rogermt
/

nsgf-plusplus

Model card Files Files and versions

132 kB

Ctrl+K

Ctrl+K

1 contributor

History: 22 commits

rogermt's picture

config.yaml: CIFAR sinkhorn batch 128→32 for T4, pool batches 2500→10000 to compensate

80b1d4b verified 29 days ago

.gitattributes

1.52 kB
initial commit about 1 month ago
README.md

1.79 kB
Upload README.md about 1 month ago
SKILL.md

19.9 kB
Update SKILL.md with DataLoader batch size mismatch bug and shared state pitfall 29 days ago
config.yaml

5.52 kB
config.yaml: CIFAR sinkhorn batch 128→32 for T4, pool batches 2500→10000 to compensate 29 days ago
dataset_loader.py

6.47 kB
Fix DataLoader batch size mismatch across training phases + --train-iters now overrides all phases 29 days ago
evaluation.py

9.18 kB
Upload evaluation.py about 1 month ago
inference.py

4.82 kB
Upload inference.py about 1 month ago
main.py

11.9 kB
main.py: --resume-phase, --checkpoint-dir, --sinkhorn-batch flags 29 days ago
model.py

12.8 kB
Upload model.py about 1 month ago
requirements.txt

167 Bytes
Upload requirements.txt about 1 month ago
sinkhorn_flow.py

8.93 kB
Fix geomloss tensor shape bug for images + optimize pool sampling 29 days ago
skill-creator.md

33.2 kB
Create skill-creator.md 29 days ago
trainer.py

15.7 kB
Add checkpointing, resume, CIFAR OOM fix, --sinkhorn-batch flag 29 days ago