Merge branch 'v0.9.2-dev_bugfix' into 'v0.9.2-dev'

bugfix: Fix the startup crash issue of the service when USE_FUSED_RMS_QUANT=1

See merge request dcutoolkit/deeplearing/vllm!344

vllm/model_executor/layers/fused_moe/layer.py

@@ -1832,9 +1832,9 @@ def moe_forward(hidden_states: torch.Tensor, router_logits: torch.Tensor,
     self = forward_context.no_compile_layers[layer_name]
     assert self.quant_method is not None
     if envs.USE_FUSED_RMS_QUANT:
-        return self.forward_impl(hidden_states, router_logits, shared_output, i_q, i_s)
+        return self.forward_impl(hidden_states, router_logits, shared_output = shared_output, i_q = i_q, i_s = i_s)
     else:
-        return self.forward_impl(hidden_states, router_logits, shared_output)
+        return self.forward_impl(hidden_states, router_logits, shared_output = shared_output)