Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
anvuong
5 months ago
|
parent
|
context
|
favorite
| on:
Helion: A high-level DSL for performant and portab...
You mean autotune? I think 10 minutes is pretty normal, torch.compile('max-autotune') can be much slower than that for large models.
Mars008
5 months ago
[–]
Add to that it can be done only once by developers before distribution for major hardware. Configs saved. Then on client side selected.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: