Skip to content

DeeperMind/AugmentedNeuralArchitectureSearch_with_REINFORCE

 
 

Repository files navigation

COMP767 Project: Augmented Neural Architecture Search with REINFORCE

Contact Information

Olga (Geya) Xu, 260520542

geya.xu@mail.mcgill.ca

Model Description

model_description

Dependencies

  • numpy==1.14.2
  • torch.egg==info
  • torchvision==0.2.0
  • torch==0.4.0a0+6a41e2d

Train Models

MNIST

python run_mnist.py

usage: run_mnist.py [-h] [--batch-size BATCH_SIZE] [--max-layers MAX_LAYERS] [--alpha ALPHA] [--gamma GAMMA] [-lr LEARNING_RATE] [--num-episodes NUM_EPISODES] [--model-dir MODEL_DIR] [--model-name MODEL_NAME] [--save-freq SAVE_FREQ]

CIFAR-10

python run_cifar.py

usage: run_cifar10.py [-h] [--batch-size BATCH_SIZE] [--max-layers MAX_LAYERS] [--alpha ALPHA] [--gamma GAMMA] [-lr LEARNING_RATE] [--num-episodes NUM_EPISODES] [--model-dir MODEL_DIR] [--model-name MODEL_NAME] [--check-memory] [--save-freq SAVE_FREQ] [--load-model LOAD_MODEL]

Test Models

MNIST

python test_mnist.py --load-model [MODEL]

usage: test_mnist.py [-h] [--batch-size BATCH_SIZE] [--max-layers MAX_LAYERS] [--alpha ALPHA] [-lr LEARNING_RATE] [--num-epochs NUM_EPOCHS] [--model-dir MODEL_DIR] [--model-name MODEL_NAME] [--load-path LOAD_PATH]

CIFAR-10

python test_cifar.py --load-model [MODEL]

usage: test_cifar.py [-h] [--batch-size BATCH_SIZE] [--alpha ALPHA] [-lr LEARNING_RATE] [--num-epochs NUM_EPOCHS] [--model-dir MODEL_DIR] [--model-name MODEL_NAME] [--load-path LOAD_PATH]

Display Replay Tree

python print_experience_tree.py --load-model [MODEL]

usage: print_experience_tree.py [-h] --load-model LOAD_MODEL

Reference

  • Neural Architecture Search with Reinforcement Learning [arxiv]
  • Convolutional Neural Network Architecture Seach with Q-Learning [paper]

Acknowledgements

Special thanks to the authors in [reinforce.py] for their REINFORCE implementation and the authors in [main.py] for their residual block implementation. I have used some parts of their implementation.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%