zilliztech/GPTCache

[Feature]: Find a more suitable similarity evaluation model

Open

#549 opened on Oct 9, 2023

View on GitHub
 (1 comment) (0 reactions) (0 assignees)Python (407 forks)batch import
good first issuehacktoberfest

Repository metrics

Stars
 (5,857 stars)
PR merge metrics
 (No merged PRs in 30d)

Description

Is your feature request related to a problem? Please describe.

Currently, GPTCache achieves an accuracy of approximately 80% in optimal conditions. However, during daily usage, it often returns unsatisfactory cache results, indicating that it does not perform well in practical production scenarios.

GPTCache utilizes user inquiries to retrieve similar data from the vector database and calculates the similarity between the user inquiries and the questions in the retrieved data. However, the effectiveness of this approach is not ideal. We attempted to improve the results by employing the cross-encoder/quora-distilroberta-base model to match answers in the retrieved data with user inquiries. Unfortunately, the outcomes were not satisfactory.

Is there another model available that better suits this particular scenario?

Describe the solution you'd like.

No response

Describe an alternate solution.

No response

Anything else? (Additional Context)

No response

Contributor guide

Research direction
Evaluate alternative similarity models such as sentence transformers (e.g., all MiniLM L6 v2) or cross encoders. Test them on the specific use case to find a model that balances accuracy and performance.
Tech stack
python
Domain
backendai
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Fresh
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
Python
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50