Floret Vectors on GPU with 40x Speedup #14014
Fazel94
started this conversation in
New Features & Project Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi,
Here is torch implementation that reach 1M training pairs per second on Colab free T4 gpu.
I was trying to train floret vectors for Persian(Farsi) language and my laptop did only 20K pairs per second so I vibed up a torch implementation and ran it on Colab free GPU that reached 1M pairs per second.
The full Farsi Wikipedia 1.3 GB bzipped amounts to 199M tokens would lead to a meager186s training time.
I hope this would prove useful to you.
All reactions