From 87d6424b2ee301469381d99d41936a37bc0eec91 Mon Sep 17 00:00:00 2001
From: aipaes <82140963+aipaes@users.noreply.github.com>
Date: Thu, 19 Mar 2026 16:43:29 +0800
Subject: [PATCH] [CI] Add nightly CI test cases for the GLM-4.7 model. (#7391)

### What this PR does / why we need it?
Add acc nightly CI test cases for the GLM-4.7 model.

### Does this PR introduce _any_ user-facing change?
no

### How was this patch tested?
through CI

- vLLM version: v0.17.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/4034c3d32e30d01639459edd3ab486f56993876d

---------

Signed-off-by: zjks98 <zhangjiakang4@huawei.com>
Co-authored-by: zjks98 <zhangjiakang4@huawei.com>
---
 .../workflows/schedule_nightly_test_a3.yaml   |  3 +
 .../single_node/models/configs/GLM-4.7.yaml   | 67 +++++++++++++++++++
 2 files changed, 70 insertions(+)
 create mode 100644 tests/e2e/nightly/single_node/models/configs/GLM-4.7.yaml

diff --git a/.github/workflows/schedule_nightly_test_a3.yaml b/.github/workflows/schedule_nightly_test_a3.yaml
index 5d9c2b7c..445bacd4 100644
--- a/.github/workflows/schedule_nightly_test_a3.yaml
+++ b/.github/workflows/schedule_nightly_test_a3.yaml
@@ -258,6 +258,9 @@ jobs:
           - name: deepseek-v3-2-w8a8
             os: linux-aarch64-a3-16
             config_file_path: DeepSeek-V3.2-W8A8.yaml
+          - name: glm-4.7-w8a8
+            os: linux-aarch64-a3-16
+            config_file_path: GLM-4.7.yaml
           - name: kimi-k2-thinking
             os: linux-aarch64-a3-16
             config_file_path: Kimi-K2-Thinking.yaml
diff --git a/tests/e2e/nightly/single_node/models/configs/GLM-4.7.yaml b/tests/e2e/nightly/single_node/models/configs/GLM-4.7.yaml
new file mode 100644
index 00000000..a99c45ad
--- /dev/null
+++ b/tests/e2e/nightly/single_node/models/configs/GLM-4.7.yaml
@@ -0,0 +1,67 @@
+# ==========================================
+# Shared Configurations
+# ==========================================
+
+_envs: &envs
+  HCCL_BUFFSIZE: "512"
+  SERVER_PORT: "DEFAULT_PORT"
+  OMP_PROC_BIND: "false"
+  OMP_NUM_THREADS: "1"
+  PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
+  HCCL_OP_EXPANSION_MODE: "AIV"
+  VLLM_ASCEND_BALANCE_SCHEDULING: "1"
+  VLLM_ASCEND_ENABLE_TOPK_OPTIMIZE: "1"
+  VLLM_ASCEND_ENABLE_FLASHCOMM1: "1"
+  VLLM_ASCEND_ENABLE_FUSED_MC2: "1"
+
+_server_cmd: &server_cmd
+  - "--enable-expert-parallel"
+  - "--tensor-parallel-size"
+  - "8"
+  - "--data-parallel-size"
+  - "2"
+  - "--port"
+  - "$SERVER_PORT"
+  - "--max-model-len"
+  - "8192"
+  - "--max-num-batched-tokens"
+  - "8192"
+  - "--max-num-seqs"
+  - "16"
+  - "--async-scheduling"
+  - "--quantization"
+  - "ascend"
+  - "--trust-remote-code"
+  - "--gpu-memory-utilization"
+  - "0.9"
+  - "--speculative-config"
+  - '{"num_speculative_tokens": 3, "method":"mtp"}'
+  - "--additional-config"
+  - '{"enable_shared_expert_dp": true, "ascend_fusion_config": {"fusion_ops_gmmswigluquant": false}}'
+
+_benchmarks: &benchmarks
+  acc:
+    case_type: accuracy
+    dataset_path: vllm-ascend/gsm8k-lite
+    request_conf: vllm_api_general_chat
+    dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_chat_prompt
+    max_out_len: 4096
+    batch_size: 8
+    baseline: 95
+    threshold: 5
+
+# ==========================================
+# ACTUAL TEST CASES
+# ==========================================
+
+test_cases:
+  - name: "GLM-4.7-TP8-DP2-decodegraph"
+    model: "Eco-Tech/GLM-4.7-W8A8-floatmtp"
+    envs:
+      <<: *envs
+    server_cmd: *server_cmd
+    server_cmd_extra:
+      - "--compilation-config"
+      - '{"cudagraph_capture_sizes": [1,2,4,8,16,32,64,128,256,512], "cudagraph_mode": "FULL_DECODE_ONLY"}'
+    benchmarks:
+      <<: *benchmarks