Datasets:

MaybeRichard
/

kermany-rae-pipeline

Modalities:

Text

Formats:

text

Size:

10K - 100K

Libraries:

Datasets

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Viewer

Auto-converted to Parquet Duplicate

Split (1)

train · 36.2k rows

text stringlengths 0 288
launched A d=0.05 seed=0 gpu=1 pid=280236 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.05_s0.json
launched A d=0.05 seed=1 gpu=2 pid=280391 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.05_s1.json
launched A d=0.05 seed=2 gpu=3 pid=280719 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.05_s2.json
launched A d=0.25 seed=0 gpu=4 pid=281046 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.25_s0.json
launched A d=0.25 seed=1 gpu=5 pid=281388 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.25_s1.json
launched A d=0.25 seed=2 gpu=6 pid=281750 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.25_s2.json
launched A d=1.0 seed=0 gpu=7 pid=282032 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d1.0_s0.json
launched A d=1.0 seed=1 gpu=1 pid=294054 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d1.0_s1.json
launched A d=1.0 seed=2 gpu=2 pid=294241 out=/data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d1.0_s2.json
A_BASELINES_DONE fail=0 Sat Jul 4 07:20:52 PM CST 2026
[kermany d0.05 A s0 ep1] val_macro_rec=44.7 (17s)
[kermany d0.05 A s0 ep2] val_macro_rec=72.0 (30s)
[kermany d0.05 A s0 ep3] val_macro_rec=81.5 (29s)
[kermany d0.05 A s0 ep4] val_macro_rec=87.1 (29s)
[kermany d0.05 A s0 ep5] val_macro_rec=87.7 (30s)
[kermany d0.05 A s0 ep6] val_macro_rec=90.1 (30s)
[kermany d0.05 A s0 ep7] val_macro_rec=90.5 (30s)
[kermany d0.05 A s0 ep8] val_macro_rec=89.6 (29s)
[kermany d0.05 A s0 ep9] val_macro_rec=90.6 (29s)
[kermany d0.05 A s0 ep10] val_macro_rec=90.9 (30s)
[kermany d0.05 A s0 ep11] val_macro_rec=91.0 (29s)
[kermany d0.05 A s0 ep12] val_macro_rec=91.0 (29s)
[kermany d0.05 A s0 ep13] val_macro_rec=91.1 (29s)
[kermany d0.05 A s0 ep14] val_macro_rec=91.0 (28s)
[kermany d0.05 A s0 ep15] val_macro_rec=90.9 (29s)
DONE /data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.05_s0.json macro=99.1 disease_mean=98.8
[kermany d0.05 A s1 ep1] val_macro_rec=40.9 (19s)
[kermany d0.05 A s1 ep2] val_macro_rec=74.8 (29s)
[kermany d0.05 A s1 ep3] val_macro_rec=77.5 (29s)
[kermany d0.05 A s1 ep4] val_macro_rec=84.9 (29s)
[kermany d0.05 A s1 ep5] val_macro_rec=88.4 (29s)
[kermany d0.05 A s1 ep6] val_macro_rec=89.5 (29s)
[kermany d0.05 A s1 ep7] val_macro_rec=90.1 (30s)
[kermany d0.05 A s1 ep8] val_macro_rec=90.3 (29s)
[kermany d0.05 A s1 ep9] val_macro_rec=90.6 (29s)
[kermany d0.05 A s1 ep10] val_macro_rec=90.2 (30s)
[kermany d0.05 A s1 ep11] val_macro_rec=90.6 (28s)
[kermany d0.05 A s1 ep12] val_macro_rec=90.8 (29s)
[kermany d0.05 A s1 ep13] val_macro_rec=90.9 (28s)
[kermany d0.05 A s1 ep14] val_macro_rec=90.8 (27s)
[kermany d0.05 A s1 ep15] val_macro_rec=90.9 (29s)
DONE /data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.05_s1.json macro=99.0 disease_mean=98.6
[kermany d0.05 A s2 ep1] val_macro_rec=47.5 (22s)
[kermany d0.05 A s2 ep2] val_macro_rec=77.6 (30s)
[kermany d0.05 A s2 ep3] val_macro_rec=65.6 (29s)
[kermany d0.05 A s2 ep4] val_macro_rec=85.7 (29s)
[kermany d0.05 A s2 ep5] val_macro_rec=88.0 (29s)
[kermany d0.05 A s2 ep6] val_macro_rec=88.9 (29s)
[kermany d0.05 A s2 ep7] val_macro_rec=90.5 (29s)
[kermany d0.05 A s2 ep8] val_macro_rec=90.5 (29s)
[kermany d0.05 A s2 ep9] val_macro_rec=90.5 (29s)
[kermany d0.05 A s2 ep10] val_macro_rec=90.5 (29s)
[kermany d0.05 A s2 ep11] val_macro_rec=90.4 (29s)
[kermany d0.05 A s2 ep12] val_macro_rec=90.8 (29s)
[kermany d0.05 A s2 ep13] val_macro_rec=90.9 (29s)
[kermany d0.05 A s2 ep14] val_macro_rec=90.8 (29s)
[kermany d0.05 A s2 ep15] val_macro_rec=90.9 (28s)
DONE /data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.05_s2.json macro=99.2 disease_mean=99.2
[kermany d0.25 A s0 ep1] val_macro_rec=84.0 (34s)
[kermany d0.25 A s0 ep2] val_macro_rec=92.2 (37s)
[kermany d0.25 A s0 ep3] val_macro_rec=92.5 (38s)
[kermany d0.25 A s0 ep4] val_macro_rec=92.9 (38s)
[kermany d0.25 A s0 ep5] val_macro_rec=92.3 (38s)
[kermany d0.25 A s0 ep6] val_macro_rec=93.8 (36s)
[kermany d0.25 A s0 ep7] val_macro_rec=93.5 (38s)
[kermany d0.25 A s0 ep8] val_macro_rec=93.8 (39s)
[kermany d0.25 A s0 ep9] val_macro_rec=94.0 (37s)
[kermany d0.25 A s0 ep10] val_macro_rec=93.9 (37s)
[kermany d0.25 A s0 ep11] val_macro_rec=93.8 (38s)
[kermany d0.25 A s0 ep12] val_macro_rec=94.2 (25s)
[kermany d0.25 A s0 ep13] val_macro_rec=94.2 (33s)
[kermany d0.25 A s0 ep14] val_macro_rec=94.2 (31s)
[kermany d0.25 A s0 ep15] val_macro_rec=94.0 (34s)
DONE /data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.25_s0.json macro=99.0 disease_mean=98.8
[kermany d0.25 A s1 ep1] val_macro_rec=87.2 (36s)
[kermany d0.25 A s1 ep2] val_macro_rec=90.6 (37s)
[kermany d0.25 A s1 ep3] val_macro_rec=91.7 (38s)
[kermany d0.25 A s1 ep4] val_macro_rec=92.5 (38s)
[kermany d0.25 A s1 ep5] val_macro_rec=92.7 (37s)
[kermany d0.25 A s1 ep6] val_macro_rec=92.7 (37s)
[kermany d0.25 A s1 ep7] val_macro_rec=93.1 (38s)
[kermany d0.25 A s1 ep8] val_macro_rec=93.7 (37s)
[kermany d0.25 A s1 ep9] val_macro_rec=93.5 (36s)
[kermany d0.25 A s1 ep10] val_macro_rec=93.7 (37s)
[kermany d0.25 A s1 ep11] val_macro_rec=93.9 (39s)
[kermany d0.25 A s1 ep12] val_macro_rec=93.9 (26s)
[kermany d0.25 A s1 ep13] val_macro_rec=93.8 (33s)
[kermany d0.25 A s1 ep14] val_macro_rec=93.7 (32s)
[kermany d0.25 A s1 ep15] val_macro_rec=93.8 (32s)
DONE /data/temp/qinshengqian/c3/kermany_downstream/jsons/A_d0.25_s1.json macro=98.9 disease_mean=98.8
[kermany d0.25 A s2 ep1] val_macro_rec=86.1 (36s)
[kermany d0.25 A s2 ep2] val_macro_rec=87.1 (38s)
[kermany d0.25 A s2 ep3] val_macro_rec=92.6 (38s)
[kermany d0.25 A s2 ep4] val_macro_rec=92.5 (37s)
[kermany d0.25 A s2 ep5] val_macro_rec=93.7 (37s)
[kermany d0.25 A s2 ep6] val_macro_rec=93.3 (38s)
[kermany d0.25 A s2 ep7] val_macro_rec=93.6 (39s)
[kermany d0.25 A s2 ep8] val_macro_rec=93.7 (36s)
[kermany d0.25 A s2 ep9] val_macro_rec=93.7 (38s)
[kermany d0.25 A s2 ep10] val_macro_rec=93.7 (38s)

End of preview. Expand in Data Studio

Kermany RAE Pipeline Snapshot

Portable snapshot for continuing the Kermany2018 OCT RAE experiments.

Included

scripts/: Kermany preparation, sampling, downstream, aggregation, and queue scripts.
configs/: Kermany configs for RAEv2 SD-VAE and OCT RAE-main FM arms.
results/kermany_downstream/: current downstream JSONs and completion markers available at snapshot time.
logs/tails/: last 500 lines of important logs.
manifests/manifest.json: counts, active arms, remote paths, and current training policy.

Not Included

Raw OCT images, generated synthetic images, and large checkpoints are intentionally excluded from this lightweight snapshot.

Active Experiment Policy

SD-VAE stops at epoch 50, then sampling and downstream run automatically.
Core FM arms only: RETFound, VisionFM, DINOv2L, MAEL.
Extra arms are disabled by default: EyeCLIP, FMUE, UrFound, SigLIP2L.
FM budget: Stage1 30 epochs, Stage2 50 epochs.

Resume Pointers

Download Kermany2018 with kagglehub.dataset_download("paultimothymooney/kermany2018").
Recreate labels and ImageFolder views with scripts in scripts/.
Regenerate configs with scripts/make_kermany_fm_configs.py.
Run scripts/run_kermany_sdvae_completion.sh or scripts/run_kermany_fm_pipeline.sh depending on current checkpoint availability.

Snapshot time: see manifests/manifest.json.

Downloads last month: 17