Get automatic type-safety via builds()’s signature inspection. Write less, stop repeating yourself, and get more out of your configs. This means that it is much easier and safer to write and maintain the configs for your Hydra applications: ZenFooConf = builds ( foo, bar = 2, baz =, populate_full_signature = True ) Now we can configure and run train_fn from the CLI exposed by train.py:įrom hydra_zen import builds def foo ( bar : int, baz : list, qux : float = 1.23 ). # Hydra records the exact, reproducible config # for each run, and saves the results in an # auto-generated, configurable output dir # hydra-zen then instantiates these configs # - creating the Model & DataLoader instances - # and passes them to train_fn, running the training code. hydra_main ( config_name = "train_fn", config_path = None, version_base = "1.3", ) # Hydra will accept configuration options from # the CLI and merge them with the stored configs. add_to_hydra_store () # Generate the CLI For train_fn zen ( train_fn ). summary = f "Model: \n -, ], ) if _name_ = "_main_" : from hydra_zen import zen store. class Model : def _init_ ( self, activation, nlayers, logits = False ) -> None : self.
#Zen of python code#
# Contents of baby_torch.py # Note: no Hydra/hydra-zen specific code here! def relu ( x ). Suppose you have the following library code. Support for using NumPy, Jax, PyTorch, and Lightning (a.k.a PyTorch-Lightning) in your It is also great for designing your data scienceĪnd machine learning research to be reproducible. Hydra-zen is fully compatible with Hydra, and is appropriate for use in both rapid Hydra-specific boilerplate from your project. Task-function wrapper, which help to eliminate most of the It also provides a custom config-store API and So by providing functions that dynamically andĪutomatically generate dataclass-based configs for your code. Hydra-zen eliminates all hand-written yaml configs from your Hydra project. Scalable: launch multiple runs of your software, be it on your local machine or across multiple nodes on a cluster.
#Zen of python full#
Repeatable: each run of your code will be self-documenting the full configuration of your software is saved alongside your results. Use hydra-zen to design your project to be:Ĭonfigurable: Change deeply-nested parameters and swap out entire pieces of your program, all from the command line. Hydra-zen is a Python library that makes the Hydra framework
🎓 Using hydra-zen for your research project? Cite us! Welcome to hydra-zen’s documentation! #
0 kommentar(er)
