gotthard_g

gotthard_g pushed to main at mx/AareLC_train

2026-10-26 10:29:03 +01:00

5bcb6d993d a warm start keeps the family of the registry version it continues; the 10-23 duplicates of the re-gated v5 versions go

gotthard_g pushed tag dataset/2026-10-26_v5h to mx/AareLC_train

2026-10-26 09:15:47 +01:00

gotthard_g pushed to main at mx/AareLC_train

2026-10-26 09:15:46 +01:00

9bd88eb52e unseen-frame comparison after the warm-start and v5h runs

22192ec1a3 run 2026-10-26_v5h-noflip-mask1: gate_failed

0e5b02896e run 2026-10-26_v5h-noflip-mask1: exported on dataset/2026-10-26_v5h

bb258eb2d2 dataset: 2026-10-26_v5h (1690 train / 298 val, 6 new pool images)

5dc38167c4 run 2026-10-25_v5-warm: gate_failed

Compare 5 commits »

gotthard_g pushed to main at mx/AareLC_train

2026-10-25 22:16:05 +01:00

09f7f2c4c4 yolo26s-seg-v5-s-noflip-mask1-2026-10-25 promoted by the gate on the honest measure (user decision 2026-10-25): F1 +0.021 in 96 % of draws, masks +0.011 in 90 %

d4765b3185 run 2026-10-25_v5-warm: exported on dataset/2026-10-22_v5

Compare 2 commits »

gotthard_g pushed to main at mx/AareLC_train

2026-10-25 21:36:48 +01:00

8f260fa7f4 unseen-frame comparison after the amp run

32f354cde6 run 2026-10-25_v5-amp: gate_failed

3c9a0c6704 run 2026-10-25_v5-amp: exported on dataset/2026-10-22_v5

Compare 3 commits »

gotthard_g pushed to main at mx/AareLC_train

2026-10-25 09:47:18 +01:00

63ee07fe61 a gold outline the reviewer kept from the model's proposal is not judged

gotthard_g pushed to main at mx/AareLC_train

2026-10-25 02:47:08 +02:00

845bd87bb3 a model is not measured on the gold frames it pre-labelled itself

gotthard_g pushed tag dataset/2026-10-01_v4 to mx/AareLC_train

2026-10-25 02:12:55 +02:00

gotthard_g pushed tag dataset/2026-10-22_v5 to mx/AareLC_train

2026-10-25 02:12:55 +02:00

gotthard_g pushed to main at mx/AareLC_train

2026-10-25 02:12:51 +02:00

0ed13a361c unseen-frame comparison after the 1024 px run

52e7e78b68 run 2026-10-23_v5-n-1024px: gate_failed

fe6b008312 run 2026-10-23_v5-n-1024px: exported on dataset/2026-10-22_v5

e5ad7fde30 cycle v5 stopped after the 26s run: m aborted (resolution first); unseen-frame comparison updated

19190188af run 2026-10-23_v5-m-noflip-mask1: exported on dataset/2026-10-22_v5

Compare 12 commits »

gotthard_g pushed to main at mx/AareLC_train

2026-10-05 22:25:19 +02:00

ddaa5a99e6 dedup study: suppressing nested duplicate outlines on the unseen gold frames

gotthard_g pushed to main at mx/AareLC_train

2026-10-05 15:48:35 +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

gotthard_g pushed to main at mx/AareLC_train

2026-10-02 15:47:19 +02:00

e8804c7762 gold: 202 frames (tpi_14688 dropped: same image as tpi_14689), mount groups and light states; comparison drawn by mount

gotthard_g pushed to main at mx/AareLC_train

2026-10-02 15:46:35 +02:00

82f0d9099b gate: bootstrap by mount, light states and excluded frames of the gold set

gotthard_g pushed to main at mx/AareLC_train

2026-10-02 13:29:22 +02:00

c7d53646ce small cluster model re-gated on the Spark in one environment: gate_passed against v3; comparison after cycle 4

gotthard_g pushed to main at mx/AareLC_train

2026-10-02 13:28:15 +02:00

f37604f87a run 2026-10-01_v4-noflip-mask1: promoted; comparison updated

95e6ccad86 ledger: 2026-09-29_v3-noflip-mask1 deployed at the beamlines

d0cc67cb7e compare_unseen: per-frame counts for head-to-head comparisons of any two models

62f568b0f6 gate: baseline and candidate measured in the same environment

831367d35d noise-aware gate, frozen core of unseen frames, and a cycle script for the Spark

Compare 9 commits »

gotthard_g pushed to main at mx/AareLC_train

2026-09-30 13:56:55 +02:00

28cbfa889d measure dir for yolo26n-seg-v3-noflip-mask1-2026-09-30; unseen-frame comparison includes the promoted v3 model

gotthard_g pushed to main at mx/AareLC_train

2026-09-30 13:55:21 +02:00

2956303ffe run 2026-09-29_v3-noflip-mask1: promoted; comparison updated

gotthard_g pushed tag dataset/2026-09-29_v3 to mx/AareLC_train

2026-09-30 07:17:48 +02:00

gotthard_g pushed to main at mx/AareLC_train

2026-09-30 07:17:45 +02:00

f1a945528e cluster runs on dataset v3: resolution and model-size levers

09cb49a3a2 ledger: v3-noflip-mask1 training on the Spark (restarted 07:13 after a memory-starved hang); v2-noflip-mask1 aborted, superseded by v3

d562b94671 run 2026-09-29_v3-recipeA: gate_passed; comparison updated

efba872cdf runs 2026-09-29_v3-recipeA and 2026-09-29_v3-noflip-mask1: exported on dataset/2026-09-29_v3

a6da2aa4aa dataset: 2026-09-29_v3 (1181 train / 210 val, 366 new pool images)

Compare 5 commits »