ViperEkura 1c7369f293 feat: add MoE auxiliary loss metrics ...

- Propagates MoE load-balancing loss through model outputs
- Logs task, auxiliary, and weighted losses across strategies
- Computes only explicitly requested callback metrics
- Preserves tensor compute_loss API and adds regression tests

2026-08-02 06:30:43 +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 grad_snr metric with EMA-based gradient SNR tracking

2026-08-01 08:54:44 +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

feat: add MoE auxiliary loss metrics

2026-08-02 06:30:43 +08:00

train_callback.py

feat: add MoE auxiliary loss metrics

2026-08-02 06:30:43 +08:00

train_context.py

feat: add MoE auxiliary loss metrics

2026-08-02 06:30:43 +08:00

trainer.py

feat: add MoE auxiliary loss metrics

2026-08-02 06:30:43 +08:00