benchmax
Build, validate, and launch reinforcement learning training environments
All Skills in This Repository (5)
Pure Emerald Level Indicators
castform-aiview-progress
Monitor castform training runs and inspect status, reward scalars, and logs.
castform-aiverify-environment
Validate Castform training environments by interpreting scorecards and reward variance.
castform-aidesign-environment
Design castform BaseEnv training environments for reinforcement-learning workflows.
castform-ailaunch-run
Launch Castform training runs after validation, upload, and argument checks.
castform-aigenerate-data
Create train and eval JSONL datasets for Castform runs.
Frequently Asked Questions
FAQPage SchemaHow to install benchmax?▼
Run `npx skills add castform-ai/benchmax --all -g -y` in your terminal to install all skills in this suite globally.
What is benchmax used for?▼
It is a Python SDK for defining reinforcement learning environments with tools, rewards, and datasets as code, then validating and launching GPU training runs on the Castform platform.
How do I validate an environment before training?▼
Run `uv run python main.py validate` to execute a two-sibling local and hosted validation. Review both outcomes and never launch while any contract error remains.
Can I do supervised finetuning instead of RL?▼
Yes. If you already have labeled conversations or input-output pairs, the train-sft skill builds a validated dataset and launches a finetuning run with no environment or rewards needed.
Does benchmax prevent accidental GPU spending?▼
Yes. The launch workflow requires explicit human confirmation before any credit-spending training run, and validation must pass first.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core