@rasbt - AI Dynamics - Page 59 of 95

Python ecosystem shifting to Rust over C++ for performance

By

–

25 August 2023 22h01

Or more like a competitor to Rust since the Python ecosystem is moving to that instead of C++ (eg see polars and some others)? And for GPU & AI stuff a competitor to Triton/CUDA?

→ View original post on X — @rasbt

25 August 2023

Clarification between inference and training processes

By

@rasbt

–

25 August 2023 21h53

Thanks for tagging me, but that’s only inference not training right?

→ View original post on X — @rasbt

25 August 2023

Potential Competitor to Triton for Python Users

By

@rasbt

–

25 August 2023 16h52

Yeah, we’ll see. It’s probably not going to be used by the average Python users, but I can see it becoming a competitor for Triton though.

→ View original post on X — @rasbt

25 August 2023

GPT-4 Behavior Changes Over Time Pose Transparency Risks

By

@rasbt

–

25 August 2023 13h29

Yes, like I showed at the bottom of the screenshot, it was basically debunked. However, the point still is that the behavior of GPT-4 changes over time, and it's quite intransparent. Now imagine you are a company who built a service on top of the GPT-4 API

→ View original post on X — @rasbt

25 August 2023

Simplifying PyTorch by removing C++ dependencies

By

@rasbt

–

25 August 2023 4h16

What I am really hoping for is that it will help simplify code bases, though. Maybe we can get rid of the C++ parts of PyTorch this way to make development easier.

→ View original post on X — @rasbt

25 August 2023

PyTorch Python overhead and realistic performance gains discussion

By

@rasbt

–

25 August 2023 4h14

2/2
PyTorch's Python overhead is usually quoted to be at ~10% (compared to PyTorch's C++ API). I think for PyTorch users (or CUDA-dependent packages), the speed advantage is probably more going to be in the single digit range, not 35,000x.
Still super exciting though!

→ View original post on X — @rasbt

25 August 2023

Python Superset Language Promise Tempered by GPU Support Concerns

By

@rasbt

–

25 August 2023 4h14

I think it is awesome! Especially the fact that it's going to be a superset of Python is really highly appreciated. But when it comes to AI workflows, I think we'll still have to see what the speed advantage is since they currently don't support GPUs and CUDA afaik.
1/2

→ View original post on X — @rasbt

25 August 2023