Practical overview
When it fits
Use ATM when you have a classification task represented by a labeled CSV and want an automated model-search workflow through its Python API.
Who it is for
- Python users who want to automate classifier selection and tuning for tabular classification data.
- Teams needing a shared record of datasets, runs and generated models across workers.
Documented uses
- Search for candidate classifiers and tune their parameters against a labeled dataset.
- Review run summaries, selected-model details and scores for tested classifiers.
- Export a selected classifier and load it to predict on new data.
Requirements & limits
- The documented input format is a comma-separated CSV with a header, a target column and feature columns.
- Datasets can be provided from local storage, HTTP or S3 locations.
- The documentation lists Python 2.7, 3.5 and 3.6 as development and test versions and recommends a virtual environment.
- ModelHub is documented as supporting collaboration among workers on multiple machines and retaining job state across failures or interruptions.
- ATM's documentation labels the project pre-alpha.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install ATM in a Python virtual environment
Create an isolated environment and install ATM using pip.
Before you start
- Python 3.6, which the documentation uses in its virtual-environment setup example.
- A system with pip.
Steps
Install virtualenv with pip.
Create an environment using Python 3.6.
Activate the environment.
Install ATM with pip.
Expected outcome
ATM is installed in the activated virtual environment.
Guide limitations
- The documentation also lists Python 2.7 and 3.5 as development and test versions, but the setup commands shown use Python 3.6.
- The documentation labels ATM pre-alpha.
Search for a classifier and inspect its results
Run ATM on a labeled CSV and examine the best classifier and comparative scores.
Before you start
- ATM installed in a Python environment.
- A comma-separated CSV with a header, a target column, feature columns and complete training samples.
- A local CSV path, HTTP URL or S3 URI.
Steps
Create an ATM instance in Python.
Pass the dataset location to atm.run.
Use the returned results object to view the run summary and best classifier.
Retrieve the scores table to compare tested classifiers.
Expected outcome
A completed datarun with a summary, best-classifier details and scores for tested classifiers.
Guide limitations
- Private S3 access requires configured AWS credentials or credentials passed when creating the ATM instance.
- The documentation labels ATM pre-alpha.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 7 November 2026
- Original website title
- ATM - Auto Tune Models — ATM 0.2.3-dev documentation · Official homepage title · retained in its source language