Megatron-DeepSpeed
Fix curriculum learning support
#134
Merged

Fix curriculum learning support #134

stas00 merged 26 commits into bigscience-workshop:main from main
conglongli
conglongli CL initial commit
99d2b37d
conglongli CL+PP support
4c9c4a39
conglongli update
82a31984
conglongli Apply suggestions from code review
21e91b97
conglongli apply code review comments
6010a3dd
stas00 make it easier to read large numbers
405c7a69
stas00 add a cl test
a90d30eb
conglongli apply review comments
fb04d2be
conglongli Update examples/curriculum_learning/README.md
8e4a4660
stas00 Merge branch 'main' of https://github.com/conglongli/Megatron-DeepSpe…
3ed70757
stas00 update
d86a4f42
stas00 fix
e5a335dd
stas00 new requirement
0c4073b6
conglongli Update megatron/learning_rates.py
d25fa9e0
conglongli Update megatron/learning_rates.py
7cd53dc0
stas00 fix samples and tokens - thank you Conglong
5a492b34
conglongli fix truncation
8ca1db7f
stas00 switch to deepspeed@master
d7301a1b
stas00 extend the doc
dbf8abdd
stas00 Trigger CI
b7fd67ed
conglongli merge upstream
c0ec861c
conglongli fix CL+PP
20d498d7
conglongli rollback accidental changes
f4c4eb34
conglongli relax assertion for corner case
df7a9d95
stas00
conglongli backward compatibility for new chkpt keys
ae24cd18
stas00
conglongli Merge remote-tracking branch 'upstream/main' into main
3e521045
stas00 stas00 merged 8dc8af51 into main 4 years ago

Login to write a write a comment.

Login via GitHub

Reviewers
No reviews
Assignees
No one assigned
Labels
Milestone