dela e7185cbf49 Pull OPUS-100 en-zh for SFT instead of a checked-in jsonl ...

train_sft --data opus-100 streams Helsinki-NLP/opus-100, writes both
directions, and skips frozen eval sentences. Runtime cache stays under
data/sft/ (gitignored).

2026-08-25 21:39:58 +08:00

..

correctness

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00

inference

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00

integration

Pull OPUS-100 en-zh for SFT instead of a checked-in jsonl

2026-08-25 21:39:58 +08:00

kernels

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00

__init__.py

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00