refactor : grad_norm 指标简化，clip_grad_norm 移至 executor

- metrics 默认加入 grad_norm，移除 grad_std/max/min/mean/nan_num - grad_norm 默认返回总 L2 范数，per_param=True 返回各参数范数 - clip_grad_norm 从 callback 移至 BaseExecutor/FSDPExecutor - FSDPExecutor 覆盖为 model.clip_grad_norm_() 保证分布式正确 - ctx_get_grad_norm 改为读取 context.grad_norm
2026-06-30 14:59:43 +08:00
parent 84d4769163
commit 0f1fcb079f
6 changed files with 38 additions and 69 deletions
@@ -31,6 +31,7 @@ class TrainContext:
    epoch: int = field(default=0)
    iteration: int = field(default=0)
    loss: float = field(default=0.0)
+    grad_norm: Optional[float] = field(default=None)
    val_dataloader: Optional[DataLoader] = field(default=None)
    val_loss: Optional[float] = field(default=None)