Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
amindiro
62 days ago
|
parent
|
context
|
favorite
| on:
I rebuilt FlashAttention in Triton to understand t...
Agreed, this observation holds true for both decode and prefill. Thanks for sharing the code
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
