Evaluate trained policies

With the local installation

Trained models can be evaluated and the results can be saved with the script evaluate_policy.py.

Usage

Flag	Description	Type	Example
--exp-id	Unique experiment ID	int	99
--n-eval-steps	Number of evaluation timesteps	int	1000
--log-info	Enable information logging at each evaluation steps	bool	0 (default) or 1
--plot-dim	Live rendering of end-effector and goal positions	int	0: do not plot (default)	2: 2D or 3: 3D
--render	Render environment during evaluation	bool	0 (default) or 1

Example:

python evaluate_policy.py --exp-id 99 --n-eval-steps 1000 --log-info 0 --plot-dim 0 --render 0

The plots are printed in the associated experiment folder, e.g. logs/exp_99/ppo/.

If --log-info was enabled during evaluation, it is possible to plot some useful information as shown in the plot below.

python scripts/plot_episode_eval_log.py --exp-id 99

An example of environment evaluation plot:

An example of experiment learning curve:

With Docker

Trained policies can be evaluated from the Docker containers. The result files and plots will generated in the working directory of the host machine.

# CPU
docker run -it --rm --network host --ipc=host --mount src=$(pwd),target=/root/rl_reach/,type=bind rlreach/rlreach-cpu:latest bash -c "python evaluate_policy.py --exp-id 99 --n-eval-steps 1000 --log-info 0 --plot-dim 0 --render 0"
# GPU
docker run -it --rm --runtime=nvidia --network host --ipc=host --mount src=$(pwd),target=/root/rl_reach/,type=bind rlreach/rlreach-gpu:latest bash -c "python evaluate_policy.py --exp-id 99 --n-eval-steps 1000 --log-info 0 --plot-dim 0 --render 0"

A Shell script is provided for ease of usability.

# CPU
./docker/run_docker_cpu.sh python evaluate_policy.py --exp-id 99 --n-eval-steps 1000 --log-info 0 --plot-dim 0 --render 0
# GPU
./docker/run_docker_gpu.sh python evaluate_policy.py --exp-id 99 --n-eval-steps 1000 --log-info 0 --plot-dim 0 --render 0

Reset the repository

It can sometimes be convenient to clean all the results and log files and reset the repository. This can be done with the following Shell script.

./cleanAll.sh

Warning

Warning, this cannot be undone!