Hacker News Reader

Comment by samstave | Hacker News Reader

ELI5 what this means?

I am losing my bibliography, etymology and vocabulary with every single AI advancement article.

Where learn AI vocab, please?

I nee an FN AI teacher to just give me daily updates on AI and verbiage, models, etc...

Hey AI - if you're so smart, build a podcast that teaches me about yourself and how to be a better meat parent whom made you.///\\\\

View on HN

Starting with a trained PyTorch model, it builds optimized C++ binaries for running inference (not training) on NVidia and AMD GPUs. Various optimizations mentioned a lot, so presumably models run faster than with just running them via regular PyTorch.

I'm curious why that is called "rendering" rather than "compiling". Is the code boiler plate and just a change in the NN's representation?

How much faster is it?

Depends on the model and GPU. Here is an example of almost 2x on a 3060 for StableDiffusion: https://www.youtube.com/watch?v=_6BsUijOWoM

Very much not an expert here but what I understand is that most deep learning frameworks (PyTorch, Tensorflow, etc.) have some overhead associated with them just being on the graphics card. This takes PyTorch code and removes the overhead by translating the network into a "native" language for the card (CUDA for NVIDIA).

What I'm not sure is what "HIP" is in this context.

The way I'm reading this is it's the difference between running code in an interpreter vs. on the bare metal (for the GPU)

HIP is AMD's open-source re-implementation of CUDA libraries.

HIP is AMD's contribution to the open source community to overcome Nvidia's CUDA software moat

You write in HIP C++ language and run them on either NVIDIA or AMD platforms. This way you get cross-platform code and are not stuck with Nvidia.

Use HIPify tool to automatically convert existing sources from CUDA to HIP.

It's been around for many years - but the fact that so many people still don't know about it - speaks for the sad state of AMD communication.

https://docs.amd.com/bundle/HIP-Programming-Guide-v5.3/page/...

HIP is a pathetic CUDA API clone. Gratuitous renames don't do it any good and are more representative of NIH rather than anything else (sadly).

They should have shipped a proper header set instead of hipify.

It doesn't really help understand what they are, but for completeness CUDA is an acronym for "Compute Unified Device Architecture" while HIP is "Heterogeneous-compute Interface for Portability"

What is an FN AI?

An autonomous Belgian firearm.

Soon. :(

A 'fUCKIn' ai'

Reply on news.ycombinator.com