Contains on #1111 for completeness.
<!-- Thanks for sending a pull request!
BEFORE SUBMITTING, PLEASE READ
https://docs.vllm.ai/en/latest/contributing/overview.html
-->
### What this PR does / why we need it?
Implement multi-stream parallelism for MoE layers with shared experts,
where computation of shared experts will be overlapped with expert token
dispatch and combine. Also, when multi-stream is enabled, weights of
shared experts will be force to replicate across all cards, regardless
of any tensor parallelism configurations, to avoid AllReduce operations.
With the expected overlaping being:
```
| shared gate_up | shared act | | shared down |
| dispatch | routed gate_up, act, down | combine |
```
<!--
- Please clarify what changes you are proposing. The purpose of this
section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster
reviews in your PR.
- Please clarify why the changes are needed. For instance, the use case
and bug description.
- Fixes #
-->
### Does this PR introduce _any_ user-facing change?
No.
<!--
Note that it means *any* user-facing change including all aspects such
as API, interface or other behavior changes.
Documentation-only updates are not considered user-facing changes.
-->
### How was this patch tested?
Tested on 1x16 910 node, with tailored 2 layer DSKv2.
<!--
CI passed with new added/existing test.
If it was tested in a way different from regular unit tests, please
clarify how you tested step by step, ideally copy and paste-able, so
that other reviewers can test and check, and descendants can verify in
the future.
If tests were not added, please describe why they were not added and/or
why it was difficult to add.
-->
---------
Signed-off-by: sdmyzlp <lrwei2@petalmail.com>
Make sure the lint test passed before start the e2e test to save compute
resource.
Updated the patch doc to make sure the CI works as expect.
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
- Set default values to fix spec decode
- To avoid oom, we need to run the test in a single process
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- CI passed, espcecially multicards CI
- For spec decode test, long term CI passed
Closes: https://github.com/vllm-project/vllm-ascend/pull/1105
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
Co-authored-by: Yizhou Liu <liu_yizhou@outlook.com>
Co-authored-by: mengwei805 <mengwei25@huawei.com>
### What this PR does / why we need it?
- Adds support for passing prompt_embeds to LLM.generate as
```bash
llm.generate({"prompt_embeds": input_embeds}, sampling_params)
```
or
```bash
llm.generate(
[{"prompt_embeds": input_embeds} for input_embeds in inputs_embeds], sampling_params
)
```
- Add `prompt_embeds` to examples
### How was this patch tested?
CI passed with new added/existing test.
and I have test with the example script in this pr, and the output seems
looks good:
```bash
[Single Inference Output]
------------------------------
The capital of France is Paris. Paris is the largest city in France and is
------------------------------
Adding requests: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 3966.87it/s]
Processed prompts: 100%|█████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 3.99it/s, est. speed input: 177.08 toks/s, output: 63.91 toks/s]
[Batch Inference Outputs]
------------------------------
Q1: Please tell me about the capital of France.
A1: The capital of France is Paris. It is located in the northern part of the
Q2: When is the day longest during the year?
A2: The day is longest during the year at the summer solstice. This typically occurs
Q3: Where is bigger, the moon or the sun?
A3: The sun is significantly bigger than the moon.
The sun has a diameter of
------------------------------
```
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
More and more config options are added to additional_config. This PR
provide a new AscendConfig to manage these config options by an easier
way to make code cleaner and readable.
This PR also added the `additional_config` doc for users.
Added the test_ascend_config.py to make sure the new AscendConfig works
as expect.
TODO: Add e2e test with torchair and deepseek once the CI resource is
available.
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
Skip benchmarks/** in vllm ascend test to reduce CI cost
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
1. remove some useless test func and file
2. fix format.sh problem
3. enable full test for singlecard and multicard
4. move long term test to long_term folder. For this kind of test, it
only runs by labeled and daily test. Include: spec decode、accuracy test
## After refactor:
There are 4 test modules
- `singlecard`: contains the test running on one NPU. It'll be run for
each PR and daily test.
- `multicard`: contains the test running on multi NPUs. It'll be run for
each PR and daily test.
- `long_term`: contains the test that cost much time(Now include `spec
decode` and `accuracy` test). It'll be run for the PR with
`long-term-test` labeled and daily test.
- `e2e`: contains the test for doc and pd feature. It'll be run for the
PR with `pd-test` labeled and daily test.
## Todo:
1. some test are skipped, they should be fixed and reenabled in the
future.
2. pyhccl test for multicard doesn't work at all. It should be enabled
as well.
3. ensure long-term-test pass by daily test.
### Know issue
Now, `ready` labels is required to start pd test or long term test. And
when `long-term-test` or `pd-test` is labeled after another one, the old
labeled test will be re-run again. So the labeled test should be ran in
the following step:
1. decide which test need run, then label it. `long-term-test` or
`pd-test` or both.
2. add `ready-for-test` label, then the test will be ran.
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
<!-- Thanks for sending a pull request!
BEFORE SUBMITTING, PLEASE READ
https://docs.vllm.ai/en/latest/contributing/overview.html
-->
### What this PR does / why we need it?
<!--
- Please clarify what changes you are proposing. The purpose of this
section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster
reviews in your PR.
- Please clarify why the changes are needed. For instance, the use case
and bug description.
- Fixes #
-->
Make spec decode support for V1 Engine
- Currently, Ascend does not support the triton kernel. PyTorch is used
to rewrite the `rejection_sampler.py` triton kernel. However, PyTorch is
not as good as Triton. Therefore, ascend c is used to implement the
function in the future.
- Currently, spec decode supports only the ngram algorithm. The eagle
algorithm needs to be further adapted.
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such
as API, interface or other behavior changes.
Documentation-only updates are not considered user-facing changes.
-->
Not change user facing.
### How was this patch tested?
<!--
CI passed with new added/existing test.
If it was tested in a way different from regular unit tests, please
clarify how you tested step by step, ideally copy and paste-able, so
that other reviewers can test and check, and descendants can verify in
the future.
If tests were not added, please describe why they were not added and/or
why it was difficult to add.
-->
test by `tests/singlecard/spec_decode/e2e/test_v1_spec_decode.py` and
`tests/sample/test_rejection_sampler.py`, test base function of
rejection sampler and e2e function of spec decode.
Signed-off-by: ponix-j <657511300@qq.com>
### What this PR does / why we need it?
Add V1Engine LoRA support.
Add LoRA e2e test on single card and multiple cards.
### Does this PR introduce _any_ user-facing change?
support lora for V1
### How was this patch tested?
CI passed with new added test
---------
Signed-off-by: jesse <szxfml@gmail.com>
Signed-off-by: paulyu <paulyu0307@gmail.com>
Signed-off-by: paulyu12 <507435917@qq.com>
Co-authored-by: jesse <szxfml@gmail.com>
Co-authored-by: paulyu <paulyu0307@gmail.com>
### What this PR does / why we need it?
#### 1. fix spec ut in vllm-ascend main and vllm main
As https://github.com/vllm-project/vllm-ascend/pull/694 and
https://github.com/vllm-project/vllm-ascend/pull/749 verify, Now,
vllm-ascend main and vllm 0.8.5, spec UT is happy, but vllm-ascend main
and vllm main, CI is fail.
I found the reason is a triton bug
https://github.com/triton-lang/triton/issues/2266, but i I didn't figure
it out that why the bug did not effect vllm-ascend main and vllm 0.8.5,
maybe the usage of triton have changed when vllm 0.8.5 to latest main
As the bug describe, I changed the minimum block_size in UT from 8 to
16, and the modification is verified locally to be effective.
#### 2. modify some case skip form.
I modified some commented out cases to skipif form, which is more
standardized.
### Does this PR introduce _any_ user-facing change?
None
### How was this patch tested?
CI
Signed-off-by: mengwei805 <mengwei25@huawei.com>
### What this PR does / why we need it?
This PR aims to fix nightly ci
[broken](https://github.com/vllm-project/vllm-ascend/actions/runs/14848150987)
We have a workflow containing multiple triggers:
- push events (to the default branch)
- pull requests (against the default branch)
- scheduled events
Our paths-filter action works great for the first two use-cases,
detecting the context and base to compare against. However, it fails for
scheduled events giving the error `This action requires 'base' input to
be configured or 'repository.default_branch' to be set in the event
payload.`
For the scheduling trigger event, we choose to skip this filter
because we don't need its results:
```
- name: Check for changes in Speculative Decode
if: github.event_name != 'schedule'
```
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
Bump vllm version to v0.8.5.post1
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
### What this PR does / why we need it?
Make CANN version bump separately from
https://github.com/vllm-project/vllm-ascend/pull/708
- Upgrade CANN version to 8.1.rc1
- Add prefix to speed up download
`m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10`
- Address tail sapce for Dockerfile.openEuler
- Add note for `/workspace` and `/vllm-workspace` as followup of
https://github.com/vllm-project/vllm-ascend/pull/741
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
CI passed
Co-authored-by: MengqingCao <cmq0113@163.com>
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
Co-authored-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
Re-enable Speculative Decode test for vLLM v0.8.5
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
### What this PR does / why we need it?
- This PR proposes a P2P version of Disaggregated Prefill based on
llm_datadist which manages data transfer.
- This solution reconstructs previous offline single-node Disaggregated
Prefill solution, and supports multi-node and online serveing now.
- Currently this solution supports 1P1D situation of Deepseek hybrid
parallelism (P: TP+EP, D: DP+EP). Note that xPyD situation is considered
in the solution design, and will be supported soon within v1 engine.
---------
Signed-off-by: hw_whx <wanghexiang7@huawei.com>
Signed-off-by: ganyi <pleaplusone.gy@gmail.com>
Co-authored-by: hw_whx <wanghexiang7@huawei.com>
Co-authored-by: ganyi <pleaplusone.gy@gmail.com>
### What this PR does / why we need it?
Add nightly CI for basic function and model usability
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
<!-- Thanks for sending a pull request!
BEFORE SUBMITTING, PLEASE READ
https://docs.vllm.ai/en/latest/contributing/overview.html
-->
### What this PR does / why we need it?
As custom deepseek modeling do some changes to support graph mode in
https://github.com/vllm-project/vllm-ascend/pull/585, so i follow it to
change custom deepseek_mtp modeling.
And some modifications for k>1 were not carried over by the
https://github.com/vllm-project/vllm-ascend/pull/429, now i add it.
In order to better take care of the MTP feature in the vllm-ascend
repository, I added cases related to graph mode(torchair), but i skip it
since torchair can not correctly clean up memory in vllmrunner.
Also i add some case for MTP quantization weights, but test weight is
not ready, so i skip it and i will open it when test quant weights is
ready.
https://github.com/vllm-project/vllm-ascend/pull/648 did not completely
fix the sample
change(https://github.com/vllm-project/vllm-ascend/issues/660) issue, I
added the relevant changes.
### Does this PR introduce _any_ user-facing change?
now, u can use following method to use mtp in deepseek v3/r1 float or
quant weights with eager mode.
```python
llm = LLM(
model="wemaster/deepseek_mtp_main_random_bf16",
tensor_parallel_size=2,
speculative_config={
"num_speculative_tokens": 1,
},
enforce_eager=True,
trust_remote_code=True,
disable_log_stats=False,
gpu_memory_utilization=0.8,
max_model_len=64,
)
```
or use mtp in deepseek v3/r1 float or quant weights with graph
mode（torchair）
```python
llm = LLM(
model="wemaster/deepseek_mtp_main_random_bf16",
tensor_parallel_size=2,
speculative_config={
"num_speculative_tokens": 1,
},
trust_remote_code=True,
additional_config={
'enable_graph_mode': True,
},
disable_log_stats=False,
gpu_memory_utilization=0.8,
max_model_len=64,
)
```
add notes:
1. now, we support k>1, so u can set num_speculative_tokens > 1 if there
is sufficient redundant computing power;
2. MTP is not supported in V1, we will support it when vLLM does it in
https://github.com/vllm-project/vllm/issues/13500.
3. if u run MTP failed by `segmentation fault`, u can follow v0.7.3
patch https://github.com/vllm-project/vllm-ascend/pull/236 file
`vllm_ascend/patch/patch_metrics.py` method
`__npu_async_metrics_collector_init__`
### How was this patch tested?
local tested passed and test by CI
Signed-off-by: mengwei805 <mengwei25@huawei.com>
### What this PR does / why we need it?
The torch-npu 2.5.1 are published:
https://pypi.org/project/torch-npu/2.5.1/
It's time to remove all torch-npu dev version from vllm-ascend code base
### Does this PR introduce _any_ user-facing change?
Yes, using torch-npu 2.5.1
### How was this patch tested?
- [ ] CI passed
- [ ] Manually test
- [ ] Grep all `dev2025`
---------
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
<!-- Thanks for sending a pull request!
BEFORE SUBMITTING, PLEASE READ
https://docs.vllm.ai/en/latest/contributing/overview.html
-->
### What this PR does / why we need it?
<!--
- Please clarify what changes you are proposing. The purpose of this
section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster
reviews in your PR.
- Please clarify why the changes are needed. For instance, the use case
and bug description.
- Fixes #
-->
This PR supports the access of vllm-acend to the piecewise_graph feature
provided by the v1 engine.
1. register unifiled_ascend_attention_with_output for piecewise_graph to
split graph.
2. support NPUGraph to accelerate kernel launch.
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such
as API, interface or other behavior changes.
Documentation-only updates are not considered user-facing changes.
-->
support npugraph to default， Users can disenable the npugraph feature by
configuring enforce_eager.
This has corresponding requirements for the versions of torch_npu and
CANN, and they need to support graph capture.
### How was this patch tested?
<!--
CI passed with new added/existing test.
If it was tested in a way different from regular unit tests, please
clarify how you tested step by step, ideally copy and paste-able, so
that other reviewers can test and check, and descendants can verify in
the future.
If tests were not added, please describe why they were not added and/or
why it was difficult to add.
-->
it turn to default
---------
Signed-off-by: Bug Hunter Yan <yanpq@zju.edu.cn>
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
Co-authored-by: Yizhou Liu <liu_yizhou@outlook.com>
### What this PR does / why we need it?
The pr will fix some bug about spec decode / MTP
The pr add a mtp e2e UT `test_mtp_correctness.py`
**vllm_ascend/attention/attention.py**
1. add support `self.attn_mask_cache` only has 1 element to cover scene
in which both spec docode and chunked prefill are enabled.
**vllm_ascend/distributed/parallel_state.py**
1. remove 2 assert because spec decode worker would use init_worker
twice
**vllm_ascend/models/deepseek_mtp.py**
1. remove unused params;
2. add support w8a8 in `CustomDeepSeekMTP`
**vllm_ascend/quantization/quant_config.py**
1. use `AscendUnquantizedFusedMoEMethod` instead of
`UnquantizedFusedMoEMethod`
**other**
1. replace `from vllm.logger import init_logger` to `from vllm.logger
import logger` all of the vllm-ascend project
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
Signed-off-by: mengwei805 <mengwei25@huawei.com>
1. Add `vllm_version_is` function to check vllm version.
2. `ensure_kv_transfer_initialized` and `get_kv_transfer_group ` have
been moved to other place in vllm main branch via
3408e47159
, this patch fix the import error.
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
Backport: https://github.com/vllm-project/vllm-ascend/pull/252
This support speculative decoding in Ascend, including speculating with
a draft model、by matching n-grams in the prompt、using MLP speculators
and using EAGLE based draft models.
Backport: https://github.com/vllm-project/vllm-ascend/pull/423
spec decode MultiStepWorker support TP1DraftModelRunner fully, support
run the draft_model_runner with multi-step prepare on the NPU directly
and support draft_model_runner use MLA.
1. before this pr, `MultiStepWorker` would not step into the branch
using NPU prepare, but only into the branch using CPU prepare (`line 52`
of `vllm_ascend/patch/patch_multi_step_worker.py`). Although this has
`no effect` on the `correct operation` of speculative decoding and the
performance of the two branches is basically the same as of the current
version, I support entering this branch in this PR. In general, there
are two main changes in `patch_multi_step_worker.py`: first, the
`is_cuda_like()` check is removed and the `TP1DraftModelRunner`
rewritten in vllm_ascend is used; second, the
`supports_gpu_multi_step()` function is made to return true on NPU
devices when outer Multi_step_worker could work correct.
3. before this pr, `TP1DraftModelRunner` only supports Attention on NPU,
but not MLA. The relevant adaptation is in
`vllm_ascend/worker/draft_model_runner.py`. Although I don’t know why
the `input_positions` of `model_input.attn_metadata` in vllm-ascend
needs to be added in `execute_model`, it is done in `model_runner.py`,
so I also made corresponding changes. Otherwise, when atten_backend is
MLA, it will prompt that input_positions cannot be found.
4. I commented out two lines in `draft_model_runner.py` in `line118` to
support the scenario of K>1.
```
# lora_mapping=model_input.lora_mapping,
# lora_requests=model_input.lora_requests,
```
I added comments. In the future, when vllm-ascend supports lora feature,
the changes here can be restored.
TODO：
- [ ] revert the patch when the related issues are addressed in vllm
### How was this patch tested?
CI passed with new added test.
- e2e test for medusa proposer:
tests/singlecard/spec_decode/e2e/test_medusa_correctness.py
- e2e test for mlp proposer:
tests/singlecard/spec_decode/e2e/test_mlp_correctness.py
- e2e test for n-gram proposer:
tests/singlecard/spec_decode/e2e/test_ngram_correctness.py
Tests for patched files:
- tests/singlecard/spec_decode/test_dynamic_spec_decode.py
- tests/singlecard/spec_decode/test_multi_step_worker.py
- tests/singlecard/spec_decode/test_ngram_worker.py
- tests/singlecard/spec_decode/test_spec_decode_worker.py
---------
Signed-off-by: MengqingCao <cmq0113@163.com>
Co-authored-by: mengwei805 <mengwei25@huawei.com>
### What this PR does / why we need it?
- add `HF_TOKEN` as global var to the runner
- add `HF_ENDPOINT` as global var to the runner
- change concurrency group, rely on current pr num
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
This PR enable custom ops build by default.
### Does this PR introduce _any_ user-facing change?
Yes, users now install vllm-ascend from source will trigger custom ops
build step.
### How was this patch tested?
By image build and e2e CI
---------
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
- Add a new runner to the continuous integration system and keep the
original CI runner until the new runner runs stably
- Add distributed test cases
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
This pr upgrades torch-npu to 0320, so that #321,
https://github.com/vllm-project/vllm-ascend/issues/267#issuecomment-2745045743
could be fixed, and #372 should be reverted after this pr
### Does this PR introduce _any_ user-facing change?
upgrade torch-npu to 0320
### How was this patch tested?
tested locally with long seq inferencing.
---------
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
Bump torch_npu version to dev20250308.3 to fix performance regression on
multi-stream case:
e04c580d07
.
### Does this PR introduce _any_ user-facing change?
NO
### How was this patch tested?
CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
Update torch-npu version to fix torch npu exponential_ accuracy
With this update, the percision issue when setting `temperature > 0` is
fixed.
---------
Signed-off-by: Mengqing Cao <cmq0113@163.com>
### What this PR does / why we need it?
Add dispatch job to leverage jobs to dynamic devices include 2 stage as
below:
The dispatch job will spend extra about `10s * parallel number + 30s`
time to wait other job launch container and release lock.
- **Stage 1: Acquire lock**
add a dispatch job, this job use lockfile to acquire locks and then get
device number dynamically
- **Stage 2.1: Launch container with dynamic device**
pass the device number via output and start the container job with
dynamic device
- **Stage 2.2: Release lock**
once the job started, release the lock.
In the backend, we use multiple path to setup multiple self host runners
as load balancer:
```
$ pwd
/home/action
$ ll | grep actions
drwx------ 6 action action 4096 Mar 7 08:55 actions-runner-01
drwx------ 6 action action 4096 Mar 7 08:55 actions-runner-02
drwx------ 6 action action 4096 Mar 7 08:55 actions-runner-03
drwx------ 6 action action 4096 Mar 7 08:56 actions-runner-04
drwx------ 4 action action 4096 Jan 24 22:08 actions-runner-05
drwx------ 4 action action 4096 Jan 24 22:08 actions-runner-06
```
```
adduser -G docker action
su action
pip3 install docker prettytable
sudo yum install procmail
```
### Does this PR introduce _any_ user-facing change?
NO
### How was this patch tested?
- CI passed
- E2E test manully, triggered 3 jobs in parallel:
- [1st
job](https://github.com/vllm-project/vllm-ascend/actions/runs/13711345757/job/38348309297)
dispatch to /dev/davinci2.
- [2nd
job](https://github.com/vllm-project/vllm-ascend/actions/runs/13711348739/job/38348316250)
dispatch to /dev/davinci3
- [3rd
job](https://github.com/vllm-project/vllm-ascend/actions/runs/13711351493/job/38348324551)
dispatch to /dev/davinci4
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
### What this PR does / why we need it?
Recover vllm-ascend dev image
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
### What this PR does / why we need it?
Add `HF_TOKEN` for downloading models that requires access rights from
huggingface hub. This will fix the CI error in #123 and #76
Signed-off-by: MengqingCao <cmq0113@163.com>
Enable CI on all branch.
Installing with the torch-npu-2.5.1.dev20250218 so that we could enable
CI on all branch and prepare for merging 0.7.1-dev to main
---------
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
Switch to cann latest version
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
