search

GAIL-4-BARK / bark-ml

Machine Learning Applied to Autonomous Driving
MIT License
0 stars 0 forks source link

Add TF2RL as a source code repository rather than a pip package #49

Closed MarcelBruckner closed 4 years ago

MarcelBruckner commented 4 years ago

I added the tf2rl repository to the bazel WORKSPACE as a github dependency

bazel now pulls the latest master of tf2rl when building and makes all code in the repo available to us.

This gives us the possibility to debug into tf2rl code and we can much easier see what happens inside the ddpg and gail networks as well as all other code that we import from the repository.

I implemented it so that we have all code available that is in the tf2rl folder of the repo. The import tf2rl in our code stays the same.

Steps to adjust to this change:

  1. After running source utils/dev_into.sh deinstall tf2rl to avoid import errors: pip uninstall tf2rl
  2. If you import tf2rl, add "@com_github_keiohta_tf2rl" to the deps = [ ... ] section in your BUILD target
  3. Add the tf2rl source code folder to your VS Code launch file: 3.1. RECOMMENDED: Generate a launch config for your file as described in section: Generate launch configuration script usage 3.2. OR Add to your "PYTHONPATH": "<...>:${workspaceFolder}/bazel-bin/<PATH-TO-THE-RUNFILES-FOLDER>/com_github_keiohta_tf2rl

If you encounter ModuleNotFoundError: No module named 'tf2rl' redo Step 3.