10 comments

[ 2.6 ms ] story [ 8.2 ms ] thread
Amazing!

Training the model, although it is small with only 15M parameters, takes quite some compute power. I wonder if the resources would have been sufficient at the time. Also, I wonder if disk space would have been sufficient to store the input corpus.

It looks like these machines run IRIX which is a unix variant on a mips processor and has some parallelization features. I wonder if the compiler he's using is taking advantage of these, or if there's potentially some speedup possible. From what I remember, llama2.c uses OMP to optionally parallelize some of the matrix-vector products which wouldn't be available on that machine.

https://en.m.wikipedia.org/wiki/IRIX

It's interesting to think about some of the earliest machines we could have run a pretty great LLM on. A Cray XMP-EA from 1986 could address up to 2GW, which would have been 16GB. The XMP or earlier Cray-1 lines could also use multiple solid-state SRAM disks that were up to 1GB in size and could stream at ~1GB/sec.
It's interesting to think about how computers might be different had SRAM generally developed and become prolific instead of DRAM.
What do you mean? It did continue to develop (right up to the latest 3nm process!), it’s just inherently lower density due to its structure.
It is not typically used outside of L1 cache. I imagine a computer using bistable memory cells as being more resilient, and a more elegant design than memory that needs circuitry to constantly refresh it.
answering the question in the xweet: researcher 28 years ago would have decent reproducible peer reviewed metrics so they would not even look at most of this.
Goes to show we have always had the "tech" just not the idea.
I wonder how much RAM is needed to run this – for example, would this work on the Indigo2's minimal config of 16 MB? (If so, I'll be running towards THINK Pascal to get this running on an SE/30...)
The prettiest machines ever. Also, had 3D glasses. We had one, they were old 20 years ago, but so cool.