DeepSpeed
56e7e0c5
- Consistent fp32 grads flow (#8056)
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
79 days ago
Consistent fp32 grads flow (#8056) Avoid bf16 buffers when doing fp32 gradient reduction and accumulation --------- Signed-off-by: tunji-ruwase_snow <tunji.ruwase@snowflake.com>
References
#8056 - Consistent fp32 grads flow
Author
sfc-gh-truwase
Parents
87f31d55
Loading