Remove the abstraction for token counting from the main evaluation API #6320
Conversation
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
github-actions
bot
added
the
area-ai-eval
shyamnamboodiripad
force-pushed
the
counter
branch
2 times, most recently
from
7511ebe to
f93c418
Compare
peterwald
reviewed
Apr 19, 2025
src/Libraries/Microsoft.Extensions.AI.Evaluation/Microsoft.Extensions.AI.Evaluation.csproj
Outdated
Show resolved
Hide resolved
peterwald
reviewed
Apr 19, 2025
src/Libraries/Microsoft.Extensions.AI.Evaluation/ChatMessageExtensions.cs
Outdated
Show resolved
Hide resolved
peterwald
reviewed
Apr 19, 2025
test/Libraries/Microsoft.Extensions.AI.Evaluation.Tests/ConversationTrimmingTests.cs
Outdated
Show resolved
Hide resolved
shyamnamboodiripad
force-pushed
the
counter
branch
2 times, most recently
from
bd3f9b0 to
987b057
Compare
This change is being made because there is still some uncertainty around what a general purpose token counting abstraction (that supports all kinds of future models, and all kinds of input modalities) should look like at the moment. We do not want to bake in an API that only supports text based inputs for the models and use cases that are prevalent today, since it would be a potential breaking change to change this API after we release a stable version of the evaluation APIs. We can always reintroduce the token counting support in a non-breaking fashion in the future if and when there is more clarity on what a general purpose token counting abstraction should look like, or if and when a general purpose token counting abstraction is introduced in a lower layer (Microsoft.Extensions.AI) in the future. In the meanwhile, callers should still be able to use the Microsoft.ML.Tokenizers library directly to count tokens in text-based content and trim down the conversation history before calling EvaluateAsync() if needed.
shyamnamboodiripad
force-pushed
the
counter
branch
from
987b057 to
5f9bc5e
Compare
peterwald
approved these changes
Apr 23, 2025
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
This change is being made because there is still some uncertainty around what a general-purpose token counting abstraction (that supports all kinds of future models, and all kinds of input modalities) should look like at the moment. We do not want to bake in an API that only supports text-based inputs for the models and use cases that are prevalent today, since it would be a potential breaking change to change this API after we release a stable version of the evaluation APIs.
We can always reintroduce the token counting support in a non-breaking fashion in the future if and when there is more clarity on what a general-purpose token counting abstraction should look like, or if and when a general-purpose token counting abstraction is introduced in a lower layer (Microsoft.Extensions.AI) in the future.
In the meanwhile, callers should still be able to use the
Microsoft.ML.Tokenizerslibrary directly to count tokens in text-based content and trim down the conversation history before callingEvaluateAsync()if needed.Fixes #6234
Microsoft Reviewers: Open in CodeFlow