EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Test] Add accuracy test for qwen3-30b-a3b-w8a8 (#3807)

Browse Source 
### What this PR does / why we need it?
Add accuracy test for qwen3-30b-a3b-w8a8
This PR depends on https://github.com/vllm-project/vllm-ascend/pull/3799

### How was this patch tested?
qwen3-30b-a3b-w8a8 accuarcy test ok:

https://github.com/vllm-project/vllm-ascend/actions/runs/19062045267/job/54443732877?pr=3807
- vLLM version: v0.11.0
- vLLM main:
83f478bb19

Signed-off-by: hfadzxy <starmoon_zhang@163.com>
This commit is contained in:
zhangxinyuehfad
2025-11-04 18:56:31 +08:00
committed by GitHub
parent 5fed166a99
commit 49e6983b3b
2 changed files with 18 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.github/workflows/accuracy_test.yaml vendored
View File

@@ -68,6 +68,8 @@ jobs:
model_name: Qwen2.5-Omni-7B model_name: Qwen2.5-Omni-7B
- runner: a2-1 - runner: a2-1
model_name: Meta-Llama-3.1-8B-Instruct model_name: Meta-Llama-3.1-8B-Instruct
- runner: a2-4
model_name: Qwen3-30B-A3B-W8A8
fail-fast: false fail-fast: false
# test will be triggered when tag 'accuracy-test' & 'ready-for-test' # test will be triggered when tag 'accuracy-test' & 'ready-for-test'
if: >- if: >-

16
tests/e2e/models/configs/Qwen3-30B-A3B-W8A8.yaml Normal file
View File

@@ -0,0 +1,16 @@
model_name: "vllm-ascend/Qwen3-30B-A3B-W8A8"
hardware: "Atlas A2 Series"
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.9
- name: "exact_match,flexible-extract"
value: 0.8
num_fewshot: 5
gpu_memory_utilization: 0.7
enable_expert_parallel: True
tensor_parallel_size: 2
apply_chat_template: False
fewshot_as_multiturn: False
quantization: ascend