https://github.com/cmpark0126/pytorch-polynomial-lr-decay

Polynomial Learning Rate Decay Scheduler for PyTorch
https://github.com/cmpark0126/pytorch-polynomial-lr-decay

deep-learning learning-rate-decay learning-rate-scheduling pytorch pytorch-extension

Last synced: about 2 months ago
JSON representation

Polynomial Learning Rate Decay Scheduler for PyTorch

• Host: GitHub
• URL: https://github.com/cmpark0126/pytorch-polynomial-lr-decay
• Owner: cmpark0126
• License: mit
• Created: 2019-07-18T04:21:42.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2021-12-25T12:34:59.000Z (about 3 years ago)
• Last Synced: 2023-03-03T04:17:35.717Z (almost 2 years ago)
• Topics: deep-learning, learning-rate-decay, learning-rate-scheduling, pytorch, pytorch-extension
• Language: Python
• Size: 10.7 KB
• Stars: 53
• Watchers: 0
• Forks: 12
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# pytorch-polynomial-lr-decay

Polynomial Learning Rate Decay Scheduler for PyTorch

This scheduler is frequently used in many DL paper. But there is no official implementation in PyTorch. So I propose this code.

## Install

```

$ pip install git+https://github.com/cmpark0126/pytorch-polynomial-lr-decay.git

```

## Usage

```python

from torch_poly_lr_decay import PolynomialLRDecay

scheduler_poly_lr_decay = PolynomialLRDecay(optim, max_decay_steps=100, end_learning_rate=0.0001, power=2.0)

for epoch in range(train_epoch):

    scheduler_poly_lr_decay.step()     # you can handle step as epoch number

    ...

```

or

```python

from torch_poly_lr_decay import PolynomialLRDecay

scheduler_poly_lr_decay = PolynomialLRDecay(optim, max_decay_steps=100, end_learning_rate=0.0001, power=2.0)

...

for batch_idx, (inputs, targets) in enumerate(trainloader):

    scheduler_poly_lr_decay.step()     # also, you can handle step as each iter number

```