[Test]update accuracy test of models (#4911)

### What this PR does / why we need it?
Delete accuracy tests for models that are no longer retained：
- Meta-Llama-3.1-8B-Instruct
- llava-1.5-7b-hf
- InternVL2-8B.yaml
- InternVL2_5-8B.yaml
- InternVL3-8B.yaml

Add accuracy tests for the new models：
- Llama-3.2-3B-Instruct
- llava-onevision-qwen2-0.5b-ov-hf
- Qwen3-VL-30B-A3B-Instruct

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: MrZ20 <2609716663@qq.com>

.github/workflows/nightly_test_a2.yaml

@@ -86,15 +86,13 @@ jobs:
               - Qwen3-8B-W8A8
               - Qwen3-VL-8B-Instruct
               - Qwen2.5-Omni-7B
-              - Meta-Llama-3.1-8B-Instruct
           - os: linux-aarch64-a2-1
             model_list:
               - ERNIE-4.5-21B-A3B-PT
-              - gemma-3-4b-it
-              - internlm-7b
               - InternVL3_5-8B-hf
-              - llava-1.5-7b-hf
               - Molmo-7B-D-0924
+              - Llama-3.2-3B-Instruct
+              - llava-onevision-qwen2-0.5b-ov-hf
           - os: linux-aarch64-a2-2
             model_list:
               - Qwen3-30B-A3B
@@ -103,6 +101,7 @@ jobs:
           - os: linux-aarch64-a2-4
             model_list:
               - Qwen3-Next-80B-A3B-Instruct
+              - Qwen3-VL-30B-A3B-Instruct
     uses: ./.github/workflows/_e2e_nightly_single_node_models.yaml
     with:
       vllm: v0.12.0

tests/e2e/models/configs/InternVL2-8B.yaml

@@ -1,11 +0,0 @@
-model_name: "OpenGVLab/InternVL2-8B"
-runner: "linux-aarch64-a2-1"
-hardware: "Atlas A2 Series"
-model: "vllm-vlm"
-tasks:
-  - name: "mmmu_val"
-    metrics:
-    - name: "acc,none"
-      value: 0.58
-max_model_len: 32768
-trust_remote_code: True

tests/e2e/models/configs/InternVL2_5-8B.yaml

@@ -1,11 +0,0 @@
-model_name: "OpenGVLab/InternVL2_5-8B"
-runner: "linux-aarch64-a2-1"
-hardware: "Atlas A2 Series"
-model: "vllm-vlm"
-tasks:
-  - name: "mmmu_val"
-    metrics:
-    - name: "acc,none"
-      value: 0.58
-max_model_len: 32768
-trust_remote_code: True

tests/e2e/models/configs/InternVL3-8B.yaml

@@ -1,11 +0,0 @@
-model_name: "OpenGVLab/InternVL3-8B"
-runner: "linux-aarch64-a2-1"
-hardware: "Atlas A2 Series"
-model: "vllm-vlm"
-tasks:
-  - name: "mmmu_val"
-    metrics:
-    - name: "acc,none"
-      value: 0.58
-max_model_len: 32768
-trust_remote_code: True

tests/e2e/models/configs/Llama-3.2-3B-Instruct.yaml

@@ -1,11 +1,10 @@
-model_name: "LLM-Research/Meta-Llama-3.1-8B-Instruct"
+model_name: "LLM-Research/Llama-3.2-3B-Instruct"
 hardware: "Atlas A2 Series"
 tasks:
 - name: "gsm8k"
   metrics:
   - name: "exact_match,strict-match"
-    value: 0.82
+    value: 0.71
   - name: "exact_match,flexible-extract"
-    value: 0.84
-
+    value: 0.76
 num_fewshot: 5

tests/e2e/models/configs/Qwen3-Omni-30B-A3B-Instruct.yaml

@@ -0,0 +1,11 @@
+model_name: "Qwen/Qwen3-Omni-30B-A3B-Instruct"
+hardware: "Atlas A2 Series"
+model: "vllm-vlm"
+tasks:
+- name: "mmmu_val"
+  metrics:
+  - name: "acc,none"
+    value: 0.52
+max_model_len: 8192
+tensor_parallel_size: 4
+enable_expert_parallel: True

tests/e2e/models/configs/accuracy.txt

@@ -5,13 +5,11 @@ Qwen2-Audio-7B-Instruct.yaml
 Qwen3-VL-30B-A3B-Instruct.yaml
 Qwen3-VL-8B-Instruct.yaml
 Qwen2.5-Omni-7B.yaml
-Meta-Llama-3.1-8B-Instruct.yaml
-InternVL2-8B.yaml
-InternVL2_5-8B.yaml
-InternVL3-8B.yaml
-InternVL3_5-8B.yaml
+Qwen3-Omni-30B-A3B-Instruct.yaml
+InternVL3_5-8B-hf.yaml
 ERNIE-4.5-21B-A3B-PT.yaml
 gemma-3-4b-it.yaml
 internlm3-8b-instruct.yaml
 Molmo-7B-D-0924.yaml
-llava-1.5-7b-hf.yaml
+llava-onevision-qwen2-0.5b-ov-hf.yaml
+Llama-3.2-3B-Instruct.yaml

tests/e2e/models/configs/gemma-3-4b-it.yaml

@@ -11,3 +11,4 @@ num_fewshot: 5
 apply_chat_template: False
 fewshot_as_multiturn: False
 gpu_memory_utilization: 0.7
+enforce_eager: True

tests/e2e/models/configs/llava-onevision-qwen2-0.5b-ov-hf.yaml

@@ -1,11 +1,10 @@
-model_name: "llava-hf/llava-1.5-7b-hf"
+model_name: "llava-hf/llava-onevision-qwen2-0.5b-ov-hf"
 hardware: "Atlas A2 Series"
 model: "vllm-vlm"
 tasks:
 - name: "ceval-valid"
   metrics:
   - name: "acc,none"
-    value: 0.30
+    value: 0.42
 trust_remote_code: True
 gpu_memory_utilization: 0.8
-dtype: "bfloat16"