NousResearch/atropos
1
0
Fork 0
mirror of https://github.com/NousResearch/atropos.git synced 2026-04-28 17:29:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

add this to our pyproject

Browse source
This commit is contained in:
Jai Suphavadeeprasit 2026-02-24 12:22:41 -05:00
parent ce85c7d95e
commit fb3228f669
2 changed files with 9 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

8
example_trainer/README.md
View file

@ -8,8 +8,8 @@ A modular training framework for fine-tuning language models with **Group Relati
```
example_trainer/
├── grpo.py # CLI entry point (dispatches to 4 training modes)
├── run.py # Unified launcher for shared_vllm mode (starts vLLM+trainer)
├── grpo.py # CLI entry point (also exposed as `atropos-grpo`)
├── run.py # Unified shared_vllm launcher (also exposed as `atropos-grpo-run`)
├── config.py # TrainingConfig Pydantic model (all hyperparameters)
├── cli.py # CLI argument parsing (modular, single source of truth)
├── api.py # Atropos API communication (registration, batch fetching)
@ -27,6 +27,10 @@ example_trainer/
└── math_zero_lora.yaml # Config for math_server_zero.py (lora mode)
```
After `pip install -e .` from the repository root, you can launch with either:
- `python -m example_trainer.grpo` or `atropos-grpo`
- `python -m example_trainer.run` or `atropos-grpo-run`
## GRPO Training Loop