Tensorpack is a neural network training interface based on TensorFlow.

Documentation link:

Reinforcement Learning DOCUMENTATION

Presentation

Codes involved

Colab link

here

How to do a local setup on UBUNTU:

First Activate the virtual environment using this link Then,

$ pip install tensorflow==1.15.0

$ git clone https://github.com/sajalgoyal113/tensorpack
$ cd tensorpack/examples/DeepQNetwork
$ pip install 'gym[atari]'
$ wget https://github.com/openai/atari-py/raw/gdb/atari_py/atari_roms/breakout.bin

$ pip install --upgrade git+https://github.com/tensorpack/tensorpack.git

Train the model

$ ./DQN.py --env breakout.bin --algo DQN

Play the trained model ( here model-2000000 is pretrained model for 80 epochs)

$ ./DQN.py --env breakout.bin --task play --load ./train_log/DQN-breakout/model-2000000

Dependencies:

• Python 3.3+.
• Python bindings for OpenCV. (Optional, but required by a lot of features)
• TensorFlow ≥ 1.5, < 2. (Not required if you only want to use tensorpack.dataflow alone as a data processing library)