Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
michaelgiba
on May 4, 2024
|
parent
|
context
|
favorite
| on:
llm.c: multi-GPU, bfloat16, flash attention, ~7% f...
it’s pretty impressive that PyTorch is only 7% slower than this given it can be used so generally
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
