Merge branch 'feature/zerobubble' of github.com:hpcaitech/ColossalAI into dev/zero_bubble

Author: duanjunwen

colossalai/shardformer/layer/linear.py

@@ -353,9 +353,14 @@ def forward(self, input_: Tensor) -> Tuple[Tensor, Tensor]:
             )
         else:
             output_parallel = linear_with_async_comm(
-                input_parallel, self.weight, bias, self.process_group, True, fp8_communication=self.fp8_communication
+                input_parallel,
+                self.weight,
+                bias,
+                self.process_group,
+                True,
+                fp8_communication=self.fp8_communication,
+                use_zbv=self.use_zbv,
             )
-
         if self.gather_output:
             # All-gather across the partitions.
             output = gather_forward_split_backward(