reverie-bio 0.1.0

Fine-tune foundation models for biology and medicine

Reverie

A training and fine-tuning tool for bio foundation models.

Installation

pip install reverie-bio[all]

For Standard Model Bio support, also install smb-utils:

pip install git+https://github.com/standardmodelbio/smb-utils.git

For development:

git clone https://github.com/zaaisvanzyl/reverie.git
cd reverie
pip install -e ".[all,dev]"

Quick Start

Option 1: Web UI (Recommended)

A browser-based interface for uploading data, configuring training, and viewing results.

# Launch the UI
reverie ui

This starts a local server at http://127.0.0.1:7860. You can customize host and port:

reverie ui --host 0.0.0.0 --port 8000

Option 2: CLI

reverie train \
    --model standard-model \
    --data ./data/synthetic_data.parquet \
    --labels ./data/synthetic_labels.csv \
    --label-column label \
    --output ./results

Foundation Models

Standard Model Bio

Column	Type	Description
subject_id	string/int	Patient identifier
time	datetime	Event timestamp
code	string	Clinical code (ICD-10, RxNorm, LOINC, etc.)

Example:

subject_id,time,code
patient_001,2023-01-15,ICD10:E11.9
patient_001,2023-01-15,RxNorm:860975
patient_002,2023-03-10,ICD10:I10

Labels

One row per patient:

subject_id,label
patient_001,0
patient_002,1

Reference: standardmodel.bio

Development

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

# Run tests
pytest

# Run tests with coverage
pytest --cov=reverie