[V1][Frontend] Add Testing For V1 Runtime Parameters #14159
Conversation
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: [email protected] <[email protected]>
robertgshaw2-redhat
requested review from
WoosukKwon,
njhill,
ywang96,
comaniac and
alexm-redhat
as code owners
|
👋 Hi! Thank you for contributing to the vLLM project. 💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging. To run CI, PR reviewers can either: Add 🚀 |
|
|
||
| # Reject all allowed token ids | ||
| with pytest.raises(ValueError): | ||
| _ = model.generate(PROMPT, SamplingParams(allowed_token_ids=[10])) |
There was a problem hiding this comment.
Do you mean an empty list here? or do you mean a full list?
There was a problem hiding this comment.
once your PR lands, I will revert this
There was a problem hiding this comment.
Oh, I see the point, you just want to disable this feature completely, and if there is anything, just raise ValueError.
| def test_allowed_token_ids(model): | ||
| """Check that we can use allowed_token_ids.""" | ||
|
|
||
| # This does not currently work due to incorrect implementation. |
There was a problem hiding this comment.
should be fixed by #14169.
robertgshaw2-redhat
added
the
ready
vllm/v1/engine/processor.py
Outdated
| raise ValueError("VLLM V1 does not yet support bad_words.") | ||
| # Logits processors not supported. | ||
| if params.logits_processors: | ||
| raise ValueError("VLLM V1 does not yet support per request " |
There was a problem hiding this comment.
perhaps remove the "yet" from this one?
| if priority != 0: | ||
| raise ValueError("V1 does not support priority yet.") | ||
| if trace_headers is not None: | ||
| raise ValueError("V1 does not support tracing yet.") | ||
|
|
There was a problem hiding this comment.
I think we need this here too?
if prompt_adapter_request is not None:
raise ValueError("V1 does not support prompt_adapter_request.")Signed-off-by: [email protected] <[email protected]>
|
This pull request has merge conflicts that must be resolved before it can be |
Signed-off-by: [email protected] <[email protected]>
The check is introduced by vllm-project#14159 and it will fail v1 benchmark with `VLLM V1 does not yet support best_of` error as the parameter is not yet supported Signed-off-by: Huy Do <[email protected]>
This was missed when merging vllm-project#14169 and vllm-project#14159 Signed-off-by: Nick Hill <[email protected]>
) Signed-off-by: [email protected] <[email protected]> Signed-off-by: Johnny <[email protected]>
SUMMARY:
SamplingParamswe can encounter at runtimeValueErrorfailures if any occur (note: this happens in the per request asyncio task, so just a single request fails when this happens and we handle the error on the client side)NOTES:
allowed_token_idssampling is incorrect.