Our take
Inactiveno commits in over a year- Activity
- 19
- Community
- 53
- Issues
- 21
- Pull requests
- 20
+0 stars in the last 30 days
EasyInstruct offers a clearly modular toolkit for instruction-data research, but its inactive maintenance status makes it a less certain choice for new production workflows.
Good fit if
- You are experimenting with instruction generation or selection methods and want existing Python components for those steps.
- Your workflow uses supported LLM APIs or locally deployed models and you can validate the framework against your own setup.
- You value an MIT-licensed project with multiple contributors and are prepared to maintain a fork if needed.
Look elsewhere if
- You need a project with recent commits, releases, or demonstrated ongoing issue handling. The repository has had no commits in over a year and no releases in the past year.
- You require CI or an included test suite, neither of which is reported for this repository.
- Your project needs a maintained, general-purpose interface for model serving rather than an instruction-processing toolkit aimed at research experiments.
All health signals
| Last commit | 2024-12-23 (1.8 years ago) |
|---|---|
| Commits, last 90 days | 0 |
| Releases, last 12 months | 0 |
| Contributors | 14 (top contributor: 54% of commits) |
| Issues closed, last 90 days | 0 |
| Pull requests merged, last 90 days | 0 |
| Project age | 3 years |
Checked on 2026-10-26 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
407Forks
36Change since Nov 17, 2026: 0 stars·0 forks
Overview
EasyInstruct is a Python package for instruction processing with large language models such as GPT-4, LLaMA, and ChatGLM. It targets researchers running LLM experiments who need a reusable way to generate instruction data, select examples, and construct prompts rather than implementing each step separately.
The framework organizes these steps into generators, selectors, and prompting methods, and also includes engines for executing prompts on locally deployed models. Its documentation and examples cover workflows such as Self-Instruct and instruction quality filtering.
Key Features
- Generate instruction datasets with methods including Self-Instruct, Evol-Instruct, Backtranslation, and KG2Instruct.
- Filter and select examples using length, perplexity, ROUGE, lexical diversity, GPT scoring, and code-oriented CIRS metrics.
- Remove duplicate instruction samples with a deduplication selector.
- Combine multiple selection methods with
MultiSelector. - Use prompt methods and engines for instruction execution, with details in the linked prompt and engine guides.
- Start workflows through a configuration-driven shell script or a local Gradio app.
- Integrate API-based models including OpenAI, Anthropic, and Cohere offerings described in the README.
Use Cases
- LLM researchers can generate synthetic instruction-following examples from seed tasks for experiments.
- Dataset curators can filter raw or generated examples for length, quality, diversity, or duplication before training.
- Researchers studying code reasoning can use the code selector and its CIRS-based filtering workflow.
- Teams prototyping instruction-processing workflows can use the Gradio app or YAML-style configurations to try the framework before adapting its Python classes.
What you need
Detected in the repository
- Python (from setup.py)
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install the current main branch:
pip install git+https://github.com/zjunlp/EasyInstruct@main
See the README and documentation for configuration and usage details.
Considerations
The measured project status is inactive: there have been no commits in over a year and no releases in the last year. The repository lists 14 contributors, and contribution activity is not marked as concentrated around one person, but recent maintenance and issue-resolution activity is absent. No CI configuration or tests are reported, so adopters should verify behavior and dependencies in their own environment. The README describes API-backed generation examples that require credentials for the relevant provider, and points to separate documentation for prompting and engines; it does not specify a general hardware requirement.
Found this useful?
Share it with someone who would like EasyInstruct.