Hydra

[Hydra @Hydra_Framework]

Hydra 1.1 is out!
- Any config can now have a Defaults List
- Recursive object instantiation
- Relative and nested interpolations
- Experimental Callback API

And much more!

Check out the blog post:
https://t.co/yL6WNZKiNW

[Hydra @Hydra_Framework]

Hydra 1.2 is released!
- Improved support for multirun, experiment management, and reproducing runs
- Improved instantiate API
- More typing support via OmegaConf 2.2
and more!
Please check out the post here to learn more: https://t.co/2LV5gETrC5

[Hydra @Hydra_Framework]

100 -> 200 -> 400 -> 800 -> 1600.

Every 4 months, like an exponential clock.

[Hydra @Hydra_Framework]

@cccntu It's just not designed for it. In a Notebook use the Compose API to create a config object.
Side effects includes changing working directory, configuring logging and more.

[Hydra @Hydra_Framework]

@cccntu If you need command line integration, use hydra.main(). However, you mentioned Jupyter notebook which does not exactly play nicely with command line arguments.
hydra.main() has side effects that are not appropriate to a Jupyter notebook like configuring the logging.

[Hydra @Hydra_Framework]

Awesome Hydra tutorial from @s_scardapane, check it out touching some of the new goodies in the upcoming Hydra 1.1.

[Simone Scardapane @s_scardapane]

*Learning Hydra for configuring ML experiments*

I wrote a lengthy tutorial on using @Hydra_Framework in #machinelearning experiments.

I go from basic configuration to advanced use-cases, including instantiating a model directly and validating the schema. 👇

1/3

[Hydra @Hydra_Framework]

Hydra 1.1 first release candidate is out!

- OmegaConf 2.1 (relative and nested interpolation)
- Every config can now have a Defaults List.
- Recursive object instantiation

Check out the release notes:
https://t.co/tk6rKdgrJ3

[Hydra @Hydra_Framework]

Hydra reached 800 dependent repositories, continuing the trend of doubling every 4 months.

[Hydra @Hydra_Framework]

@MorganFunto of @huggingface just released Tune, a Hydra powered Transformers performance & evaluation framework.
Check out out the blog post for some hard-core details.

https://t.co/5fjHgzFSvk

[Hydra @Hydra_Framework]

Lightning-transformers is an awesome new library from the folks at @gridai_ combining Hydra, @huggingface's transformers and @PyTorchLightnin to form a super flexible transformers framework.

[Lightning AI ⚡️ @LightningAI]

We love #HuggingFace Transformers. Wonder how you can train them with the advanced Lightning Trainer options?

Lightning Transformers offers ready-to-go #Hydra configs for tasks and datasets, no boilerplate required, powered by Lightning.⚡

Learn More: https://t.co/o731O9occJ

[Hydra @Hydra_Framework]

Check out the blog post by @tousifsays: He is using Hydra to compose a config for Vivado- a hardware design tool - instead of having to manually draw in the GUI or write XML configs.
https://t.co/cPM037c97I

Exciting to see Hydra used in the Hardware design space.

[Hydra @Hydra_Framework]

1 Million downloads: check.
https://t.co/VVdwKTi2Z8.*

[Hydra @Hydra_Framework]

Join us on the PyTorch Ecosystem Day on April 21, 2021 and learn more about how Hydra helps simplify your code and workflow!
In addition to poster sessions, we will also host a breakout room on 9:50AM- 10:30AM PST!
Register: https://t.co/MSYqxRPIHH
#PTED21

[Hydra @Hydra_Framework]

A great blog post by @samscub, showcasing how he is deploying his research code with Hydra, Ray and Docker.

[Samuel Stanton @samuel_stanton_]

I wrote a blog post about deploying research code to the cloud without having to write boilerplate or learn some new framework. It's based on @Hydra_Framework and the corresponding @raydistributed plugin, and I love it!

https://t.co/4ZEqKOxPdO

[ray @raydistributed]

You can configure and Scale ML with @Hydra_Framework and Ray on AWS or local Ray clusters.

Blog Post: https://t.co/LwYb7SOKXh

[Hydra @Hydra_Framework]

ML Best practices 101.

[Nicholas Vadivelu @nicvadivelu]

Excited to present "Establishing a Productive DS & ML Workflow" this Wednesday from 6-7:30pm (EST)! Come if you're interested in learning more about best practices for experiment configuration (@Hydra_Framework), productivity tools, and more!

Syllabus: https://t.co/JrNPF7CbMg.

[Hydra @Hydra_Framework]

We are adding a Ray (@raydistributed) Launcher to Hydra's Launcher family! With the new Launcher, you can launch your applications to an AWS or local Ray cluster. Please check out the blog post: https://t.co/rHKonYLm9B

[Hydra @Hydra_Framework]

@tryolabs @tiangolo I want to call out that Hydra also enables leveraging Python type hints to drive configuration and command line arguments.

Here is a minimal example (It's recommended to read the tutorial from the start though):

https://t.co/ttDf5MAA6H

[Hydra @Hydra_Framework]

Hydra and OmegaConf featured in the top 10 Python libraries of 2020 by Tryolabs!
"A must for doing configuration management in the 21st century!"

[Tryolabs @tryolabs]

Our yearly list of favorite #Python libraries is out! 🚀

This time we also added some honorable mentions to our main picks. Did we miss any major one? Give us a shout.

Some highlights on the thread. 👇
https://t.co/fQtoGqIlrm