[EPLB]Fix ci.

Signed-off-by: offline0806 <z00858301@china.huawei.com>

Author: offline0806

vllm_ascend/eplb/core/eplb_utils.py

@@ -39,7 +39,8 @@ def determine_default_expert_map(global_expert_num, world_size, rank_id,
         end = global_expert_num
         local_count = global_expert_num - rank_id * local_num_experts
 
-    if isinstance(global_redundant_expert_num, int) and rank_id < global_redundant_expert_num:
+    if isinstance(global_redundant_expert_num,
+                  int) and rank_id < global_redundant_expert_num:
         local_count += 1
         if end < global_expert_num:
             end += 1

vllm_ascend/ops/common_fused_moe.py

@@ -246,8 +246,7 @@ def __init__(self, *args, **kwargs):
         self.global_redundant_expert_num = ascend_config.init_redundancy_expert
         # static eplb initializing with expert_map_path
         if self.expert_map_path and os.path.exists(
-                self.expert_map_path) and os.access(
-            self.expert_map_path, os.R_OK):
+                self.expert_map_path) and os.access(self.expert_map_path, os.R_OK):
             self.expert_load_balancer = ExpertLoadBalancer(
                 self.expert_map_path, self.global_num_experts)
             self.local_num_experts, self.expert_map = (
@@ -256,8 +255,7 @@ def __init__(self, *args, **kwargs):
             self.log2phy = self.expert_load_balancer.get_rank_log2phy_map(
                 self.moe_instance_id, self.ep_rank).npu()
             self.global_redundant_expert_num = (
-                self.expert_load_balancer.get_global_redundant_expert_num(
-                ))
+                self.expert_load_balancer.get_global_redundant_expert_num())
         else:
             # init moe.
             self.local_num_experts, self.expert_map = determine_expert_map(

vllm_ascend/ops/fused_moe.py

@@ -244,8 +244,7 @@ def __init__(
         self.global_num_experts = num_experts + self.global_redundant_expert_num
         # static eplb initializing with expert_map_path
         if self.expert_map_path and os.path.exists(
-                self.expert_map_path) and os.access(
-            self.expert_map_path, os.R_OK):
+                self.expert_map_path) and os.access(self.expert_map_path, os.R_OK):
             self.expert_load_balancer = ExpertLoadBalancer(
                 self.expert_map_path, self.global_num_experts)
             self.local_num_experts, self.expert_map = (
@@ -254,8 +253,7 @@ def __init__(
             self.log2phy = self.expert_load_balancer.get_rank_log2phy_map(
                 self.moe_instance_id, self.ep_rank).npu()
             self.global_redundant_expert_num = (
-                self.expert_load_balancer.get_global_redundant_expert_num(
-                ))
+                self.expert_load_balancer.get_global_redundant_expert_num())
         else:
             # init moe.
             self.local_num_experts, self.expert_map = determine_expert_map(
@@ -269,8 +267,8 @@ def __init__(
                 self.log2phy = determine_default_log2phy_map(
                     self.global_num_experts, self.ep_size, self.ep_rank,
                     self.global_redundant_expert_num)
-        local_num_experts = (torch.sum(self.expert_map != -1) if
-                             self.expert_map is not None else num_experts)
+        local_num_experts = (torch.sum(self.expert_map != -1)
+                             if self.expert_map is not None else num_experts)
         if self.dynamic_eplb:
             self.moe_load = torch.zeros(local_num_experts, dtype=torch.int64)
 

vllm_ascend/quantization/w4a8_dynamic.py

@@ -23,8 +23,8 @@
 from vllm.config import get_current_vllm_config
 from vllm.distributed import get_ep_group
 from vllm.forward_context import get_forward_context
-from vllm_ascend.ascend_config import get_ascend_config
 
+from vllm_ascend.ascend_config import get_ascend_config
 from vllm_ascend.distributed.parallel_state import get_mc2_group
 from vllm_ascend.ops.moe.experts_selector import select_experts
 

vllm_ascend/quantization/w8a8_dynamic.py

@@ -230,8 +230,7 @@ def apply(
                 w1_scale=layer.w13_weight_scale,
                 w2_scale=layer.w2_weight_scale,
                 expert_map=expert_map,
-                dynamic_eplb=self.dynamic_eplb
-            )
+                dynamic_eplb=self.dynamic_eplb)
 
         # this is a naive implementation for experts load balance so as
         # to avoid accumulating too much tokens on a single rank.

vllm_ascend/torchair/ops/torchair_fused_moe.py

@@ -1019,8 +1019,7 @@ def __init__(
         self.global_num_experts = num_experts + self.global_redundant_expert_num
         # static eplb initializing with expert_map_path
         if self.expert_map_path and os.path.exists(
-                self.expert_map_path) and os.access(
-            self.expert_map_path, os.R_OK):
+                self.expert_map_path) and os.access(self.expert_map_path, os.R_OK):
             self.expert_load_balancer = ExpertLoadBalancer(
                 self.expert_map_path, self.global_num_experts)
             self.local_num_experts, self.expert_map = (
@@ -1029,8 +1028,7 @@ def __init__(
             self.log2phy = self.expert_load_balancer.get_rank_log2phy_map(
                 self.moe_instance_id, self.ep_rank).npu()
             self.global_redundant_expert_num = (
-                self.expert_load_balancer.get_global_redundant_expert_num(
-                ))
+                self.expert_load_balancer.get_global_redundant_expert_num())
         else:
             # init moe.
             self.local_num_experts, self.expert_map = determine_expert_map(
@@ -1044,9 +1042,10 @@ def __init__(
                 self.log2phy = determine_default_log2phy_map(
                     self.global_num_experts, self.ep_size, self.ep_rank,
                     self.global_redundant_expert_num)
-        local_num_experts = (torch.sum(self.expert_map != -1) if
-                             self.expert_map is not None else num_experts)
-        self.moe_load = torch.zeros(local_num_experts, dtype=torch.int64)
+        local_num_experts = (torch.sum(self.expert_map != -1)
+                             if self.expert_map is not None else num_experts)
+        if self.dynamic_eplb:
+            self.moe_load = torch.zeros(local_num_experts, dtype=torch.int64)
 
         self.torchair_graph_enabled = ascend_config.torchair_graph_config.enabled
         self.enable_multistream_moe = \