Skip to content

Latest commit

 

History

History
22 lines (20 loc) · 1.73 KB

README.md

File metadata and controls

22 lines (20 loc) · 1.73 KB
Raw

COCOB-optimizer

PyTorch implementation of COCOB optimizer in Training Deep Networks without Learning Rates Through Coin Betting [1]

Usage

  1. Put cocob_bp.py in YOUR_PYTHON_PATH/site-packages/torch/optim.
  2. Open YOUR_PYTHON_PATH/site-packages/torch/optim/__init__.py add the following code:
from .cocob_bp import COCOB_Backprop                                                                                  
del cocob_bp
  1. Save __init__.py and restart your python.
    Use COCOB_Backprop as
optimizer = optim.COCOB_Backprop(net.parameters())                                                                    
...                                                                                                                   
optimizer.step()

Implemented by Huidong Liu

Reference

[1] Francesco Orabona and Tatiana Tommasi, Training Deep Networks without Learning Rates Through Coin Betting, NIPS 2017.