Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
NavinF
on July 7, 2022
|
parent
|
context
|
favorite
| on:
Google ML Compiler Inlining Achieves 3-7% Reductio...
It’s pretty easy to train models deterministically. Just use the same batch size, float type, initial seed, and flip this flag:
https://www.tensorflow.org/api_docs/python/tf/config/experim...
rfoo
on July 7, 2022
[–]
Caveat:
> on the same hardware
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
