Utilities for making hydra scale to ML workflows
Project description
hydra-zen helps you configure your project using the power of Hydra, while enjoying the Zen of Python!
hydra-zen eliminates the boilerplate code that you write to configure, orchestrate, and organize the results of large-scale projects, such as machine learning experiments. It does so by providing Hydra-compatible tools that dynamically generate “structured configurations” of your code, and enables Python-centric workflows for running configured instances of your code.
hydra-zen offers:
Functions for automatically and dynamically generating structured configs that can be used to fully or partially instantiate objects in your application.
The ability to launch Hydra jobs, complete with parameter sweeps and multi-run configurations, from within a notebook or any other Python environment.
Incisive type annotations that provide enriched context about your project’s configurations to IDEs, type checkers, and other tooling.
Runtime validation of configurations to catch mistakes before your application launches.
Equal support for both object-oriented libraries (e.g., torch.nn) and functional ones (e.g., jax and numpy).
These functions and capabilities can be used to great effect alongside PyTorch Lightning to design boilerplate-free machine learning projects!
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for hydra_zen-0.1.0rc4-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 682a7b507639de5ffc2090ac291de49e709040f92b44285820b23a7158f362c9 |
|
MD5 | ea284ffcf104fb6c98f86dc110538279 |
|
BLAKE2b-256 | b79e103de529195639fe33fcb2812548cb2e65d0abf79847721008fc8f07aa22 |