[OpenAI] Support logprobs and top_logprobs in ChatCompletion #333
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Depends on apache/tvm#16650 for exposing |
|
Below are output of streaming: 
|
CharlieFRuan
added a commit
that referenced
this pull request
Mar 14, 2024
Changes in WebLLM: - Stateful chat completion: #330 - OpenAI's `logit_bias`: #331 - OpenAI's `logprobs` and `top_logprobs`: #333 Changes in TVMjs: - apache/tvm#16650 - Fix param download issues (already reflected in 0.2.26, but at the time this PR was not merged yet) - Expose `sampleTopPFromProb` to support `logprobs` (new in 0.2.27)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Support
logprobsandtop_logprobsfor OpenAI APIChatCompletion. Subsequently, these are added toGenerationConfigas well. See this link for reference.Implementation-wise, since we need the actual distribution, we use
tvmjs.applySoftmaxWithTemperature()andtvmjs.sampleTopPFromProb()instead oftvmjs.sampleTopPFromLogits().Performance-wise, despite all sampling being done on CPU, with
top_logprobs=2, we observe no degradation with Llama-2-7b-q4f32 on M3 Macbook. Note that withoutlogprobs, exact same behavior (hence performance) is maintained.We also support
CompletionUsagein this PR for token counting.