waterloo intern

model perf @baseten
eng @uwaterloo

[waterloo intern @waterloo_intern]

- 230 training runs
- 1,623 GPU hours (67 B200 days)
- 76 TB of training data
- a 2x faster model

Every paper said it can't be done.
Quantization Aware Distillation made it possible.

[waterloo intern @waterloo_intern]

https://t.co/GRwnDpMSzA

[waterloo intern @waterloo_intern]

@elliotarledge ahhh

[waterloo intern @waterloo_intern]

@wivincent flexxx

[waterloo intern @waterloo_intern]

@BrianLi23 i intentionally brought yours down to steal your gpus

[waterloo intern @waterloo_intern]

anyways if you work for an inference lab you get 17x8 B200 nodes to try stuff out
...
~1.2 EFLOPS FP8 for an intern

[roon @tszzl]

there is no such thing as running out of compute. for the right price someone will sell you compute. it’s an elastic resource like all other markets. when RSI arrives running that program will be so valuable that all clouds will mostly shut down and sell compute to the singularity

[waterloo intern @waterloo_intern]

@urarthur yessir!

[waterloo intern @waterloo_intern]

@Jorjy_k only at Baseten

[waterloo intern @waterloo_intern]

@alphatozeta8148 yes pleaseee

[waterloo intern @waterloo_intern]

@_Suresh2 😂

[waterloo intern @waterloo_intern]

@AzeSaah you got this!!

[waterloo intern @waterloo_intern]

@JackRNewhouse to get moreeeeee

[waterloo intern @waterloo_intern]

@NVIDIAAI a reply from NVIDIA is crazyyy...give me more GPUs please?

[waterloo intern @waterloo_intern]

@KyrieBlunders never

[waterloo intern @waterloo_intern]

@aishwarya_08 hahahahaha

[waterloo intern @waterloo_intern]

@GPTWare never enough lol

[waterloo intern @waterloo_intern]

@IKorovinsky mmmmaybe _||_

[waterloo intern @waterloo_intern]

@ZayaanMulla my good sir

[waterloo intern @waterloo_intern]

@philipkiely sorry : )

[waterloo intern @waterloo_intern]

@baseten legalmogging the competition

[waterloo intern @waterloo_intern]

@mudithj @tenderizzation sorry m9 but we're going to take over