Skip to content

Conversation

zhangxinyuehfad
Copy link
Contributor

@zhangxinyuehfad zhangxinyuehfad commented Jun 18, 2025

What this PR does / why we need it?

Update accuracy report test

  1. Add Record commit hashes and GitHub links for both vllm and vllm-ascend in accuracy reports
  2. Add accuracy result verification checks to ensure output correctness
  3. Creat PR via forked repository workflow

Does this PR introduce any user-facing change?

How was this patch tested?

dense-accuracy-test: https://github.com/vllm-project/vllm-ascend/actions/runs/15745619485
create pr via forked repository workflow: https://github.com/zhangxinyuehfad/vllm-ascend/actions/runs/15747013719/job/44385134080
accuracy report pr: #1292

Currently, the accuracy report used is old and needs to be merged into pr, retest, update new report, then close #1292 .

@vllm-ascend-ci vllm-ascend-ci added dense-accuracy-test accuracy-test enable all accuracy test for PR ready-for-test start test by label for PR and removed dense-accuracy-test accuracy-test enable all accuracy test for PR labels Jun 18, 2025
@zhangxinyuehfad zhangxinyuehfad force-pushed the zxy_accuracy_report branch 2 times, most recently from 4832fa4 to 7052731 Compare June 18, 2025 17:50
@vllm-ascend-ci vllm-ascend-ci added ready-for-test start test by label for PR and removed ready-for-test start test by label for PR labels Jun 18, 2025
type: choice
options:
- main
- v0.9.0-dev
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
- v0.9.0-dev
- v0.9.1-dev

type: choice
options:
- main
- v0.9.0-dev
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please remove accuracy_report.yaml

  1. Please append accuracy_report after accuracy test
  2. Report only when dispatch
  3. Using bot

datasets=datasets)
model = model_name.split("/")[1]
preamble = f"""# 🎯 {model} Accuracy Test
preamble = f"""# {ACCURACY_FLAG}🎯 {model}
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I curious why this is not a markdown format to reduce maintainence cost

Please update hash as follow:

# 🎯 Qwen2.5-7B-Instruct
**vLLM Version**: vLLM: 0.9.1 ([b6553be](https://github.com/vllm-project/vllm/commit/b6553be1bc75f046b00046a4ad7576364d03c835)) , **vLLM Ascend**: refs/pull/1288/merge([c59e3895](https://github.com/vllm-project/vllm-ascend/commit/c59e3895e227dfc07cef71f0163b6ca8ad3649a6))

Preview:

🎯 Qwen2.5-7B-Instruct

vLLM Version: vLLM: 0.9.1 (b6553be) , vLLM Ascend: refs/pull/1288/merge(c59e3895)

f"| {n_shot:6} "
f"| {metric:<6} "
f"| {value:>5.4f} "
f"| {value:>5.4f} "
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add ✅ ❌ here



def main(args):
global ACCURACY_FLAG
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove this.

parser.add_argument("--torch_npu_version", type=str, required=False)
parser.add_argument("--vllm_version", type=str, required=False)
parser.add_argument("--cann_version", type=str, required=False)
parser.add_argument("--vllm_commit", type=lambda s: s[:7], required=False)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

only vllm_commit and vllm_ascend_commit are needed, other var can be generate from them

@@ -0,0 +1,28 @@
{
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's move this to run_accuracy to keep it simple

@zhangxinyuehfad zhangxinyuehfad force-pushed the zxy_accuracy_report branch 4 times, most recently from fe8dd57 to b0b31fc Compare June 23, 2025 01:41
Copy link

codecov bot commented Jun 23, 2025

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 27.21%. Comparing base (c30ddb8) to head (9a0fc54).
⚠️ Report is 548 commits behind head on main.

Additional details and impacted files
@@            Coverage Diff             @@
##             main    #1288      +/-   ##
==========================================
- Coverage   27.39%   27.21%   -0.19%     
==========================================
  Files          56       56              
  Lines        6191     6214      +23     
==========================================
- Hits         1696     1691       -5     
- Misses       4495     4523      +28     
Flag Coverage Δ
unittests 27.21% <ø> (-0.19%) ⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@zhangxinyuehfad zhangxinyuehfad force-pushed the zxy_accuracy_report branch 2 times, most recently from 10af3a6 to ae04ba2 Compare June 23, 2025 11:41
Signed-off-by: hfadzxy <starmoon_zhang@163.com>
- name: Get vLLM-Ascend commit hash and URL
working-directory: ./vllm-ascend
run: |
VLLM_ASCEND_COMMIT=$(git rev-parse HEAD)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

short

Copy link
Collaborator

@Yikun Yikun left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, let's merge this to see it work or not. some nits could address in separate PR.

run: |
VLLM_ASCEND_COMMIT=$(git rev-parse HEAD)
echo "VLLM_ASCEND_COMMIT=$VLLM_ASCEND_COMMIT" >> $GITHUB_ENV
echo "VLLM_ASCEND_COMMIT_URL=https://github.com/vllm-project/vllm-ascend/commit/$VLLM_ASCEND_COMMIT" >> $GITHUB_ENV
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

remove

@Yikun Yikun merged commit 0060886 into vllm-project:main Jun 25, 2025
20 of 22 checks passed
@Yikun
Copy link
Collaborator

Yikun commented Jun 25, 2025

Yikun pushed a commit that referenced this pull request Jun 25, 2025
### What this PR does / why we need it?
fix accuracy test:
1. fix accuracy report
like:https://vllm-ascend--1429.org.readthedocs.build/en/1429/developer_guide/evaluation/accuracy_report/Qwen2.5-7B-Instruct-V0.html
2. fix create pr for report

Signed-off-by: hfadzxy <starmoon_zhang@163.com>
weijinqian0 pushed a commit to weijinqian0/vllm-ascend that referenced this pull request Jun 30, 2025
### What this PR does / why we need it?
Update accuracy report test
1. Add Record commit hashes and GitHub links for both vllm and
vllm-ascend in accuracy reports
2. Add accuracy result verification checks to ensure output correctness
3. Creat PR via forked repository workflow

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

### How was this patch tested?
dense-accuracy-test:
https://github.com/vllm-project/vllm-ascend/actions/runs/15745619485
create pr via forked repository workflow:
https://github.com/zhangxinyuehfad/vllm-ascend/actions/runs/15747013719/job/44385134080
accuracy report pr:
vllm-project#1292

Currently, the accuracy report used is old and needs to be merged into
pr, retest, update new report, then close vllm-project#1292 .


Signed-off-by: hfadzxy <starmoon_zhang@163.com>
weijinqian0 pushed a commit to weijinqian0/vllm-ascend that referenced this pull request Jun 30, 2025
…llm-project#1435)

### What this PR does / why we need it?
fix accuracy test:
1. fix accuracy report
like:https://vllm-ascend--1429.org.readthedocs.build/en/1429/developer_guide/evaluation/accuracy_report/Qwen2.5-7B-Instruct-V0.html
2. fix create pr for report

Signed-off-by: hfadzxy <starmoon_zhang@163.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready-for-test start test by label for PR

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants