Back to the catalog

ACautomata/manifold · openwiki

Bundle OKF 0.1 · 7 conceitos · ACautomata/manifold

Open source Repository Open in the app JSON README (API)

About

# Files

- [Architecture and Source Map](architecture.md) - Component boundaries, data/config layers, evaluation/reporting boundaries, domain vocabulary, and where to look in source.
- [Callback registry and training spine](callback-registry.md) - CallbackRegistry two-phase resolve/build, the spec contract, and TrainingSpine as the single caller that composed the five training CLIs (ADR-0029 + ADR-0032).
- [Before/after GRPO evaluation](evaluation.md) - Shipped workflow and source map for manifold-eval, the same-noise before/after GRPO comparison, 3D paired-fidelity scoring, slice-grid output, and the self-contained comparison page builder. Also records the accepted but not yet implemented in-training paired-fidelity monitor.
- [Frozen arms and per-rank device policy](frozen-arm-and-device-policy.md) - FrozenArmMixin (register + dual-exclude off the optimizer / checkpoint) and DevicePolicy (the per-rank CUDA device decision that replaced resolve_warm_device and the pre-PG set_device tw

Details

Kind
OKF bundles
Topic
Maps, weather & travel
Publisher
acautomata
Origin
okf_github
Category
dados
Version
0.1
Last push
2026-09-08T13:16:13Z
Repository state
ativo
Language
Python
License
MIT
Added
2026-09-08 22:08:51
Updated
2026-09-08 22:08:51
Origin id
ACautomata/manifold:openwiki/index.md

README

# manifold

[![Python](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12%20%7C%203.13-blue)](https://www.python.org/)
[![License](https://img.shields.io/badge/license-MIT-green)](./LICENSE)

Pretraining and medical-imaging experiments built on top of
[stable-pretraining](https://github.com/galilai-group/stable-pretraining) and
[MONAI](https://monai.io/). Follows the
[diffusers](https://github.com/huggingface/diffusers) architecture: pipeline,
scheduler, and models.

## Setup

This project uses a `src/` layout and is managed with `pyproject.toml`.
Dependencies are installed into a conda environment named after the project.

```bash
# Create the conda environment
conda create -n manifold python=3.12 -y
conda activate manifold

# Install the package (editable) with all dependencies
pip install -e .
```

## Development

```bash
# Install with dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Lint
ruff check .
```

## Layout

```
manifold/
├── pyproject.toml
├── README.md
└── src/
    └── manifold/
        ├── __init__.py
        └── py.typed
```

## License

MIT — see [LICENSE](./LICENSE).

More