Run Sample
Run a sample script in samples directory. Enter the following command:
python samples/<FILE_NAME>
Example:
python samples/cartpole_v1_ppo.py
then, you can see the training information in your shell:
+----------------------------------------------+
| AINE-DRL Training Start! |
|==============================================|
| ID: CartPole-v1_PPO |
| Output Path: results/CartPole-v1_PPO |
|----------------------------------------------|
| Training INFO: |
| number of environments: 3 |
| total time steps: 20000 |
| summary frequency: 1000 |
| agent save frequency: 10000 |
|----------------------------------------------|
| PPO Agent: |
| n_steps: 16 |
| epoch: 3 |
| mini_batch_size: 8 |
| gamma: 0.99 |
| lam: 0.95 |
| advantage_normalization: False |
| epsilon_clip: 0.2 |
| value_loss_coef: 0.5 |
| entropy_coef: 0.001 |
| device: cpu |
+----------------------------------------------+
[AINE-DRL] training time: 1.18, time steps: 1000, cumulated reward: 36.15
[AINE-DRL] training time: 2.28, time steps: 2000, cumulated reward: 90.64
[AINE-DRL] training time: 3.38, time steps: 3000, cumulated reward: 94.22
When the training is finished, you can see the results (tensorboard, log message, agent save file) in results directory.
Open the tensorboard result by entering the command:
tensorboard --logdir=results
or
tensorboard --logdir=results/<SUB_DIRECTORY>
If you want to change the inference export format like gif, png (default: real-time rendering), you need to change the Inference setting in the configuration file. Follow the example:
Inference:
Config:
export: gif # default: render_only
export detail options: None, render_only, gif, png