Etude

Launch training runs on your own or rented machines, and track what they record.

# job.py
from etude.launch.records import Job


def runs() -> Job:
    return Job(command="uv run python train.py", max_hours=2)

# train.py
from etude.track.records import Step
from etude.track.run import init

run = init(experiment="capacity-pilot", clocks=[Step("updates", "optimizer updates")])
for update in range(1, 1001):
    loss = train_one_update()
    run.log({"loss": loss}, updates=update)
    if update % 500 == 0:
        run.checkpoint(save_model(), updates=update)
run.finish()

etude launch job.py          # run it on this computer, an SSH host or a rented GPU
etude status                 # where runs stand
etude logs --run RUN          # preparation, every attempt, cleanup
etude reports write --open   # charts of every run, as one HTML file per experiment

Everything a run records goes to your own directory or S3 bucket. There is no server to run and no account to make. A run that loses its machine resumes from its last checkpoint on another, and reads back as one run.

Install

Etude needs Python 3.12 and uv.

uv tool install etude-sdk          # the `etude` command
uv add etude-sdk                   # the library, in a training project
uv add 'etude-sdk[s3]'             # with S3 storage

Read more

• Tracking: runs, clocks, checkpoints, evaluations, reports.
• The launcher: pools, SSH hosts, rented machines, cohorts, limits and recovery.

Develop

uv sync
uv run pytest
uv run mypy
uv run ruff check . && uv run ruff format --check .

AGENTS.md holds the conventions, for people and coding agents.

Licence

MIT. See LICENSE.