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
2026-05-17 11:20:13 +08:00
parent 8a11a7d444
commit a44fd22a99
5 changed files with 19 additions and 4 deletions
@@ -40,7 +40,7 @@ CONFIGS = [
            "n_routed_experts": 4,
            "n_shared_experts": 1,
            "n_activated_experts": 2,
-            "moe_topk_method": "greedy",
+            "topk_method": "greedy",
        },
        id="gqa_moe",
    ),