Megatron-DeepSpeed
349c45c7
- skip params on bwd; add argmin/argmax
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
4 years ago
skip params on bwd; add argmin/argmax
References
#155 - [debug] ModelInspector
Author
stas00
Parents
0aeab4f8
Loading