MLSTAC
Machine learning model sharing, built on the STAC MLM extension and Safetensors.
Overview
MLSTAC gives you a single, consistent way to publish and consume machine learning models. Each model is described by a STAC Item using the MLM extension, and its weights travel as Safetensors. From that metadata you can load a model from many backends (HTTP, local disk, S3, Google Cloud Storage), download it, inspect it, and turn it into a usable PyTorch model.
Status: experimental. The API may change between minor versions.
Installation
pip install mlstac
Quickstart
import mlstac
# Load only the metadata first (no weights are downloaded yet)
model = mlstac.load("https://example.com/my-model/mlm.json")
# Inspect what you have
print(model.get_model_summary())
model.print_schema()
# Download every file into a local folder; you get back a loader
# pointing at the local copy
local = model.download("./my-model")
# Build a usable model from the local files
net = local.compiled_model()
Loading from different sources
mlstac.load is the main entry point. It accepts a URL, a local path, or a list of
.pt2 files for an ad-hoc ensemble.
# Remote metadata
mlstac.load("https://example.com/model/mlm.json")
# A local directory that contains an mlm.json
mlstac.load("./my-model")
# An ensemble built directly from .pt2 files
ensemble = mlstac.load(["model_a.pt2", "model_b.pt2"])
ensemble.is_ensemble # True
Downloading
You can download from the loader instance, or call the module-level function directly.
# From an existing loader (reuses the source it was created from)
local = model.download("./my-model")
# Or as a one-shot call
local = mlstac.download("https://example.com/model/mlm.json", "./my-model")
Supported backends
| Scheme | Example |
|---|---|
http / https / ftp |
https://example.com/model/mlm.json |
local |
./my-model or /abs/path/mlm.json |
s3 |
s3://bucket/key/mlm.json (needs boto3) |
gs |
gs://bucket/key/mlm.json (needs google-cloud-storage) |
Documentation
Full docs live at tacofoundation.github.io/mlstac.
Development
# install with dev dependencies
poetry install --with dev
# run the test suite
poetry run pytest tests/ -v
License
See LICENSE.
Metadata
Release files for mlstac 0.4.9
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| mlstac-0.4.9.tar.gz | 14.2 kB | Without provenance | Details for mlstac-0.4.9.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| mlstac-0.4.9-py3-none-any.whl | Python 3 | none | any | Details for mlstac-0.4.9-py3-none-any.whl |
Total release size: 29.7 kB
Release files / mlstac-0.4.9.tar.gz
| Download URL | mlstac-0.4.9.tar.gz |
|---|---|
| Size | 14.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
4f1fc327e703c63429d03846ed4cdd365708aec89aecfa0c24085e745b40e593
|
|
BLAKE2b-256 checksum How to use checksums |
e97a6a27ea57039ed38187579b0c0ad676a4d8aae823eafb5e1381d5af9f152d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
poetry/2.4.1 CPython/3.12.13 Windows/11
|
Release files / mlstac-0.4.9-py3-none-any.whl
| Download URL | mlstac-0.4.9-py3-none-any.whl |
|---|---|
| Size | 15.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
5758d4e1de03734aba540d5200ede73facf4286b838198b06dfe5495a4e4463d
|
|
BLAKE2b-256 checksum How to use checksums |
78d82246bb7178f78d9d4e23d1b6134fa999a16c314bd0b9ffc1a2331d8f84d6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
poetry/2.4.1 CPython/3.12.13 Windows/11
|
Security for mlstac 0.4.9
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.