tapiocaTakeshi
/

Qubit-7B-ja-training

Model card Files Files and versions

Qubit-7B-ja-training / code

45.4 kB

Ctrl+K

Ctrl+K

1 contributor

History: 7 commits

tapiocaTakeshi's picture

train: per-rank FSDP flat-param + optimizer shard checkpoints (replace DCP), local-first then bucket copy

518d1d7 verified about 1 month ago

prep_data.py

13.9 kB
prep: stream FineWeb2 test split (avoid preparing full train split) about 1 month ago
train_qubit7b.py

31.4 kB
train: per-rank FSDP flat-param + optimizer shard checkpoints (replace DCP), local-first then bucket copy about 1 month ago