[CI] Add new runner and enable QwQ multinpu test #417
Conversation
Potabk
force-pushed
the
ci
branch
2 times, most recently
from
e8aa80c to
dd0a9c2
Compare
Potabk
force-pushed
the
ci
branch
4 times, most recently
from
1599f82 to
ebdad10
Compare
Yikun
changed the title
| tensor_parallel_size=4, | ||
| distributed_executor_backend=distributed_executor_backend, | ||
| ) as vllm_model: | ||
| vllm_model.generate_greedy(example_prompts, max_tokens) |
There was a problem hiding this comment.
I remember last offline discussion, we'd also want to add transformer as compare, does it works (stable works)？
There was a problem hiding this comment.
will finish at next pull request
| fi | ||
| pip install /root/.cache/pta/torch_npu-2.5.1.dev20250320-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl | ||
| - name: Run vllm-project/vllm-ascend test |
There was a problem hiding this comment.
V1 engine should be tested as well.
There was a problem hiding this comment.
I think we shouldn't test v1 like this. v1 doesn't fully support in some features. in addition we should mock the VLLM_USE_V1 in the related test modules if in need
There was a problem hiding this comment.
why？ We only test V1 for vllm-ascend test, not vllm test.
There was a problem hiding this comment.
I mean we will construct pytest fixture to decide which cases need to be tested on v1
Potabk
force-pushed
the
ci
branch
4 times, most recently
from
7b0213b to
59578d7
Compare
| env: | ||
| VLLM_USE_V1: 1 | ||
| VLLM_WORKER_MULTIPROC_METHOD: spawn | ||
| VLLM_USE_V1: 0 |
There was a problem hiding this comment.
add it here @pytest.mark.parametrize("use_v1", ["1", "0"])
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Potabk
force-pushed
the
ci
branch
2 times, most recently
from
5605150 to
83df77b
Compare
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
| @@ -1,5 +1,8 @@ | |||
| [pytest] | |||
| minversion = 6.0 | |||
| markers = | |||
There was a problem hiding this comment.
this markers is useless now, right?
There was a problem hiding this comment.
this markers help pytest detect what kind of cases should be run, just like pytest -m 'multinpu'
|
I'll fix the nit in the follow-up PR |
### 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? - 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>
3 participants
What this PR does / why we need it?
Does this PR introduce any user-facing change?
No
How was this patch tested?
CI passed