News for hackers. Stuff that matters.
Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by Havoc on Thursday February 22, 2024 @08:18PM (#39475690)
> 307MB of Enron emails

Wait what?

Is that some sort of inside joke?

Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by _aavaa_ on Thursday February 22, 2024 @08:32PM (#39475783)
Nope!

See for example: https://www.cs.cmu.edu/~./enron/

Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by Havoc on Friday February 23, 2024 @10:32AM (#39481809)
> This data is valuable; to my knowledge it is the only substantial collection of "real" email that is public.

Interesting. Something good came out of Enron after all

Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by kuprel on Thursday February 22, 2024 @04:56PM (#39473865)
This adds PyTorch/CUDA training support to Andrej Karpathy's minbpe. It takes 2min 28sec (148 seconds) on an RTX4090 to train the BasicTokenizer with a vocab_size of 512 on 307MB of Enron emails. The original code takes 2hrs 15min (8076 seconds) on an M2 Air with Python 3.11 to do this. That is a 55x speedup.
Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by threesevenths on Friday February 23, 2024 @10:23AM (#39481658)
Am I reading this right? A 55x improvement while also going from an M2 Air to an RTX 4090?

If so this doesn’t seem like a logical comparison and the 55x claim would likely not translate when using the same hardware.

Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by kuprel on Friday February 23, 2024 @12:06PM (#39483206)
The M2 Air was actually much faster than whatever CPU was on the cloud RTX4090 machine I rented. I chose the stronger benchmark to compare to
Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by kuprel on Friday February 23, 2024 @03:16PM (#39485562)
Using int16 and an H100 the speedup is actually 108x over the M2 air
Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by lostmsu on Saturday February 24, 2024 @01:22AM (#39489569)
Why is it surprising? CPU-only M2 probably has under 1 teraops while RTX 4090 has 77. M2's GPU was not used, but even it only provides around 4 teraops, so would have been ~20x slower than 4090.
Re:55x Speedup of Andrej Karpathy's Minbpe LLM Tokenizer wit... by erichocean on Friday February 23, 2024 @08:41AM (#39480439)
Now someone needs to do a Mojo version, and write up the blog post.