fix(autogen-ext): correct claude-3-haiku context window in OpenAI-compatible model table - #8199
Open
Shxiao (Shxiao101) wants to merge 1 commit into
Open
Conversation
…patible model table claude-3-haiku-20240307 has a 200k token context window (matching the Anthropic table in this repo and Anthropic's docs); the 50000 entry made the OpenAI-compatible path over-count usage and truncate early.
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
claude-3-haiku-20240307was listed with a 50,000-token context window in the OpenAI-compatible model info table (autogen_ext.models.openai._model_info). This PR corrects it to 200,000.Why
The same repository's Anthropic table (
autogen_ext.models.anthropic._model_info) and Anthropic's docs both list claude-3-haiku's context window as 200k tokens. When claude-3-haiku is used throughOpenAIChatCompletionClient(e.g. via an OpenAI-compatible endpoint),remaining_tokens()— and thereforeTokenLimitedChatCompletionContext— under-counts the available budget by 150k tokens and silently drops conversation history far earlier than necessary.How I checked
get_token_limit("claude-3-haiku-20240307")returns 200000 after the change (50000 before)pytest python/packages/autogen-ext/tests/models/test_openai_model_client.py -k "model_info or token"— 6 passedruff check/ruff format --checkon the changed file — cleanOne-line change, no API surface change; fits the maintenance-mode policy (bug fixes only).