Hi everyone, I want to show you pytask, a tool for managing workflows in your research or data analysis project! #datascience #econtwitter #python #reproducibility https://t.co/nURATMVLK4 🧵 1/16

observation since moving in with gf washing clothes seems like an area where people heavily optimize for their personal local minima. If you bring two people together and they try to mix clothes and approaches, nothing works anymore

I realized GPT-4o knows about pytask but only proposes code that follows removed syntax. I guess I can be happy that I have a little userbase and I am not flooded with issues.

Old: Deprecation cycles are as long as necessary for efficient communication and user adjustment. New: Deprecation cycles should be as long as needed for LLM APIs to incorporate the latest syntax in their training checkpoints, ensuring they don't suggest outdated code.

✨ Bonus: Project Templates with pixi cookiecutter-pytask-project v1.7.0 has been released. It now uses @prefix_dev's pixi, which offers a fantastic developer experience. 🙏Thanks to everyone who contributed via code, feedback, and discussions! https://t.co/XaYtS0DInB

🤖 Task Generators Following up on the previous example, we can use task generators to process each downloaded file individually, copy it to another location, and add the .txt extension. https://t.co/ho4uCJfAHw

🤖 Provisional Nodes Provisional nodes can be used when you do not know how many products are created by a task. For example, downloading all files without file extension from a GitHub repository. https://t.co/0BdYRlDqXD

Caveats when working with remote backends. ❌ Limited cross-platform support. Windows users should use the WSL2. ⚠ Local files. pytask automatically syncs local files to remote workers. Store files in S3, etc., to avoid time-consuming uploads. https://t.co/7rIriwoQdA

You can also run every task in a cloud VM using the coiled.Cluster. https://t.co/OlS9RJtxrk

Using pytask-parallel with coiled, you can run single tasks in a cloud VM of your choice with the hardware you need! And it scales automatically with the load! https://t.co/e6Gu5VVqQk

✨pytask v0.5 release announcement✨ This release comes with two big features: 🤖 Provisional Nodes and Task Generators 🚀 Support for HPC and your favorite cloud provider (AWS, GCP, Azure) using dask / @CoiledHQ ✨Bonus at the end Find out more ⬇ https://t.co/7dSRYEDXsm

I understand creating this one button at the end feels like a waste of time after so many iterations and checks. I would argue you should build this button right from the start to be more confident in your code, to find errors quicker. Maybe use my own https://t.co/ayTn2mZgVo 😄

👉pytask.mark.parametrize is deprecated and will be removed in v0.4 (coming in ~1-2 months). Switch to the new approach! https://t.co/AkCMoNWUQ7 👉pytask is even lazier (or more careful) and uses hashes to detect changes in task modules. Thank you to all contributors!

One week late, but ... pytask v0.3.2 has been released 🥳. https://t.co/ZLvyNpeeV7 There are two important changes👇

I wrote a blogpost about dags, a package I created with @tobraab. Dags provides tools to combine several interrelated Python functions into one function. If you want to know what that means and why it is useful, check out the post:

I know it is an example of emotional data viz and I know that I am the joke here 🙃

My climbing book casually introduced me to time travel.

It’s not my profile of course. I cannot manage my time.

Cherish your weekends!

Ever ran an optimization with @scipy but none of the algorithms worked? To use other algorithms, you had to learn a new library and rewrite your code? Say hi to #estimagic, a modern #python package that wraps all algorithms you’ll ever need 🚀 Of course, there’s more: A 🧵

Wrong column.