@@ -0,0 +151,4 @@
+            update_group(g, ng) for g, ng in zip(groups, saved_groups)]
+        self.__setstate__({'state': state, 'param_groups': param_groups})
+
+    def step(self, grads, closure=None):

@@ -0,0 +52,4 @@
+        #     self.__class__.step.hooked = True
+
+    def register_step_pre_hook():
+        raise NotImplementedError("For optimizer, 'register_step_pre_hook' is not supported yet.")

@@ -0,0 +10,4 @@
+    weight = ms_torch.nn.Parameter(ms_torch.tensor([1, 2]).to(ms_torch.float))
+    opt = ms_torch.optim.SGD([weight], lr=0.01)
+    for group in opt.param_groups:
+        group['lr'] = 0.8

@@ -26,0 +23,4 @@
+from msadapter.pytorch.optim.adam import Adam
+from msadapter.pytorch.optim.adamw import AdamW
+
+__all__ = ['Momentum', 'LARS', 'Adam', 'AdamW', 'LazyAdam', 'AdamOffload',