ViperEkura d9240ab149 refactor: split train context build steps ...

- separate checkpoint, model, data, and strategy setup\n- keep build orchestration concise and readable

2026-08-08 18:15:14 +08:00

..

__init__.py

chore : 升级 torch 2.11.0+cu128，移除自定义 Muon，修复 gloo device_id

2026-06-27 16:10:37 +08:00

metric_util.py

feat: add moe auxloss and metrics

2026-08-05 18:12:28 +08:00

rollout.py

fix: correct online rollout lifecycle

2026-07-23 19:01:37 +08:00

schedule.py

fix: scheduler division-by-zero, loss_mask bool

2026-07-03 22:04:55 +08:00

strategy.py

refactor: stateless MoE routing with grouped dispatch

2026-08-05 18:42:12 +08:00

train_callback.py

feat: add moe auxloss and metrics

2026-08-05 18:12:28 +08:00

train_context.py

refactor: split train context build steps

2026-08-08 18:15:14 +08:00

trainer.py

refactor: emit strategy metrics as floats

2026-08-02 06:38:28 +08:00