Live data from Hacker News

Fault Tolerant Llama training

pytorch.org

1–10 of 16 posts

Re: Fault Tolerant Llama training

#4
post #3

300 L40s? What's this, 1998?

I was curious about this so I had o3 do a bit of research. Turns out 300 L40s have more compute than any supercomputer before 2013 (and arguably before 2016, depending on how you count reduced-precision FLOPs).

https://chatgpt.com/share/685dea79-26ec-8002-bd62-7ed83aedf4...

Re: Fault Tolerant Llama training

#5
post #3

300 L40s? What's this, 1998?

Hey Tim, how's it going?

Interested in lending PyTorch some compute? :)

torchft can handle much larger scales but for public multi-day demonstration run this is what we had available. Point of this blog was to demonstrate correctness of the quorum algorithm and recovery with a stock PyTorch stack and not so much peak flops.

Stay tuned though -- planning on doing some much larger demos on B200s!

Re: Fault Tolerant Llama training

#7
post #6

This is severely underrated work, why aren't there more mid sized companies helping this? Ultra Ethernet just got released.

Ultra Ethernet will do almost nothing. It’s a rubber stamped version of Broadcom’s design and Marcel/Cisco/etc will just add it to their asics. Remains to be seen if SpecrumX will or Connectix. If not, none of it matters.

These chips are $30m-$100m projects a pop. After the embarrassingly brutal failure of Barefoot nobody is going to do ASICs.

Re: Fault Tolerant Llama training

#8
This is awesome, can’t wait to try out these techniques. At least a week a year of my time for the past few years has gone towards recovering from a fault crashing a training run. Sometimes environment related, sometimes shared storage, sometimes just because a slightly faulty IB cable.
Post reply on HN