fix: 修复训练与模型参数传递问题

- state_dict_fn 传入 CheckpointCallback,修复多卡 DDP 下 key 前缀丢失
- MLA 增加 use_qk_norm 支持,消除参数静默丢失
- moe_topk_method 统一命名为 topk_method
- checkpoint 回调移至最前
This commit is contained in:
2026-05-17 11:20:13 +08:00
parent 8a11a7d444
commit a44fd22a99
5 changed files with 19 additions and 4 deletions
+1 -1
View File
@@ -67,4 +67,4 @@ class ModelConfig(BaseModelConfig):
n_routed_experts: Optional[int] = None
n_shared_experts: Optional[int] = None
n_activated_experts: Optional[int] = None
moe_topk_method: Optional[str] = None
topk_method: Optional[str] = None