-
Notifications
You must be signed in to change notification settings - Fork 977
Pull requests: EleutherAI/gpt-neox
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Save checkpoint metadata when using SlurmRunner
#733
by dashstander
was merged Dec 8, 2022
Loading…
updated Dec 8, 2022
Adds the proper DeepSpeed bf16 config
#786
by dashstander
was closed Feb 14, 2023
Loading…
updated Feb 14, 2023
Add DeepSpeed bf16 configuration
#787
by dashstander
was merged May 16, 2023
Loading…
updated May 16, 2023
Updates bf16 demo config and mixed precision docutmentation.
#941
by dashstander
was merged May 18, 2023
Loading…
updated May 18, 2023
Return to broadcasting megatron and deepspeed configs as training.py arguments
#948
by dashstander
was merged May 23, 2023
Loading…
updated May 23, 2023
Bump transformers version and update enwik8 link
#1024
by dashstander
was merged Sep 13, 2023
Loading…
updated Sep 13, 2023
Fix bf16 for zero > 0 and pipeline parallelism > 0
#1032
by dashstander
was merged Sep 18, 2023
Loading…
updated Sep 18, 2023
Remove support for lazy dataset implementation
#1033
by dashstander
was merged Sep 18, 2023
Loading…
updated Sep 18, 2023
Remove the NeoX implementation of GPT2Tokenizer
#1042
by dashstander
was merged Sep 25, 2023
Loading…
updated Sep 25, 2023
Pre-compute RoPE embeddings in fp32
#1041
by dashstander
was merged Sep 25, 2023
Loading…
updated Sep 25, 2023
Improve FLOPS Calculation
#1044
by dashstander
was merged Sep 27, 2023
Loading…
updated Sep 27, 2023
Add section to the README detailing how to start distributed jobs
#1048
by dashstander
was merged Sep 29, 2023
Loading…
updated Sep 29, 2023
Organize the
tools
directory
#1055
by dashstander
was merged Oct 2, 2023
Loading…
updated Oct 2, 2023
Add documentation about using labelled datasets
#1056
by dashstander
was merged Oct 4, 2023
Loading…
updated Oct 4, 2023
LR scheduler fix no longer breaks inference
#1060
by dashstander
was merged Oct 17, 2023
Loading…
updated Oct 17, 2023
ProTip!
Updated in the last three days: updated:>2024-06-28.