OPAL
OPAL is a PyTorch optimizer wrapper that lets a coding agent rewrite training settings while a model trains, for developers training neural networks.
What it is good for
- Adjusting learning rates and optimizer settings during a long training run
- Reducing the need to hand-tune an optimizer before training starts
- Comparing optimizer results against a plain PyTorch run
At a glance
- Who it is for
- Setup
- Requires programming
Level 5 of 5. You need to write code to use it.
- Cost
- Free and open source
- Type
- Framework or SDK
- Works with
- PyTorch, Claude Code, Codex, Cursor, Hugging Face Trainer
Momentum
- Rank this week
- Not ranked yet
- Best rank
- None yet
- GitHub stars
- 257
GitHub numbers as of Nov 2, 2026. Updated weekly.
On Hacker News
- Reactive Optimizers: rewrite the optimizer's source code during training 32 points, 3 comments, Oct 21, 2026
Official links
Setup steps are on the official pages. We do not repeat them here, since they change.
Questions
- Is OPAL free?
- Yes. It is free and open source (MIT license).
- Do I need to know how to code to use OPAL?
- Yes. You need to write code to use it.
- What does OPAL work with?
- According to its own documentation, it works with PyTorch, Claude Code, Codex, Cursor, and Hugging Face Trainer.
- What is OPAL good for?
- Adjusting learning rates and optimizer settings during a long training run, reducing the need to hand-tune an optimizer before training starts, and comparing optimizer results against a plain PyTorch run.
Similar agents
- OpenEvolve
OpenEvolve is an open-source evolutionary coding agent that uses LLMs to find and refine better code for a given problem for developers and researchers.
- mathema
mathema is a Python tool that checks functions against stated mathematical claims and records the results for developers who want evidence their code is correct.
- REA
REA is an MCP server that lets AI agents reverse engineer apps and native binaries to explain how a feature works.
- Firecrawl
Firecrawl is an API that searches the web and pulls clean page content for AI agents and apps.
- browser-use
Browser Use is an open-source Python library and cloud service that lets AI agents control a web browser to finish tasks for you.
- TradingAgents
TradingAgents is a multi-agent framework that simulates a trading firm of LLM analysts, researchers, and managers to evaluate stock trades for researchers and developers.