[EPLB]Record expert map without dynamic eplb. (#3409)

What this PR does / why we need it?
1.Record expert map without dynamic eplb.
2.Add export PYTHONOPTIMIZE=1  when using dynamic eplb.
3.change eplb doc

Does this PR introduce any user-facing change?
How was this patch tested?
Qwen3_moe in A3.

- vLLM version: v0.11.0

---------

Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>

vllm_ascend/torchair/quantization/torchair_w4a8_dynamic.py

@@ -134,6 +134,7 @@ class TorchairAscendW4A8DynamicFusedMoEMethod:
         self.ep_group = get_ep_group()
 
         ascend_config = get_ascend_config()
+        self.dynamic_eplb = ascend_config.dynamic_eplb or ascend_config.expert_map_record_path
         self.torchair_graph_enabled = ascend_config.torchair_graph_config.enabled
 
         vllm_config = get_current_vllm_config()
@@ -336,7 +337,8 @@ class TorchairAscendW4A8DynamicFusedMoEMethod:
                 is_torchair=self.torchair_graph_enabled,
                 quantized_x_for_share=shared_gate_up,
                 dynamic_scale_for_share=shared_dequant_scale,
-                mc2_mask=kwargs.get("mc2_mask", None))
+                mc2_mask=kwargs.get("mc2_mask", None),
+                dynamic_eplb=self.dynamic_eplb)
         else:
             # The current implementation of deepseek moe splits hidden_states
             # according to tp_size before they are feed into layers module.

vllm_ascend/torchair/quantization/torchair_w8a8_dynamic.py

@@ -848,7 +848,7 @@ class TorchairAscendW8A8DynamicFusedMoEMethod:
         self.ep_group = get_ep_group()
 
         ascend_config = get_ascend_config()
-        self.dynamic_eplb = ascend_config.dynamic_eplb
+        self.dynamic_eplb = ascend_config.dynamic_eplb or ascend_config.expert_map_record_path
         self.torchair_graph_enabled = ascend_config.torchair_graph_config.enabled
         self.enable_shared_expert_dp = ascend_config.enable_shared_expert_dp