Megatron-DeepSpeed
189f0547
- Test out the loss from the fp32 weights and optimizer states
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
3 years ago
Test out the loss from the fp32 weights and optimizer states
References
thomas/fix_layer_norm
#271 - Sync layer norm
Author
thomasw21
Parents
c3844b5c
Loading