archiki / grips Goto Github PK
View Code? Open in Web Editor NEWCode for our paper: "GrIPS: Gradient-free, Edit-based Instruction Search for Prompting Large Language Models"
License: MIT License
Code for our paper: "GrIPS: Gradient-free, Edit-based Instruction Search for Prompting Large Language Models"
License: MIT License
Hi Archiki Prasad,
I want to achieve the result after the Instruction + Examples run_search.py of the InstructGPT curie of table 6, which is 59.44 .
But there is only "Instruction + Positive Examples" mode in function custom_instruction_prompt, no "Instruction + Examples" mode. Thus, I failed to achieve the best results. I don't know if there is a problem with my understanding of the code, or the existing code does not achieve this effect.
I think I need your help to achieve this result, or do you have any suggestions, thanks!
Hi Archiki Prasad,
I want to reproduce the results of GrIPS initializing with task-agnostic instructions with Instruction-Only prompts in your paper.
With the hyperparameters in the paper which is l = 1, m = 5, n = 10, and P = 2, I have achieved the accuracy of model InstructGPT babbage in table 3, which is 54.41. However, I cannot achieve the accuracy of model GPT-2 XL in table 3, which is 54.29. My result is about 53.00.
I have tried several train seeds and different hyperparameters, but still failed to achieve the accuracy 54.29.
I need your help to reproduce the result. Could you give me some suggestions.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.