transformers
FEAT / Optim: Add GaLore optimizer
#29588
Merged

FEAT / Optim: Add GaLore optimizer #29588

younesbelkada
younesbelkada add galore v1
b31ce79c
HuggingFaceDocBuilderDev
younesbelkada
muellerzr
muellerzr commented on 2024-03-11
younesbelkada add import
58169f15
younesbelkada add tests and doc
9032635c
younesbelkada fix doctest
136f104c
younesbelkada younesbelkada marked this pull request as ready for review 2 years ago
forward contrib credits from discussions
a5483b36
forward contrib credits from discussions
887d3adc
younesbelkada younesbelkada requested a review from muellerzr muellerzr 2 years ago
younesbelkada younesbelkada requested a review from pacman100 pacman100 2 years ago
younesbelkada younesbelkada changed the title DRAFT / Optim: Add GaLore optimizer FEAT / Optim: Add GaLore optimizer 2 years ago
muellerzr
muellerzr approved these changes on 2024-03-11
younesbelkada Apply suggestions from code review
d6f119fb
younesbelkada Merge remote-tracking branch 'upstream/main' into HEAD
3fae2290
younesbelkada fix failing tests'
c8c50f80
younesbelkada younesbelkada requested a review from amyeroberts amyeroberts 2 years ago
BenjaminBossan
BenjaminBossan commented on 2024-03-11
amyeroberts
amyeroberts commented on 2024-03-11
PenutChen
pacman100
pacman100 commented on 2024-03-12
hiyouga
younesbelkada Merge remote-tracking branch 'upstream/main' into add-galore-optimizer
2bdda681
younesbelkada switch to `optim_target_modules` and clarify docs
630bd13c
younesbelkada more clarification
a871b75a
younesbelkada Merge remote-tracking branch 'upstream/main' into add-galore-optimizer
cb6cd7e9
younesbelkada enhance lookup logic
51b7b292
younesbelkada update a test to add peak memory
3da3b90d
younesbelkada add regex, all-linear and single string support
9115c94d
younesbelkada add layer-wise optimization through DummyOptimizers and LRSchedulers
0b4ba838
hiyouga forward contrib credits from discussions and original idea
3e5930ef
younesbelkada
younesbelkada younesbelkada requested a review from BenjaminBossan BenjaminBossan 2 years ago
younesbelkada younesbelkada requested a review from pacman100 pacman100 2 years ago
younesbelkada younesbelkada requested a review from amyeroberts amyeroberts 2 years ago
younesbelkada younesbelkada requested a review from muellerzr muellerzr 2 years ago
hiyouga
younesbelkada
younesbelkada add a section about DDP not supported in layerwise
a16d3a87
muellerzr
muellerzr commented on 2024-03-13
younesbelkada Update src/transformers/trainer.py
29e7e94f
younesbelkada fix self
18ea144a
younesbelkada check only if layer_wise
7800bf1f
kiddyboots216
peterjc123
peterjc123 commented on 2024-03-13
younesbelkada
peterjc123
peterjc123 commented on 2024-03-13
peterjc123
kiddyboots216
RefractAI
amyeroberts
amyeroberts commented on 2024-03-13
BenjaminBossan
BenjaminBossan commented on 2024-03-13
PenutChen
PenutChen commented on 2024-03-14
younesbelkada Update src/transformers/training_args.py
e022bdda
younesbelkada oops
830c68df
younesbelkada make use of intervals
b640e980
younesbelkada clarify comment
14a89b2f
younesbelkada add matching tests
6f7102db
younesbelkada GaLoRe -> GaLore
c11cb63e
younesbelkada move to `get_scheduler`
36782019
younesbelkada add note on docs
fdc4b2a2
younesbelkada add a warning
e7ce9b7d
younesbelkada
younesbelkada younesbelkada requested a review from amyeroberts amyeroberts 2 years ago
younesbelkada younesbelkada requested a review from BenjaminBossan BenjaminBossan 2 years ago
younesbelkada younesbelkada requested a review from muellerzr muellerzr 2 years ago
BenjaminBossan
BenjaminBossan approved these changes on 2024-03-14
younesbelkada adapt a bit the docs
91d64368
younesbelkada update docstring
b9e338a0
winglian
younesbelkada support original API
6ff37620
younesbelkada
hiyouga
hiyouga commented on 2024-03-17
younesbelkada
younesbelkada commented on 2024-03-17
younesbelkada Update docs/source/en/trainer.md
0d0440a1
BenjaminBossan
BenjaminBossan commented on 2024-03-18
younesbelkada slightly refactor
832f2be9
matthewdouglas
matthewdouglas commented on 2024-03-18
younesbelkada Update docs/source/en/trainer.md
898a3c5a
amyeroberts
amyeroberts approved these changes on 2024-03-18
winglian
winglian
winglian
younesbelkada Update src/transformers/training_args.py
ed3ad4ad
younesbelkada fix args parsing and add tests
57e7096e
younesbelkada
younesbelkada remove warning for regex
64ccfa6b
younesbelkada Merge remote-tracking branch 'upstream/main' into add-galore-optimizer
4413f074
younesbelkada fix type hint
73dcabb8
younesbelkada add note about extra args
1987b7ae
younesbelkada make `is_regex` return optional
db2bf219
younesbelkada younesbelkada merged f6261d7d into main 2 years ago
fakerybakery
NicolasMejiaPetit
kiddyboots216
NicolasMejiaPetit
kiddyboots216

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone