Megatron-DeepSpeed
f3307058
- Fix adaptive_seq_len via resetting activation shape
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
3 years ago
Fix adaptive_seq_len via resetting activation shape
References
#212 - Eval harness
#291 - BigScience Eval Harness
#313 - Prefix LM Eval
Author
Muennighoff
Parents
1c11b107
Loading