Soumith Chintala: I like the TPU vs GPU competition, a...| Robocrunch AI

I like the TPU vs GPU competition, and MFU as a metric is pretty healthy.

Congrats to Google and NVIDIA for hitting 50%+ MFUs on their large training runs:https://t.co/P5ywgxhP3o https://t.co/tOepgknGUV
— Soumith Chintala (@soumithchintala) May 12, 2022

Related

https://www.kaggle.com/c/ultra-mnist/

Do you want an NVIDIA RTX 3080Ti GPU for FREE? 😍 . . . . . . . . . . . . . . . . . . . . . . . . . . . . Now that I have your attention, I must ...

Shared by abhishek at 3/17/2022 0-10

https://github.com/abhishekkrthakur/tez

Finally updating Tez and adding multi-gpu and hopefully tpu support 😉 https://t.co/BxKAJkeGh8 800 stars only. Give some love to the simplest Py ...

Shared by abhishek at 3/16/2022 3-9

https://forms.gle/mLtPEXbpjjvWvpxq8

Next Wednesday on Physics ∩ ML, hear me explain how to tune GPT-3 on a single GPU using new theoretical advances! What are the practical and the ...

Shared by Greg Yang at 3/16/2022 0-3

https://link.medium.com/iJcJNIyC5pb

Train 18-billion-parameter GPT models with a single GPU on your personal computer! https://t.co/4uRx1OINVe

Shared by Yang You at 5/16/2022 0-6

https://twitter.com/ylecun/status/1485748926003167232

M/FAIR(?) unveils their GPU cluster, whose main motivation seems to be training.... Transformers! 😁 I hope ConvNets will be allowed to keep it ...

Shared by Lucas Beyer at 1/24/2022 1-25

https://nvda.ws/3JG2spU

Get an in-depth look at the new NVIDIA H100 GPU and the new features of the NVIDIA Hopper architecture to improve strong scaling for #AI and #HP ...

Shared by NVIDIA AI at 3/24/2022 5-11

768