Reproduce pensieve.
Use the commands below to easily set up virtual envs and requirements without worrying about version issues.
Set up conda env
conda create -n venev_tf1 pyhton=2.7
conda activate
Install dependencies
pip install tensorflow==1.1.0
pip install tflearn==0.3.1
conda install matplotlib
python multi_agent.py
To see how reward is changing over time.
python plot_reward.py
If you need to train on different videos, setup video files and run:
python get_video_sizes.py
Copy the model to abr_eval/models, change path in rl_no_training.py file.
python rl_no_training.py
python bb.py
python mpc.py
python dp.py