Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR adds the necessary changes (that I found, at least) to run the adapter with vllm 0.5.3.
The main change was the removal of
TextTokensPrompt
, it was removed with this PR: vllm-project/vllm@739b61aI'm not 100% sure if simply replacing with
LLMInputs
is the correct fix here, but it does seem to work. (I think I see some more complex processing happening on the upstream openai serving engine)The initializers for all of the
OpenAIServing*
classes changed, I copied over the new defs from upstream.Also I noticed that there was no initialization for the tokenizer, so I added it.
How Has This Been Tested?
Booting up a pod with a mig gpu slice, installing vllm@5.3 and vllm-tgis-adapter@llm-inputs, and running the default facebook/opt-125m model with
python3 -m vllm_tgis_adapter
.Requests sent using the swagger page at
/docs
and grpcui (though grpc should be unaffected)Merge criteria: