AI_for_prosthetics

NIPS 2018 competition: AI for prothetics Process of training and testing

1. Install

   • Install Anaconda3
   • Install packages from environment.yml

2. Training

   • cd opensim_http
   • ./run.sh
   • cd ../training
   • ipython -i ddpg2_v2.py
     • newpool(num_pools)
     • r(0,1000000) # to run 1000000 episodes of training

   Note: the training models are saved in model2 folder

3. Testing:

   • cd opensim_http
   • python server -l 127.0.0.1 -p 5000 open new command line
   • cd training
   • ipython -i ddpg2_test.py
     • agent.load_weights() # load saved episode in models (please copy your selected episode weight to models folder)
     • test()

4. Challenge: follow the instructions in nips2018-ai-for-prosthetics-round2-starter-kit instructions

5. Reference This work is inspired from https://github.com/ctmakro/stanford-osrl