mx/AareLC_train
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 3 Wiki Activity

Commit Graph

  • ddaa5a99e6 dedup study: suppressing nested duplicate outlines on the unseen gold frames main jungfrauandClaude Fable 5.1 2026-10-05 22:26:18 +02:00
  • 8c3c538244 gold: 285 frames (batch 8 added by mount: 83 corrected frames, 2-per-mount cap, 20 siblings held out); every model re-measured jungfrauandClaude Fable 5.1 2026-10-05 15:49:30 +02:00
  • e8804c7762 gold: 202 frames (tpi_14688 dropped: same image as tpi_14689), mount groups and light states; comparison drawn by mount jungfrauandClaude Fable 5.1 2026-10-02 15:48:02 +02:00
  • 82f0d9099b gate: bootstrap by mount, light states and excluded frames of the gold set GotthardGandClaude Fable 5.1 2026-10-02 15:46:31 +02:00
  • c7d53646ce small cluster model re-gated on the Spark in one environment: gate_passed against v3; comparison after cycle 4 jungfrauandClaude Fable 5.1 2026-10-02 13:30:10 +02:00
  • f37604f87a run 2026-10-01_v4-noflip-mask1: promoted; comparison updated jungfrau 2026-10-02 13:25:42 +02:00
  • 95e6ccad86 ledger: 2026-09-29_v3-noflip-mask1 deployed at the beamlines jungfrauandClaude Fable 5.1 2026-10-02 11:30:41 +02:00
  • d0cc67cb7e compare_unseen: per-frame counts for head-to-head comparisons of any two models GotthardGandClaude Fable 5.1 2026-10-02 11:13:45 +02:00
  • 62f568b0f6 gate: baseline and candidate measured in the same environment GotthardGandClaude Fable 5.1 2026-10-02 11:02:39 +02:00
  • 831367d35d noise-aware gate, frozen core of unseen frames, and a cycle script for the Spark GotthardGandClaude Fable 5.1 2026-10-02 09:51:50 +02:00
  • 591f7b0b31 train_run: a run that already has a gate verdict is not trained again without --force jungfrauandClaude Fable 5.1 2026-10-02 07:32:48 +02:00
  • a67e6b8e8e run 2026-10-01_v4-recipeA: gate_failed; comparison updated jungfrau 2026-10-02 04:13:26 +02:00
  • ae889d1255 runs 2026-10-01_v4-recipeA and 2026-10-01_v4-noflip-mask1: exported on dataset/2026-10-01_v4 jungfrau 2026-10-01 23:52:09 +02:00
  • 7688cfa713 dataset: 2026-10-01_v4 (1610 train / 285 val, 493 new pool images) jungfrau 2026-10-01 23:52:08 +02:00
  • a9192406d3 yolo26s-seg-v3-s-640px-2026-10-01 Duan 2026-10-02 10:47:39 +02:00
  • 28cbfa889d measure dir for yolo26n-seg-v3-noflip-mask1-2026-09-30; unseen-frame comparison includes the promoted v3 model v1.3.0 jungfrauandClaude Fable 5.1 2026-09-30 13:56:58 +02:00
  • 2956303ffe run 2026-09-29_v3-noflip-mask1: promoted; comparison updated jungfrau 2026-09-30 13:56:06 +02:00
  • 9941d8d456 add ra train script duan_j 2026-09-30 11:45:26 +02:00
  • 8ff20f9f84 add train script on merlin Duan 2026-09-30 10:40:32 +02:00
  • f1a945528e cluster runs on dataset v3: resolution and model-size levers jungfrauandClaude Fable 5.1 2026-09-30 07:18:06 +02:00
  • 09cb49a3a2 ledger: v3-noflip-mask1 training on the Spark (restarted 07:13 after a memory-starved hang); v2-noflip-mask1 aborted, superseded by v3 jungfrauandClaude Fable 5.1 2026-09-30 07:16:27 +02:00
  • d562b94671 run 2026-09-29_v3-recipeA: gate_passed; comparison updated jungfrau 2026-09-29 23:42:41 +02:00
  • efba872cdf runs 2026-09-29_v3-recipeA and 2026-09-29_v3-noflip-mask1: exported on dataset/2026-09-29_v3 jungfrau 2026-09-29 22:41:13 +02:00
  • a6da2aa4aa dataset: 2026-09-29_v3 (1181 train / 210 val, 366 new pool images) dataset/2026-09-29_v3 jungfrau 2026-09-29 22:41:12 +02:00
  • 486b077f24 ledger: 2026-09-29_alc-v1-noflip-mask1 deployed at the beamlines (weights only) jungfrauandClaude Fable 5.1 2026-09-29 20:47:44 +02:00
  • b4d86d3f31 experiments: loop-type classifier, first cross-validated test (broken loops AUROC 0.985) GotthardGandClaude Fable 5.1 2026-09-29 18:18:27 +02:00
  • 0ec5138409 all models re-measured on the 203-frame gold set; comparison on the unseen frames updated jungfrau 2026-09-29 18:00:49 +02:00
  • acaf09483e slurm: base weights via LFS, and the cluster run 2026-09-29_v2-noflip-mask1 spelled out GotthardGandClaude Fable 5.1 2026-09-29 16:53:44 +02:00
  • 3d39d40560 run 2026-09-29_v2-recipeA: gate_failed; comparison updated jungfrau 2026-09-29 17:57:18 +02:00
  • ad8ae0fce4 runs 2026-09-29_v2-recipeA and 2026-09-29_v2-noflip-mask1: exported on dataset/2026-09-29_v2 jungfrau 2026-09-29 17:06:19 +02:00
  • 0742281af0 baseline: yolo26n-seg-alc-v1-noflip-mask1-2026-09-29 re-measured on the gold set grown to 203 frames jungfrau 2026-09-29 17:06:19 +02:00
  • a6efca1c10 base weights yolo26n-seg.pt under LFS, so a cluster clone needs no download jungfrau 2026-09-29 17:06:12 +02:00
  • 039cdcba1c dataset: 2026-09-29_v2 (876 train / 156 val, 315 new pool images) jungfrau 2026-09-29 17:06:12 +02:00
  • 70276e8381 dataset: 2026-09-29_v2 (876 train / 156 val, 315 new pool images) dataset/2026-09-29_v2 jungfrau 2026-09-29 17:06:12 +02:00
  • babe197cb2 recipe: no horizontal flip and full-resolution masks by default (winner of the 2026-09-28/29 comparison) GotthardGandClaude Fable 5.1 2026-09-29 11:13:00 +02:00
  • e7a02cd1ea run 2026-09-29_alc-v1-noflip-mask1: promoted (same detection as A, tightest masks); loop switches to it v1.2.0 jungfrau 2026-09-29 11:13:44 +02:00
  • 703bfc2e11 run 2026-09-29_alc-v1-noflip-mask1: gate_passed, comparison updated jungfrau 2026-09-29 10:37:36 +02:00
  • d4534fc66c run 2026-09-29_alc-v1-noflip-mask1: exported (noflip + mask_ratio 1) jungfrau 2026-09-29 07:31:18 +02:00
  • 308b4c7704 ignore Ultralytics label caches inside datasets GotthardG 2026-09-29 00:35:06 +02:00
  • d2d525487c experiments 2026-09-28: comparison on the unseen gold frames jungfrau 2026-09-29 00:31:03 +02:00
  • 40dc9e54e3 run 2026-09-28_alc-v1-800px: gate_failed jungfrau 2026-09-29 00:31:02 +02:00
  • 6000995228 run 2026-09-28_alc-v1-overlap: aborted on purpose (nested classes need overlap_mask off) jungfrau 2026-09-28 23:55:33 +02:00
  • be6f1d77a7 run 2026-09-28_alc-v1-800px: exported (training size 800 px) jungfrau 2026-09-28 23:55:28 +02:00
  • cc868e9856 run 2026-09-28_alc-v1-overlap: training jungfrau 2026-09-28 23:55:28 +02:00
  • b5a398ebb9 run 2026-09-28_alc-v1-overlap: exported (overlap_mask on) jungfrau 2026-09-28 23:54:53 +02:00
  • 13f4d6e3f7 run 2026-09-28_alc-v1-mask1: gate_failed jungfrau 2026-09-28 23:54:53 +02:00
  • 07716e7b0b run 2026-09-28_alc-v1-mask1: exported (full-resolution masks (mask_ratio=1)) jungfrau 2026-09-28 21:30:31 +02:00
  • e84bcfb658 run 2026-09-28_alc-v1-noflip: gate_passed jungfrau 2026-09-28 21:30:31 +02:00
  • 381680470c run 2026-09-28_alc-v1-noflip: exported (no horizontal flip (fliplr=0): the camera never shows the pin on the left) jungfrau 2026-09-28 21:03:25 +02:00
  • 1bdf1fc6b1 train_run: --set KEY=VALUE overrides for Ultralytics train arguments, recorded in the ledger GotthardGandClaude Fable 5.1 2026-09-28 21:01:31 +02:00
  • ed208bff95 run 2026-09-28_alc-v1-humanonly: gate passed, not promoted (A/B on clean gold: masks no tighter without accepted-unchanged labels, crystal worse) jungfrau 2026-09-28 20:44:27 +02:00
  • 84b5092766 run 2026-09-28_alc-v1-humanonly: exported (twin without accepted-unchanged labels) jungfrau 2026-09-28 20:01:45 +02:00
  • 8898f4212a dataset: 2026-09-28_alc-v1-humanonly (476 train / 84 val, 7 new pool images) dataset/2026-09-28_alc-v1-humanonly jungfrau 2026-09-28 20:01:45 +02:00
  • 46bfcb7ea5 runs/README.md: render jungfrau 2026-09-28 19:18:29 +02:00
  • 7bfd799c26 run 2026-09-28_alc-v1: promoted by the gate after the mask-term fix (F1 +0.045, mask score 0.678 -> 0.710) jungfrau 2026-09-28 19:18:27 +02:00
  • 3794330fc3 gate: mask term over all gold objects (unmatched = 0), so finding more objects is not punished; tests GotthardGandClaude Fable 5.1 2026-09-28 19:16:01 +02:00
  • e8475684d4 run 2026-09-28_alc-v1: gate_failed (F1 +0.045, mask IoU -0.022 vs the deployed model) jungfrau 2026-09-28 19:08:07 +02:00
  • e0b685507f run 2026-09-28_alc-v1: exported jungfrau 2026-09-28 16:51:45 +02:00
  • 71ee34cd17 baseline: gold measure of the deployed yolo26n-seg-overlap-false-2026-04-12 jungfrau 2026-09-28 16:51:45 +02:00
  • 489cc22dc3 dataset: 2026-09-28_alc-v1 (606 train / 108 val, 714 new pool images) dataset/2026-09-28_alc-v1 jungfrau 2026-09-28 16:49:58 +02:00
  • 34a3e93d40 measure_weights: gold measure of any weights into the registry (the gate's baseline) GotthardGandClaude Fable 5.1 2026-09-28 16:48:24 +02:00
  • 4698dbeaf9 training loop: run ledger, studio export importer with an image pool, train_run with gold gate and auto-promotion, SLURM template, CI GotthardGandClaude Fable 5.1 2026-09-28 16:06:06 +02:00
  • 098bc8f0d6 generate a ci template for later v0.1.0 duan_j 2026-04-14 16:43:58 +02:00
  • 2ab0599306 ignore local files duan_j 2026-04-14 16:41:48 +02:00
  • 4d05adb7a9 ignore some files duan_j 2026-04-14 16:22:48 +02:00
  • 4fba05631e track base models with git lfs duan_j 2026-04-14 16:15:10 +02:00
  • 391c357c84 migration and splitting AareLC duan_j 2026-04-14 16:07:31 +02:00