Hardware Acceleration of LLMs: A comprehensive survey and comparison

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by koolala on Friday September 06, 2024 @11:13PM (#41471242)

I'd love to watch a LLM run in WebGL where everything is Textures. Would be neat to visually see the difference in architectures.

[ Parent | on HN ]

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by vanviegen on Saturday September 07, 2024 @01:47AM (#41471729)

Wouldn't that be just like watching static noise?

[ Parent | on HN ]

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by archerx on Saturday September 07, 2024 @04:42AM (#41472501)

I think some patterns would appear.

[ Parent | on HN ]

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by Twirrim on Saturday September 07, 2024 @10:50AM (#41474221)

Why would you expect patterns to appear?

[ Parent | on HN ]

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by koolala on Saturday September 07, 2024 @09:16PM (#41477520)

Everything is a different pattern. Noise isn't random - we just call 1 very specific unique pattern of it 'random'.

[ Parent | on HN ]

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by archerx on Sunday September 08, 2024 @05:57AM (#41479383)

I was really into Convolutional Neural Networks back in the day and when you visualize the layers you’d get interesting patterns. I’m guessing the same would happen with LLMs but the patterns will be a lot more abstract but still some what recognizable.

Here’s an example of visualizing CNN layers; https://www.researchgate.net/figure/Visualization-of-example...

[ Parent | on HN ]

Re:Hardware Acceleration of LLMs: A comprehensive survey and... by iml7 on Saturday September 07, 2024 @08:23AM (#41473516)

Doesn't Google have a tool that allows you to check the activation status of the matrix? Gemma scope

[ Parent | on HN ]