "peak_gpu_reserved_mb": 2640.0}, "distillation": {"seconds": 317.06, "peak_cpu_rss_mb": 4653.7, "peak_gpu_alloc_mb": 3826.7, "peak_gpu_reserved_mb": 4148.0, "kd_epochs": 5}, "active_learning": {"seconds": 467.89, "peak_cpu_rss_mb": 4775.4, "peak_gpu_alloc_mb": 3947.0, "peak_gpu_reserved_mb": 4156.0, "rounds": 6}, "inference": {"seconds": 10.03, "peak_cpu_rss_mb": 4775.4, "peak_gpu_alloc_mb": 2149.5, "peak_gpu_reserved_mb": 4156.0, "patches": 175, "images": 5, "patches_per_second": 17.4, "seconds_per_image": 2.006}}, "peak_gpu_alloc_mb": 3947.0, "peak_cpu_rss_mb": 4775.4, "total_seconds": 813.8} +[13:30:37][ 817s][TEST] patch_dice=0.8918 patch_iou=0.8456 patch_precision=0.9652 patch_recall=0.8761 patch_vi=0.0894 image_dice=0.9069 image_iou=0.8332 image_precision=0.9621 image_recall=0.8655 image_vi=0.1115 cal_ece=0.0078 cal_nll=0.0716 cal_brier=0.0105 cal_mean_confidence=0.9951 cal_accuracy=0.9875 cal_predicted_prevalence=0.0822 cal_true_prevalence=0.0886 cal_mean_probability=0.0821 +[13:30:37][ 817s][DONE] multiclass macs-v11-norestart-acq-random b5 s2 | test patch Dice=0.8918 VI=0.0894 | image Dice=0.9069 VI=0.1115 | 817s diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/result.json b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/result.json new file mode 100644 index 0000000000000000000000000000000000000000..a82f34ba6e0ddaca58a22abff77d565f264f727e --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/result.json @@ -0,0 +1,647 @@ +{ + "target": "multiclass", + "variant": "macs-v11-norestart-acq-random", + "budget": 5.0, + "seed": 2, + "config": { + "task": "mitochondria", + "target": "multiclass", + "variant": "macs-v11-norestart-acq-random", + "budget": 5.0, + "seed": 2, + "sources": "", + "run_tag": "", + "arch": "unet", + "teacher_ckpt": "model_best.pth", + "models_dir": "/mnt/sdc1/connectomics/MACS-2/models", + "cache_dir": "/mnt/sdc1/connectomics/MACS-2/cache/macs_v2", + "output_dir": "experiments/macs_v2/outputs-r26", + "checkpoint_dir": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2", + "device": "cuda:0", + "num_workers": 8, + "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": "bayzid-system-2", + "python": "3.12.11", + "torch": "2.5.1+cu118", + "argv": [ + "experiments/macs_v2/run.py", + "sweep", + "--targets", + "multiclass", + "--variants", + "macs-v11-norestart-acq-random", + "--budgets", + "5", + "--seeds", + "2", + "--num-workers", + "8", + "--output-dir", + "experiments/macs_v2/outputs-v3", + "--task", + "mitochondria", + "--output-dir", + "experiments/macs_v2/outputs-r26" + ], + "cwd": "/mnt/sdc1/connectomics/MACS-2", + "gpu": "NVIDIA GeForce RTX 3070", + "cuda": "11.8", + "commit": "aecf380", + "dirty": true + }, + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "drosophila-vnc" + ], + "alphas": [ + 0.01152117817089438, + 0.011843492157429523, + 0.9683960236342841, + 0.008239306037391987 + ], + "weights_all": { + "sources": [ + "fly-brain-mito-benchmark", + "kasthuri-pp", + "lucchi-pp", + "drosophila-vnc" + ], + "mixture": [ + 0.01152117817089438, + 0.011843492157429523, + 0.9683960236342841, + 0.008239306037391987 + ], + "per_source": [ + 0.014321460439305007, + 0.018762755437246332, + 0.9619826418368663, + 0.0049331422865822525 + ], + "uniform": [ + 0.25, + 0.25, + 0.25, + 0.25 + ], + "mixture_info": { + "objective": 0.004128238346556883, + "mmd2_mixture": 0.005874251861260116, + "mmd2_uniform": 0.2484069500788153, + "mmd2_single_source": [ + 0.6121996102822926, + 0.5667987013145535, + 0.00484548842698207, + 0.6689345615903355 + ], + "median_squared_distance": 0.20474058073505946, + "gammas": [ + 1.2210573942031921, + 2.4421147884063843, + 4.884229576812769, + 9.768459153625537, + 19.536918307251074 + ], + "max_weight": 0.9683960236342841, + "entropy": 0.17460135147032324, + "effective_sources": 1.1907714227038573, + "min_weight": 0.008239306037391987 + }, + "per_source_info": { + "mmd2_own_rkhs": [ + 0.6529838206918517, + 0.6112545706593259, + 0.0030196534926445473, + 0.8176336137299548 + ], + "standardized": [ + 0.426446119561544, + 0.2913887851649038, + -1.6771725126169683, + 0.9593376078905211 + ], + "tau": 0.5, + "max_weight": 0.9619826418368663, + "entropy": 0.19889648186407677, + "effective_sources": 1.2200556614747875, + "min_weight": 0.0049331422865822525 + }, + "lam": 0.01, + "tau": 0.5, + "single_mmd_source": "lucchi-pp", + "seconds": 1.3 + }, + "teacher_diagnostics": { + "teacher_val_dice": [ + 0.25842427994689, + 0.17499716869513837, + 0.9675528820923397, + 0.17625340082547544 + ], + "ensemble_val_dice": { + "mixture": 0.9673725087302072, + "mixture@raw": 0.9675092802728925, + "mixture@logit": 0.9667406060014452, + "per_source": 0.9673224803379604, + "per_source@raw": 0.9674927929469518, + "per_source@logit": 0.9664724666731698, + "uniform": 0.2475083000162392, + "uniform@raw": 0.20821212102935532, + "uniform@logit": 0.19508060271550287 + }, + "val_patches": 350 + }, + "temperatures": [ + 1.3822230100631714, + 2.5090832710266113, + 3.80090069770813, + 2.767641305923462 + ], + "top_source": "lucchi-pp", + "weight_stats": { + "max_weight": 0.9683960236342841, + "entropy": 0.17460135147032324, + "effective_sources": 1.1907714227038573, + "min_weight": 0.008239306037391987 + }, + "correlations": { + "spearman_weight_teacher_dice": 0.39999999999999997, + "spearman_weight_taxonomy": -0.8944271909999159, + "top_source": "lucchi-pp", + "top_source_is_best_teacher": true + }, + "init_teacher": "lucchi-pp", + "student_init": "top-source", + "blank_fraction": 0.0, + "kd_seconds": 317.1, + "round0_val": { + "val_dice": 0.9664809506362138, + "val_vi": 0.042570740310092864, + "val_ece": 0.008499366466997168 + }, + "rounds": [ + { + "round": 1, + "n_labeled": 15, + "selection_seconds": 49.2, + "round_seconds": 248.1, + "teacher_dice_labeled": [ + 0.3643246058846837, + 1.0803219359369092e-05, + 0.9695205143038016, + 0.005870504217068626 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 92564, + "evidence_nonempty_patches": 15, + "first_correction": true, + "restart": "kept", + "restart_student_dice": 0.9715444706603675, + "restart_teacher_dice": 0.9695205143038016, + "weights": [ + 2.799617717272887e-05, + 7.531593559647689e-07, + 0.9999706950841266, + 5.555793445810171e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9695205143038016, + "uniform": 0.1341940295498982, + "top1": 0.9695205143038016, + "top2": 0.9695205143038016, + "top3": 0.9695205143038016 + }, + "val_dice": 0.9574874035032602, + "val_vi": 0.04754816383522154, + "val_ece": 0.0023955603806004424 + }, + { + "round": 2, + "n_labeled": 30, + "selection_seconds": 0.2, + "round_seconds": 17.3, + "teacher_dice_labeled": [ + 0.3490785549522218, + 5.687182228692972e-06, + 0.9666151600995794, + 0.0038201736958619515 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 175833, + "evidence_nonempty_patches": 30, + "first_correction": false, + "restart": "", + "weights": [ + 2.4746009095485483e-05, + 7.753251102664906e-07, + 0.9999739183139422, + 5.603518519935929e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9666151600995794, + "uniform": 0.11897515444893156, + "top1": 0.9666151600995794, + "top2": 0.9666151600995794, + "top3": 0.9666151600995794 + }, + "pseudo_student_dice": 0.9664208248123394, + "pseudo_ensemble_dice": 0.9634128001058755, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9240949079384414, + "val_vi": 0.05008252675464989, + "val_ece": 0.0022567909196813092 + }, + { + "round": 3, + "n_labeled": 45, + "selection_seconds": 0.2, + "round_seconds": 22.8, + "teacher_dice_labeled": [ + 0.40267579900842326, + 3.1609658647296267e-06, + 0.9667449033157767, + 0.004431323940818959 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 316358, + "evidence_nonempty_patches": 45, + "first_correction": false, + "restart": "", + "weights": [ + 4.223806131424505e-05, + 7.74286724331039e-07, + 0.9999564246058827, + 5.630460785874754e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9667449033157767, + "uniform": 0.12163197207015711, + "top1": 0.9667449033157767, + "top2": 0.9667449033157767, + "top3": 0.9667449033157767 + }, + "pseudo_student_dice": 0.971355759429154, + "pseudo_ensemble_dice": 0.9669085558064213, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9487542388420963, + "val_vi": 0.045581266036163054, + "val_ece": 0.0016897782894312102 + }, + { + "round": 4, + "n_labeled": 60, + "selection_seconds": 0.2, + "round_seconds": 28.4, + "teacher_dice_labeled": [ + 0.39971331751841854, + 2.343841556310793e-06, + 0.96803074489061, + 0.003988285434921942 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 426649, + "evidence_nonempty_patches": 59, + "first_correction": false, + "restart": "", + "weights": [ + 4.048130599406144e-05, + 7.643894719031811e-07, + 0.9999582009081988, + 5.533963352570889e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9680318878258827, + "uniform": 0.09669547761110801, + "top1": 0.96803074489061, + "top2": 0.9680318878258827, + "top3": 0.9680318878258827 + }, + "pseudo_student_dice": 0.9754581536430786, + "pseudo_ensemble_dice": 0.9716943865527904, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.958549319136628, + "val_vi": 0.04301469301085579, + "val_ece": 0.001705662325275334 + }, + { + "round": 5, + "n_labeled": 75, + "selection_seconds": 0.2, + "round_seconds": 34.2, + "teacher_dice_labeled": [ + 0.3733748582177732, + 2.0400669141947856e-06, + 0.9675505507895473, + 0.004090116941919932 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 490179, + "evidence_nonempty_patches": 73, + "first_correction": false, + "restart": "", + "weights": [ + 3.125773458121997e-05, + 7.680735981722924e-07, + 0.999967417560055, + 5.566317655579274e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9675505507895473, + "uniform": 0.09871244635193133, + "top1": 0.9675505507895473, + "top2": 0.9675505507895473, + "top3": 0.9675505507895473 + }, + "pseudo_student_dice": 0.9684497256326241, + "pseudo_ensemble_dice": 0.9642822782374447, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9294766279461192, + "val_vi": 0.04565340438972718, + "val_ece": 0.0015906719899368964 + }, + { + "round": 6, + "n_labeled": 90, + "selection_seconds": 0.2, + "round_seconds": 39.8, + "teacher_dice_labeled": [ + 0.37307549282809366, + 1.691466382951372e-06, + 0.9687181155392535, + 0.010721465642690713 + ], + "teacher_dice_pool": null, + "evidence_fg_pixels": 591202, + "evidence_nonempty_patches": 88, + "first_correction": false, + "restart": "", + "weights": [ + 3.0802565550486793e-05, + 7.591556727851116e-07, + 0.999967850387404, + 5.878913727004378e-07 + ], + "ensemble_choice": "mixture", + "candidate_dice": { + "mixture": 0.9687172350198938, + "uniform": 0.1162386108911664, + "top1": 0.9687181155392535, + "top2": 0.9687172350198938, + "top3": 0.9687172350198938 + }, + "pseudo_student_dice": 0.9808506099527219, + "pseudo_ensemble_dice": 0.9744040817144805, + "pseudo_source": "ensemble", + "consistency_scale": 1.0, + "val_dice": 0.9057036990387399, + "val_vi": 0.0608781853803356, + "val_ece": 0.004348281597833228 + }, + { + "round": 7, + "n_labeled": 90, + "selection_seconds": 0.0, + "round_seconds": 61.8, + "final_consolidation": true, + "val_dice": 0.9342754520438474, + "val_vi": 0.04501878291782856, + "val_ece": 0.0019047621555511002 + } + ], + "uncertainty_quality": { + "spearman_uncertainty_error": 0.0054727604455829205, + "auroc_top20_error": 0.488566270224715 + }, + "al_seconds": 467.9, + "source_archs": [ + "unet", + "unet", + "unet", + "unet" + ], + "alphas_final": [ + 3.0802565550486793e-05, + 7.591556727851116e-07, + 0.999967850387404, + 5.878913727004378e-07 + ], + "weight_stats_final": { + "max_weight": 0.999967850387404, + "entropy": 0.0003712550762131482, + "effective_sources": 1.0003713239999081, + "min_weight": 5.878913727004378e-07 + }, + "top_source_final": "lucchi-pp", + "ensemble_choice_final": "", + "restart": "kept", + "test_raw_student": { + "patch_dice": 0.9026430632906233, + "patch_iou": 0.8633032402987841, + "patch_precision": 0.9475904090022813, + "patch_recall": 0.8898730999661695, + "patch_vi": 0.08065235337995162, + "n_patches": 175, + "cal_ece": 0.008353477806945755, + "cal_nll": 0.0826653203660908, + "cal_brier": 0.009963482896945177, + "cal_mean_confidence": 0.9970253528069457, + "cal_accuracy": 0.988671875, + "cal_predicted_prevalence": 0.08271763392857143, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08245870931736407 + }, + "test_notta": { + "patch_dice": 0.9026430632906233, + "patch_iou": 0.8633032402987841, + "patch_precision": 0.9475904090022813, + "patch_recall": 0.8898730999661695, + "patch_vi": 0.08065235337995162, + "n_patches": 175, + "cal_ece": 0.008353477806945755, + "cal_nll": 0.0826653203660908, + "cal_brier": 0.009963482896945177, + "cal_mean_confidence": 0.9970253528069457, + "cal_accuracy": 0.988671875, + "cal_predicted_prevalence": 0.08271763392857143, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08245870931736407 + }, + "test": { + "patch_dice": 0.8917828004717795, + "patch_iou": 0.8455950059771068, + "patch_precision": 0.9651558159237785, + "patch_recall": 0.8761341084587753, + "patch_vi": 0.08940179843323945, + "n_patches": 175, + "image_dice": 0.9068740727765864, + "image_iou": 0.833201119212236, + "image_precision": 0.9621373509757929, + "image_recall": 0.8654878800808575, + "image_vi": 0.11154909686357835, + "n_images": 5, + "cal_ece": 0.007805100428565471, + "cal_nll": 0.07163893213079582, + "cal_brier": 0.010517803031265599, + "cal_mean_confidence": 0.9951151155141179, + "cal_accuracy": 0.9875334821428572, + "cal_predicted_prevalence": 0.0822265625, + "cal_true_prevalence": 0.08861049107142857, + "cal_mean_probability": 0.08205247475370536 + }, + "eval_seconds": 10.0, + "total_seconds": 816.9, + "resources": { + "phases": { + "setup": { + "seconds": 0.76, + "peak_cpu_rss_mb": 2173.3, + "peak_gpu_alloc_mb": 592.4, + "peak_gpu_reserved_mb": 594.0 + }, + "features_and_weights": { + "seconds": 18.09, + "peak_cpu_rss_mb": 4175.1, + "peak_gpu_alloc_mb": 2298.2, + "peak_gpu_reserved_mb": 2640.0 + }, + "distillation": { + "seconds": 317.06, + "peak_cpu_rss_mb": 4653.7, + "peak_gpu_alloc_mb": 3826.7, + "peak_gpu_reserved_mb": 4148.0, + "kd_epochs": 5 + }, + "active_learning": { + "seconds": 467.89, + "peak_cpu_rss_mb": 4775.4, + "peak_gpu_alloc_mb": 3947.0, + "peak_gpu_reserved_mb": 4156.0, + "rounds": 6 + }, + "inference": { + "seconds": 10.03, + "peak_cpu_rss_mb": 4775.4, + "peak_gpu_alloc_mb": 2149.5, + "peak_gpu_reserved_mb": 4156.0, + "patches": 175, + "images": 5, + "patches_per_second": 17.4, + "seconds_per_image": 2.006 + } + }, + "peak_gpu_alloc_mb": 3947.0, + "peak_cpu_rss_mb": 4775.4, + "total_seconds": 813.8 + }, + "student_parameters": 31042369, + "teacher_parameters": [ + 31042369, + 31042369, + 31042369, + 31042369 + ], + "checkpoint": "/mnt/sdc1/connectomics/MACS-2/models/experiments/macs_v2/multiclass/macs-v11-norestart-acq-random_b5_s2.pth" +} \ No newline at end of file diff --git a/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/rounds.csv b/macs_v2_r26/runs/mitochondria/multiclass/macs-v11-norestart-acq-random/budget-5/seed-2/rounds.csv new file mode 100644 index 0000000000000000000000000000000000000000..6b8f81237a8d9f0e8b88db7e1c6c219de2bfd33c --- /dev/null +++ b/macs_v2_r26/runs/mitochondria/multiclass/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 +1,15,49.2,248.1,"[0.3643246058846837, 1.0803219359369092e-05, 0.9695205143038016, 0.005870504217068626]",,92564,15,True,kept,0.9715444706603675,0.9695205143038016,"[2.799617717272887e-05, 7.531593559647689e-07, 0.9999706950841266, 5.555793445810171e-07]",mixture,"{'mixture': 0.9695205143038016, 'uniform': 0.1341940295498982, 'top1': 0.9695205143038016, 'top2': 0.9695205143038016, 'top3': 0.9695205143038016}",0.9574874035032602,0.04754816383522154,0.0023955603806004424,,,,, +2,30,0.2,17.3,"[0.3490785549522218, 5.687182228692972e-06, 0.9666151600995794, 0.0038201736958619515]",,175833,30,False,,,,"[2.4746009095485483e-05, 7.753251102664906e-07, 0.9999739183139422, 5.603518519935929e-07]",mixture,"{'mixture': 0.9666151600995794, 'uniform': 0.11897515444893156, 'top1': 0.9666151600995794, 'top2': 0.9666151600995794, 'top3': 0.9666151600995794}",0.9240949079384414,0.05008252675464989,0.0022567909196813092,0.9664208248123394,0.9634128001058755,ensemble,1.0, +3,45,0.2,22.8,"[0.40267579900842326, 3.1609658647296267e-06, 0.9667449033157767, 0.004431323940818959]",,316358,45,False,,,,"[4.223806131424505e-05, 7.74286724331039e-07, 0.9999564246058827, 5.630460785874754e-07]",mixture,"{'mixture': 0.9667449033157767, 'uniform': 0.12163197207015711, 'top1': 0.9667449033157767, 'top2': 0.9667449033157767, 'top3': 0.9667449033157767}",0.9487542388420963,0.045581266036163054,0.0016897782894312102,0.971355759429154,0.9669085558064213,ensemble,1.0, +4,60,0.2,28.4,"[0.39971331751841854, 2.343841556310793e-06, 0.96803074489061, 0.003988285434921942]",,426649,59,False,,,,"[4.048130599406144e-05, 7.643894719031811e-07, 0.9999582009081988, 5.533963352570889e-07]",mixture,"{'mixture': 0.9680318878258827, 'uniform': 0.09669547761110801, 'top1': 0.96803074489061, 'top2': 0.9680318878258827, 'top3': 0.9680318878258827}",0.958549319136628,0.04301469301085579,0.001705662325275334,0.9754581536430786,0.9716943865527904,ensemble,1.0, +5,75,0.2,34.2,"[0.3733748582177732, 2.0400669141947856e-06, 0.9675505507895473, 0.004090116941919932]",,490179,73,False,,,,"[3.125773458121997e-05, 7.680735981722924e-07, 0.999967417560055, 5.566317655579274e-07]",mixture,"{'mixture': 0.9675505507895473, 'uniform': 0.09871244635193133, 'top1': 0.9675505507895473, 'top2': 0.9675505507895473, 'top3': 0.9675505507895473}",0.9294766279461192,0.04565340438972718,0.0015906719899368964,0.9684497256326241,0.9642822782374447,ensemble,1.0, +6,90,0.2,39.8,"[0.37307549282809366, 1.691466382951372e-06, 0.9687181155392535, 0.010721465642690713]",,591202,88,False,,,,"[3.0802565550486793e-05, 7.591556727851116e-07, 0.999967850387404, 5.878913727004378e-07]",mixture,"{'mixture': 0.9687172350198938, 'uniform': 0.1162386108911664, 'top1': 0.9687181155392535, 'top2': 0.9687172350198938, 'top3': 0.9687172350198938}",0.9057036990387399,0.0608781853803356,0.004348281597833228,0.9808506099527219,0.9744040817144805,ensemble,1.0, +7,90,0.0,61.8,,,,,,,,,,,,0.9342754520438474,0.04501878291782856,0.0019047621555511002,,,,,True