Update modeling_densebackward_olmoe0125.py

modeling_densebackward_olmoe0125.py

@@ -42,7 +42,7 @@ class DenseBackwardOlmoeSparseMoeBlock(OlmoeSparseMoeBlock):
         all_expert_outputs = torch.zeros((N_tokens, self.num_experts, hidden_dim),
                                          dtype=flat_hidden.dtype, device=flat_hidden.device)
         
-        for expert_idx in tqdm(range(self.num_experts), desc="密集计算-所有专家"):
+        for expert_idx in range(self.num_experts):
             expert_layer = self.experts[expert_idx]
             # 对所有token都计算当前专家的输出
             expert_output = expert_layer(flat_hidden)  # (N_tokens, hidden_dim)