accelerate
Make the Scheduler adjust the steps taken relative to the gradient accumulation steps
#1187
Merged

Make the Scheduler adjust the steps taken relative to the gradient accumulation steps #1187

muellerzr merged 22 commits into main from scheduler-docs
muellerzr
muellerzr Make scheduler actually adjust the length
94181fda
muellerzr Print
df904b19
muellerzr Use last_epoch
099bfe06
muellerzr Try now
8006f8c9
muellerzr Make scheduler step based on gradient accumulation
01f9a0a9
muellerzr muellerzr added enhancement
muellerzr muellerzr requested a review from sgugger sgugger 2 years ago
muellerzr Put decorator in the right place
8744f408
HuggingFaceDocBuilderDev
muellerzr clean
79ff7c56
sgugger
sgugger commented on 2023-03-13
muellerzr Rework it all, better version and working now with tests
60f3009f
muellerzr Fix tests
250fe900
muellerzr Bring back grad_accum_steps
71619d5a
muellerzr muellerzr marked this pull request as draft 2 years ago
muellerzr Use plugin
f19a4b0e
muellerzr Fix docstrings
744686e6
muellerzr Formatting nit
b39fa980
muellerzr
muellerzr muellerzr marked this pull request as ready for review 2 years ago
sgugger
sgugger approved these changes on 2023-03-14
muellerzr Proper kwarg
3acda862
muellerzr muellerzr requested a review from sgugger sgugger 2 years ago
muellerzr Raise err
97d1af23
muellerzr Move around check for TPU
0d922d4a
sgugger
sgugger approved these changes on 2023-03-14
muellerzr fix para name
4a473335
muellerzr Fix steps w/ deepspeed
04495041
muellerzr Good import
bb5a374b
muellerzr Try with this
f3bce841
muellerzr Try now?
79268088
muellerzr Should be working now, just need to test on CI
248070f4
muellerzr muellerzr merged e4620984 into main 2 years ago
muellerzr muellerzr deleted the scheduler-docs branch 2 years ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone