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

fix: moe quant bug and attention bug

See merge request dcutoolkit/deeplearing/vllm!275

vllm/envs.py

@@ -172,8 +172,8 @@ if TYPE_CHECKING:
     VLLM_USE_LIGHTOP_MOE_SUM: bool = False
     VLLM_USE_LIGHTOP_MOE_ALIGN: bool = False
     VLLM_USE_MERGE_ATTN_STATES_OPT: bool = False
-    USE_FUSED_RMS_QUANT: bool = True
-    USE_FUSED_SILU_MUL_QUANT: bool = True
+    USE_FUSED_RMS_QUANT: bool = False
+    USE_FUSED_SILU_MUL_QUANT: bool = False
     VLLM_P2P_ASYNC: bool = False
     VLLM_P2P_BUF_TOKENS: int = 30000
     VLLM_SCHED_ENABLE_MINIMAL_INJECTION: bool = False
@@ -1158,12 +1158,12 @@ environment_variables: dict[str, Callable[[], Any]] = {
         
     # vllm will use rmsquant fused op 
     "USE_FUSED_RMS_QUANT": 
-    lambda: bool(int(os.getenv("USE_FUSED_RMS_QUANT", "1"))),
+    lambda: bool(int(os.getenv("USE_FUSED_RMS_QUANT", "0"))),
     # vllm will use silu_mul_quant fused op,
     # This variable has a default value of true, 
     # but it is still controlled by CRQ and RQ.
     "USE_FUSED_SILU_MUL_QUANT":
-    lambda: bool(int(os.getenv("USE_FUSED_SILU_MUL_QUANT", "1"))),
+    lambda: bool(int(os.getenv("USE_FUSED_SILU_MUL_QUANT", "0"))),
     
     # vllm pd separation will be used async
     "VLLM_P2P_ASYNC":

vllm/model_executor/layers/fused_moe/layer.py

@@ -377,7 +377,7 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
         shared_output: Optional[torch.Tensor] = None,
         use_nn_moe: Optional[bool] = False,
         routed_scaling_factor: Optional[float] = 1.0,
-        use_fused_gate: Optional[bool] = False,
+        use_fused_gate: Optional[bool] = False, **_
     ) -> torch.Tensor:
         if enable_eplb:
             raise NotImplementedError(
@@ -1542,6 +1542,7 @@ class FusedMoE(torch.nn.Module):
             hidden_states, router_logits = get_ep_group().dispatch(
                 hidden_states, router_logits)
 
+        if envs.USE_FUSED_RMS_QUANT:
             # Matrix multiply.
             final_hidden_states = self.quant_method.apply(
                 layer=self,
@@ -1570,6 +1571,32 @@ class FusedMoE(torch.nn.Module):
                 i_q=i_q,
                 i_s=i_s
             )
+        else:
+            final_hidden_states = self.quant_method.apply(
+                layer=self,
+                x=hidden_states,
+                router_logits=router_logits,
+                top_k=self.top_k,
+                renormalize=self.renormalize,
+                use_grouped_topk=self.use_grouped_topk,
+                global_num_experts=self.global_num_experts,
+                expert_map=self.expert_map,
+                topk_group=self.topk_group,
+                num_expert_group=self.num_expert_group,
+                custom_routing_function=self.custom_routing_function,
+                scoring_func=self.scoring_func,
+                e_score_correction_bias=self.e_score_correction_bias,
+                activation=self.activation,
+                apply_router_weight_on_input=self.apply_router_weight_on_input,
+                enable_eplb=self.enable_eplb,
+                expert_load_view=self.expert_load_view,
+                logical_to_physical_map=self.logical_to_physical_map,
+                logical_replica_count=self.logical_replica_count,
+                shared_output=shared_output,
+                use_nn_moe=self.use_nn_moe,
+                routed_scaling_factor=self.routed_scaling_factor,
+                use_fused_gate=self.use_fused_gate,
+            )
 
         if do_naive_dispatch_combine:
             final_hidden_states = get_ep_group().combine(final_hidden_states)
@@ -1645,8 +1672,11 @@ def moe_forward(hidden_states: torch.Tensor, router_logits: torch.Tensor,
     forward_context: ForwardContext = get_forward_context()
     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)
+    else:
+        return self.forward_impl(hidden_states, router_logits, shared_output)
+        
 
 
 def moe_forward_fake(hidden_states: torch.Tensor, router_logits: torch.Tensor,

vllm/v1/attention/backends/flash_attn.py

@@ -576,6 +576,7 @@ class FlashAttentionImpl(AttentionImpl):
                         layer._k_scale, layer._v_scale
                     )
                 else:
+                    from vllm.attention.utils.fa_utils import reshape_and_cache_cuda
                     reshape_and_cache_cuda(
                         key,
                         value,