lessw2020 Ranger21 issues

lessw2020 / Ranger21

Ranger deep learning optimizer rewrite to use newest components

Apache License 2.0

323 stars 46 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

- Enable / Disable showing settings

#51 okbalefthanded opened 9 months ago
0
link to https://github.com/kozistr/pytorch_optimizer

#50 wassname opened 1 year ago
0
Error when using DDP

#49 smartbarbarian opened 1 year ago
1
learning rate scheduler

#48 shyhyawJou opened 1 year ago
0
Recommended settings for transformers?

#47 OhadRubin opened 2 years ago
1
hi，please help me

#46 wudizuixiaosa closed 2 years ago
0
Nice name of your project)

#45 Ranger21 opened 2 years ago
0
Use collections.abc.Callable

#44 JackKelly closed 2 years ago
1
AttributeError: module 'collections' has no attribute 'Callable'

#43 JackKelly closed 2 years ago
0
What it the best hyper-parameter setting?

#42 NoOneUST opened 2 years ago
2
Gradient normalization lowers the maximum learning rate that can converge.

#41 Handagot opened 2 years ago
0
Can ranger be used for NLP transformers?

#40 LifeIsStrange closed 2 years ago
2
Not support pytorch _1.3.1

#39 huangnengCSU opened 2 years ago
0
Require an documentation

#38 huangnengCSU opened 2 years ago
2
decouple the lr scheduler and optimizer?

#36 hiyyg opened 3 years ago
5
sample usage in fastai

#35 nikky4D opened 3 years ago
1
local variable 'neg_grad_ma' referenced before assignment when momentum_type is not "pnm"

#34 lechmazur closed 3 years ago
0
removed torch.linalg as LA mention

#33 pranshurastogi29 closed 3 years ago
0
allow parallel, division by zero error

#32 ryancinsight closed 1 year ago
0
Allow parallel patch based training

#31 ryancinsight closed 3 years ago
3
RuntimeError: hit nan for variance_normalized

#30 gcp opened 3 years ago
7
About gradient normalization

#29 julightzhong10 opened 3 years ago
0
error in warmdown - lr below min lr. current lr = 2.999999999999997e-0518 [07:50<00:04, 4.66s/it] auto handling but please report issue!

#28 neuronflow opened 3 years ago
2
Performance of ResNet50 on ImageNet

#27 juntang-zhuang opened 3 years ago
1
Update README.md

#26 nestordemeure closed 3 years ago
1
SAM paper

#25 ryanstout opened 3 years ago
0
File "/home/.../site-packages/ranger21/ranger21.py", line 680, in step raise RuntimeError("hit nan for variance_normalized")

#24 neuronflow closed 3 years ago
1
Activate/deactivate softplus for MADGRAD & choosing beta softplus

#23 TheZothen closed 3 years ago
1
Some fixes when using MADGRAD (Softplus and Stable weight decay)

#22 TheZothen closed 3 years ago
2
comparing ranger21 to SAM optimizer

#21 nikky4D closed 3 years ago
2
error when training with batch_size = 1

#20 neuronflow opened 3 years ago
0
hit nan for variance_normalized

#19 jimmiebtlr closed 3 years ago
7
resuming training with ranger21?

#18 neuronflow opened 3 years ago
3
Multi GPU problem

#17 zsgj-Xxx opened 3 years ago
4
lr below min_lr check too aggressive

#16 kai-tub opened 3 years ago
2
added reference to Ranger21 paper

#13 nestordemeure closed 3 years ago
1
optimizer = Ranger21(params=model.parameters(), lr=learning_rate) File "/mnt/Drive1/florian/msblob/Ranger21/ranger21/ranger21.py", line 179, in __init__ self.total_iterations = num_epochs * num_batches_per_epoch TypeError: unsupported operand type(s) for *: 'NoneType' and 'NoneType'

#12 neuronflow opened 3 years ago
6
adaptive gclipping: unable to process len of 5 - currently must be <= 4

#11 HungYu-Wu closed 3 years ago
6
Adaptive Gradient Clipping

#10 benihime91 closed 3 years ago
3
Changes in lr

#9 zsgj-Xxx closed 3 years ago
7
torch.grad removed in PyTorch 1.8.1?

#8 jszym closed 3 years ago
2
Augmentation requests

#5 LifeIsStrange closed 3 years ago
2
Example

#3 johnyquest7 closed 3 years ago
1
Package and repo reorg

#2 BrianPugh closed 3 years ago
1
Adaptive Gradient Clipping

#1 kayuksel closed 3 years ago
2