Environment and reproducibility

mlenv

Snapshot the ML stack to a file and diff two snapshots.

0 downloads/month on PyPI

Install

pip install mlenv-cli

If you use uv, also as an isolated CLI:

uv tool install mlenv-cli

What it does

The classic "it trained on my machine" usually comes down to something that moved: the CUDA build behind PyTorch, a minor Python bump, a driver. mlenv captures all of it in one JSON you can commit and compare.

In action

A real recorded session of the CLI running.

Read it as text

python
key             value
version         3.12.3
implementation  CPython
executable      /home/delcenjo/jmweb-org/mlenv/.venv/bin/python3
platform
key      value
system   Linux
release  6.17.0-35-generic
machine  x86_64
libc     glibc 2.39
packages
key             value
annotated-doc   0.0.4
coverage        7.14.3
iniconfig       2.3.0
markdown-it-py  4.2.0
mdurl           0.1.2
mlenv           0.2.0
mlenv-cli       0.2.0
packaging       26.2
pluggy          1.6.0
pygments        2.20.0
pytest          9.1.1
pytest-cov      7.1.0
rich            15.0.0
ruff            0.15.20
shellingham     1.5.4
typer           0.26.8

$ mlenv show
python
key             value
version         3.12.3
implementation  CPython

platform
key      value
system   Linux
release  6.17.0-35-generic
machine  x86_64
libc     glibc 2.39

Features

—Python, platform, packages, CUDA, cuDNN and GPUs.
—Diff with the dangerous changes ranked first.
—Optional GPU capture through NVML.
—CI gate and JSON output.

View the code on GitHub →

Other tools

dsdiff

A git-style diff between two datasets, with distribution drift.

splitcheck

Detect rows leaking between train, validation and test splits.

pii-sweep

Scan datasets for PII before sharing them.