transformers
91fb62d0
- Speedup training by using numpy instead of jnp for batch shuffling (#15963)
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
4 years ago
Speedup training by using numpy instead of jnp for batch shuffling (#15963) Speedup training by using numpy instead of jnp for batch shuffling Co-authored-by: Yeb Havinga <y.t.havinga@mgrid.net>
References
#15963 - Speedup T5 Flax training by using Numpy instead of JAX for batch shuffling
Author
yhavinga
Parents
ea07064a
Loading