Skip to content

[CI]Add EPLB CI.#3568

Merged
wangxiyuan merged 77 commits intovllm-project:mainfrom
offline893:main_1009
Oct 21, 2025
Merged

[CI]Add EPLB CI.#3568
wangxiyuan merged 77 commits intovllm-project:mainfrom
offline893:main_1009

Conversation

@offline893
Copy link
Copy Markdown
Contributor

@offline893 offline893 commented Oct 20, 2025 •
edited by github-actions bot
Loading

What this PR does / why we need it?

1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params.

Does this PR introduce any user-facing change?

How was this patch tested?

Qwen in A3.

@offline0806
[bugfix]change log2phy map to npu.
3cf7290 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]add patch to test_torch_air_init.
e2d177c 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]add patch to test torch air fused moe.
a73699a 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[]�t]add return value to patch.
08a22fa 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]add patch of torch.npu.is_availible to mock dis env.
44d09dd 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]add pytest.fixture to test torch air fused moe.
6d43490 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[bugfix]init expert map of mtp layer when using expert_map_path.
edb7593 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[bugfix]add switch to enable whether using gmm swiglu or gmm.
c836dd1 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[bugfix]fix nz tensor can not use gmm swiglu by clone buffer tensor.
702857c 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[bugfix]clone src tensor when prepare p2p task.
c0e7d7c 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[bugfix]add log2phy to apply.
549eab4 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[log]change d2d log level to warning.
4cc5cf0 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
Merge remote-tracking branch 'upstream_gitee/main' into main_1009
befd600
@offline0806
[log]add log when init expert map of mtp.
a349723 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ci]fix pre lint.
404a765 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]mock tensor to expert param per layer.
9ef551c 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]change test generate task.
1cb817c 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[[ut]fix ci.
4dade01 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ut]fix ut.
acc6aef 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[[ut]fix ut.
976ba59 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[log]add log info when init new expert_map.
fbaecb9 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[log]add log info when init new expert_map.
4b357ad 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[env]add tmp env to enable eplb.
4d13920 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
Merge remote-tracking branch 'upstream_gitee/main' into main_1009
59a4b92 
# Conflicts:
#	vllm_ascend/ops/common_fused_moe.py
#	vllm_ascend/torchair/ops/torchair_fused_moe.py
#	vllm_ascend/worker/model_runner_v1.py
@offline0806
[EPLB]record expert map without dynamic eplb.
dfd431b 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[doc]change doc of eplb.
45084b8 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
Merge remote-tracking branch 'upstream_gitee/main' into main_1009
83e926b
@offline0806
[[EPLB]Merge the judegement for dynamic eplb and expert map record path.
4cae86f 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[[ci]fix ci.
8260dac 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[log]if expert map is tensor,logonce.
6cf8676 
nce.
Signed-off-by: offline0806 <3337230449@qq.com>
@wangxiyuan wangxiyuan added the ready-for-test start test by label for PR label Oct 21, 2025
@offline0806
[CI]change eplb ci to nightly.
8662b0f 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[CI]fix pre commit.
3939a52 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[CI]fix pre commit.
d048543 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[CI]fix pre commit.
6606e35 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[CI]fix pre commit.
16581b3 
Signed-off-by: offline0806 <3337230449@qq.com>
@MengqingCao MengqingCao added the ready read for review label Oct 21, 2025
wangxiyuan
wangxiyuan reviewed Oct 21, 2025
View reviewed changes
vllm_ascend/patch/platform/patch_common/__init__.py Outdated
"EXPERT_MAP_RECORD", "false") == "true"
if dynamic_eplb:
import vllm_ascend.patch.platform.patch_common.patch_multiproc_executor # noqa
logger.warning(
Copy link
Copy Markdown
Collaborator

@wangxiyuan wangxiyuan Oct 21, 2025

Choose a reason for hiding this comment

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

logger.info

Copy link
Copy Markdown
Contributor Author

@offline893 offline893 Oct 21, 2025

Choose a reason for hiding this comment

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

Revised according to the review comments.

vllm_ascend/worker/model_runner_v1.py Outdated
dtype=torch.bool,
device=self.device,
)
self.dynamic_eplb = self.ascend_config.dynamic_eplb
Copy link
Copy Markdown
Collaborator

@wangxiyuan wangxiyuan Oct 21, 2025

Choose a reason for hiding this comment

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

duplicate with L482

Copy link
Copy Markdown
Contributor Author

@offline893 offline893 Oct 21, 2025

Choose a reason for hiding this comment

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

Revised according to the review comments.

vllm_ascend/worker/model_runner_v1.py Outdated
)
self.dynamic_eplb = self.ascend_config.dynamic_eplb
self.expert_map_record_path = self.ascend_config.expert_map_record_path
EPLBParamUtils.check_dynamic_eplb(self.ascend_config.dynamic_eplb)
Copy link
Copy Markdown
Collaborator

@wangxiyuan wangxiyuan Oct 21, 2025

Choose a reason for hiding this comment

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

if self.ascend_config.dynamic_eplb:
    EPLBParamUtils.check_dynamic_eplb(self.ascend_config.dynamic_eplb)

Copy link
Copy Markdown
Contributor Author

@offline893 offline893 Oct 21, 2025

Choose a reason for hiding this comment

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

Revised according to the review comments.

@github-actions
Copy link
Copy Markdown
Contributor

github-actions bot commented Oct 21, 2025

This pull request has conflicts, please resolve those before we can evaluate the pull request.

wangxiyuan
wangxiyuan reviewed Oct 21, 2025
View reviewed changes
tests/e2e/nightly/models/test_deepseek_r1_w8a8_eplb.py Outdated
str(tp_size), "--data-parallel-size",
str(dp_size), "--port",
str(port), "--max-model-len", "36864", "--max-num-batched-tokens",
"36864", "--block-size", "128", "--trust-remote-code", "quantization",
Copy link
Copy Markdown
Collaborator

@wangxiyuan wangxiyuan Oct 21, 2025

Choose a reason for hiding this comment

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

--quantization

Copy link
Copy Markdown
Contributor Author

@offline893 offline893 Oct 21, 2025

Choose a reason for hiding this comment

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

Revised according to the review comments.

@offline0806
[CI]fix ci.
1e213e7 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
Merge remote-tracking branch 'upstream_gitee/main' into main_1009
3f2ef47 
# Conflicts:
#	.github/workflows/vllm_ascend_test_nightly.yaml
#	vllm_ascend/patch/platform/patch_common/__init__.py
@offline0806
[patch]change patch of v1 executor.
ef0c94f 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[CI]fix pre commit.
6f52bbd 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[CI]fix ci.
033eb79 
Signed-off-by: offline0806 <3337230449@qq.com>
@offline0806
[ci[ci]skip mooncake ut.
1325287 
Signed-off-by: offline0806 <3337230449@qq.com>
@wangxiyuan wangxiyuan merged commit e916265 into vllm-project:main Oct 21, 2025
16 of 17 checks passed
Yikun
Yikun reviewed Oct 21, 2025
View reviewed changes
Copy link
Copy Markdown
Member

@Yikun Yikun left a comment •
edited
Loading

Choose a reason for hiding this comment

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

I don't know why this PR can be merged. @wangxiyuan

  1. This PR remove many exsiting tests like qwen3-32b-in8-a3, qwen3-32b-in8-a2
  2. It was merged without test pass, so if no special reason, I prefer to revert this PR.

@wangxiyuan
Copy link
Copy Markdown
Collaborator

wangxiyuan commented Oct 22, 2025

  1. this is a mistake by rebase. @offline0806 will create a new PR to add it back.
  2. test failure due to other problem, not releated to this PR. and it has been fixed on main now.

ZYang6263 pushed a commit to rjg-lyh/vllm-ascend that referenced this pull request Oct 23, 2025
@offline893 @offline0806 @ZYang6263
[CI]Add EPLB CI. (vllm-project#3568)
a81d1ad 
### What this PR does / why we need it?
1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params. 
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
Qwen in A3.


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>
luolun pushed a commit to luolun/vllm-ascend that referenced this pull request Nov 19, 2025
@offline893 @offline0806
[CI]Add EPLB CI. (vllm-project#3568)
efc12c5 
### What this PR does / why we need it?
1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params. 
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
Qwen in A3.


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>
Signed-off-by: luolun <luolun1995@cmbchina.com>
hwhaokun pushed a commit to hwhaokun/vllm-ascend that referenced this pull request Nov 19, 2025
@offline893 @offline0806 @hwhaokun
[CI]Add EPLB CI. (vllm-project#3568)
85307d7 
### What this PR does / why we need it?
1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params.
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
Qwen in A3.

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>
Signed-off-by: hwhaokun <haokun0405@163.com>
NSDie pushed a commit to NSDie/vllm-ascend that referenced this pull request Nov 24, 2025
@offline893 @offline0806 @NSDie
[CI]Add EPLB CI. (vllm-project#3568)
7f08a18 
### What this PR does / why we need it?
1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params.
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
Qwen in A3.

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>
Signed-off-by: nsdie <yeyifan@huawei.com>
Clorist33 pushed a commit to Clorist33/vllm-ascend that referenced this pull request Dec 9, 2025
@offline893 @offline0806
[CI]Add EPLB CI. (vllm-project#3568)
61b60af 
### What this PR does / why we need it?
1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params. 
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
Qwen in A3.


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>
@MrZ20 MrZ20 mentioned this pull request Mar 2, 2026
Merged
wangxiyuan pushed a commit that referenced this pull request Mar 3, 2026
@MrZ20
[Nightly][Refactor]Migrate nightly single-node model tests from .py…
859f2c2 
… to `.yaml` (#6503)

### What this PR does / why we need it?
This PR refactors the nightly single-node model test by migrating test
configurations from Python scripts to a more maintainable `YAML-based`
format.

| Original PR | Python (`.py`) | YAML (`.yaml`) |
| :--- | :--- | :--- |
| [#3568](#3568) |
`test_deepseek_r1_0528_w8a8_eplb.py` | `DeepSeek-R1-0528-W8A8.yaml` |
| [#3631](#3631) |
`test_deepseek_r1_0528_w8a8.py` | `DeepSeek-R1-0528-W8A8.yaml` |
| [#5874](#5874) |
`test_deepseek_r1_w8a8_hbm.py` | `DeepSeek-R1-W8A8-HBM.yaml` |
| [#3908](#3908) |
`test_deepseek_v3_2_w8a8.py` | `DeepSeek-V3.2-W8A8.yaml` |
| [#5682](#5682) |
`test_kimi_k2_thinking.py` | `Kimi-K2-Thinking.yaml` |
| [#4111](#4111) |
`test_mtpx_deepseek_r1_0528_w8a8.py` | `MTPX-DeepSeek-R1-0528-W8A8.yaml`
|
| [#3733](#3733) |
`test_prefix_cache_deepseek_r1_0528_w8a8.py` |
`Prefix-Cache-DeepSeek-R1-0528-W8A8.yaml` |
| [#6543](#6543) |
`test_qwen3_235b_w8a8.py` | `Qwen3-235B-A22B-W8A8.yaml` |
| [#6543](#6543) |
`test_qwen3_235b_a22b_w8a8_eplb.py` | `Qwen3-235B-A22B-W8A8.yaml` |
| [#3973](#3973) |
`test_qwen3_30b_w8a8.py` | `Qwen3-30B-A3B-W8A8.yaml` |
| [#3541](#3541) |
`test_qwen3_32b_int8.py` | `Qwen3-32B-Int8.yaml` |
| [#3757](#3757) |
`test_qwq_32b.py` | `QwQ-32B.yaml` |
| [#5616](#5616) |
`test_qwen3_next_w8a8.py` | `Qwen3-Next-80B-A3B-Instruct-W8A8.yaml` |
| [#3541](#3541) |
`test_qwen2_5_vl_7b.py` | `Qwen2.5-VL-7B-Instruct.yaml` |
| [#5301](#5301) |
`test_qwen2_5_vl_7b_epd.py` | `Qwen2.5-VL-7B-Instruct-EPD.yaml` |
| [#3707](#3707) |
`test_qwen2_5_vl_32b.py` | `Qwen2.5-VL-32B-Instruct.yaml` |
| [#3676](#3676) |
`test_qwen3_32b_int8_a3_feature_stack3.py` |
`Qwen3-32B-Int8-A3-Feature-Stack3.yaml` |
| [#3709](#3709) |
`test_prefix_cache_qwen3_32b_int8.py` |
`Prefix-Cache-Qwen3-32B-Int8.yaml` |
| [#5395](#5395) |
`test_qwen3_next.py` | `Qwen3-Next-80B-A3B-Instruct-A2.yaml` |
| [#3474](#3474) |
`test_qwen3_32b.py` | `Qwen3-32B.yaml` |
| [#3541](#3541) |
`test_qwen3_32b_int8.py` | `Qwen3-32B-Int8-A2.yaml` |
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.15.0
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.15.0

---------

Signed-off-by: MrZ20 <2609716663@qq.com>
LCAIZJ pushed a commit to LCAIZJ/vllm-ascend that referenced this pull request Mar 7, 2026
@MrZ20 @LCAIZJ
[Nightly][Refactor]Migrate nightly single-node model tests from .py…
093a573 
… to `.yaml` (vllm-project#6503)

### What this PR does / why we need it?
This PR refactors the nightly single-node model test by migrating test
configurations from Python scripts to a more maintainable `YAML-based`
format.

| Original PR | Python (`.py`) | YAML (`.yaml`) |
| :--- | :--- | :--- |
| [vllm-project#3568](vllm-project#3568) |
`test_deepseek_r1_0528_w8a8_eplb.py` | `DeepSeek-R1-0528-W8A8.yaml` |
| [vllm-project#3631](vllm-project#3631) |
`test_deepseek_r1_0528_w8a8.py` | `DeepSeek-R1-0528-W8A8.yaml` |
| [vllm-project#5874](vllm-project#5874) |
`test_deepseek_r1_w8a8_hbm.py` | `DeepSeek-R1-W8A8-HBM.yaml` |
| [vllm-project#3908](vllm-project#3908) |
`test_deepseek_v3_2_w8a8.py` | `DeepSeek-V3.2-W8A8.yaml` |
| [vllm-project#5682](vllm-project#5682) |
`test_kimi_k2_thinking.py` | `Kimi-K2-Thinking.yaml` |
| [vllm-project#4111](vllm-project#4111) |
`test_mtpx_deepseek_r1_0528_w8a8.py` | `MTPX-DeepSeek-R1-0528-W8A8.yaml`
|
| [vllm-project#3733](vllm-project#3733) |
`test_prefix_cache_deepseek_r1_0528_w8a8.py` |
`Prefix-Cache-DeepSeek-R1-0528-W8A8.yaml` |
| [vllm-project#6543](vllm-project#6543) |
`test_qwen3_235b_w8a8.py` | `Qwen3-235B-A22B-W8A8.yaml` |
| [vllm-project#6543](vllm-project#6543) |
`test_qwen3_235b_a22b_w8a8_eplb.py` | `Qwen3-235B-A22B-W8A8.yaml` |
| [vllm-project#3973](vllm-project#3973) |
`test_qwen3_30b_w8a8.py` | `Qwen3-30B-A3B-W8A8.yaml` |
| [vllm-project#3541](vllm-project#3541) |
`test_qwen3_32b_int8.py` | `Qwen3-32B-Int8.yaml` |
| [vllm-project#3757](vllm-project#3757) |
`test_qwq_32b.py` | `QwQ-32B.yaml` |
| [vllm-project#5616](vllm-project#5616) |
`test_qwen3_next_w8a8.py` | `Qwen3-Next-80B-A3B-Instruct-W8A8.yaml` |
| [vllm-project#3541](vllm-project#3541) |
`test_qwen2_5_vl_7b.py` | `Qwen2.5-VL-7B-Instruct.yaml` |
| [vllm-project#5301](vllm-project#5301) |
`test_qwen2_5_vl_7b_epd.py` | `Qwen2.5-VL-7B-Instruct-EPD.yaml` |
| [vllm-project#3707](vllm-project#3707) |
`test_qwen2_5_vl_32b.py` | `Qwen2.5-VL-32B-Instruct.yaml` |
| [vllm-project#3676](vllm-project#3676) |
`test_qwen3_32b_int8_a3_feature_stack3.py` |
`Qwen3-32B-Int8-A3-Feature-Stack3.yaml` |
| [vllm-project#3709](vllm-project#3709) |
`test_prefix_cache_qwen3_32b_int8.py` |
`Prefix-Cache-Qwen3-32B-Int8.yaml` |
| [vllm-project#5395](vllm-project#5395) |
`test_qwen3_next.py` | `Qwen3-Next-80B-A3B-Instruct-A2.yaml` |
| [vllm-project#3474](vllm-project#3474) |
`test_qwen3_32b.py` | `Qwen3-32B.yaml` |
| [vllm-project#3541](vllm-project#3541) |
`test_qwen3_32b_int8.py` | `Qwen3-32B-Int8-A2.yaml` |
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.15.0
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.15.0

---------

Signed-off-by: MrZ20 <2609716663@qq.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@Yikun Yikun Yikun left review comments

@wangxiyuan wangxiyuan wangxiyuan approved these changes

+1 more reviewer

@gemini-code-assist gemini-code-assist[bot] gemini-code-assist[bot] left review comments

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

dist-test documentation Improvements or additions to documentation module:tests ready read for review ready-for-test start test by label for PR

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

5 participants

@offline893 @wangxiyuan @Yikun @MengqingCao @offline0806