[Test] Add accuracy test for qwen3-8b-w8a8

Signed-off-by: hfadzxy <[email protected]>

Author: zhangxinyuehfad

.github/workflows/accuracy_test.yaml

@@ -59,6 +59,8 @@ jobs:
             model_name: DeepSeek-V2-Lite
           - runner: a2-4
             model_name: Qwen3-Next-80B-A3B-Instruct
+          - runner: a2-1  
+            model_name: Qwen3-8B-W8A8
       fail-fast: false
     # test will be triggered when tag 'accuracy-test' & 'ready-for-test'
     if:  >-

tests/e2e/models/configs/Qwen3-8B-W8A8.yaml

@@ -0,0 +1,12 @@
+model_name: "vllm-ascend/Qwen3-8B-W8A8"
+hardware: "Atlas A2 Series"
+tasks:
+- name: "gsm8k"
+  metrics:
+  - name: "exact_match,strict-match"
+    value: 0.80
+  - name: "exact_match,flexible-extract"
+    value: 0.82
+num_fewshot: 5
+enable_thinking: False
+quantization: ascend

tests/e2e/models/test_lm_eval_correctness.py

@@ -48,7 +48,7 @@ def build_model_args(eval_config, tp_size):
     }
     for s in [
             "max_images", "gpu_memory_utilization", "enable_expert_parallel",
-            "tensor_parallel_size", "enforce_eager", "enable_thinking"
+            "enforce_eager", "enable_thinking", "quantization"
     ]:
         val = eval_config.get(s, None)
         if val is not None: