#544: Wheel Next + Packaging PEPs

Talk Python To Me

Michael Kennedy

Technology

4.8 • 642 Ratings

🗓️ 10 April 2026

⏱️ 71 minutes

🔗️ Recording | iTunes | RSS

🧾️ Download transcript

Summary

When you pip install a package with compiled code, the wheel you get is built for CPU features from 2009. Want newer optimizations like AVX2? Your installer has no way to ask for them. GPU support? You're on your own configuring special index URLs. The result is fat binaries, nearly gigabyte-sized wheels, and install pages that read like puzzle books. A coalition from NVIDIA, Astral, and QuanSight has been working on Wheel Next: A set of PEPs that let packages declare what hardware they need and let installers like uv pick the right build automatically. Just uv pip install torch and it works. I sit down with Jonathan Dekhtiar from NVIDIA, Ralf Gommers from Quansight and the NumPy and SciPy teams, and Charlie Marsh, founder of Astral and creator of uv, to dig into all of it.

Transcript

Click on a timestamp to play from that location

0:00.0	When you PIP install a package with compiled code, the wheel you get is built for CPU features from 2009.
0:06.5	Want newer optimizations like AVX2?
0:09.5	Your installer has no way to ask for them.
0:11.7	Want GPU support?
0:13.0	You're on your own configuring special index URLs.
0:15.6	The result is fat binaries, nearly gigabyte-sized wheels, and install pages that read like puzzle books.
0:22.6	A coalition from Nvidia, Astral, and Quantite has been working on Wheel Next, a set of
0:27.2	peps that let packages declare what hardware they need, and let installers like UV pick the right
0:32.9	build automatically. Just UV-P-P install torch, and it'll work. I see down with Jonathan DeCher
0:38.7	from Nvidia, Ralph Gomer's from Quantite and the NumPy and Cypine teams, and Charlie Marsh,
0:44.0	founder of Astral and creator of UV, to dig into it all. This is Talk Python to Me. Episode
0:48.7	544 recorded March 2, 2006. Talk Python to me, yeah, we ready to roll.
0:57.0	Upgrading the code.
0:58.0	No fear of getting old.
1:00.0	New frameworks in sight, geeky rap on deck.
1:03.0	Quark crew, it's time to unite.
1:05.0	We started in pyramid, cruising old school lanes, had that stable base, yes, sir.
1:10.0	Welcome to Talk Python, python to me the number one
1:11.8	python podcast for developers and data scientists this is your host michael kennedy i'm a pf fellow
1:17.9	who's been coding for over 25 years let's connect on social media you'll find me and talk python
1:23.3	on mastodon blue sky and x the social links are all in your show notes.
1:28.2	You can find over 10 years of past episodes at talk python.fm.
	...

Please login to see the full transcript.

Previous episode | Next episode

Disclaimer: The podcast and artwork embedded on this page are from Michael Kennedy, and are the property of its owner and not affiliated with or endorsed by Tapesearch.

Generated transcripts are the property of Michael Kennedy and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.