Lunarium 2: Machine Learning Agent Experiment Day 2
Machine Learning Hummingbird agent
Scene set up and script walkthrough to understand essential
Components that make up the machine learning set up and functionality.
Ray set up, down and front observes objects in the scene
Similar to Lidar
Rays are set away from beak to avoid false detection
Environment boxed in to facilitate bird training
Nectar Full Flower
On Empty, Flower changes colour
Download and install Anaconda
Creating a new environment
Python 3.7 was incorrect : use Python 3.8 (64 bit)
Download latest update from pip - use:
pip install six==1.10.0 $ pip install cherrypy==11.0.0
Fix Courtesy of Pradyuns
;conta activatr ml-agents -3.0 activates this environment
Pypi is deprecated
Set up 8 Training Environments identical to each other.
Ensure Hummingbird is spelt correctly in behaviour
Use this yml file instead
behaviors:
Hummingbird:
trainer_type: ppo
hyperparameters:
batch_size: 2048
buffer_size: 20480
learning_rate: 0.0003
beta: 0.005
epsilon: 0.2
lambd: 0.95
num_epoch: 3
learning_rate_schedule: linear
network_settings:
normalize: false
hidden_units: 256
num_layers: 2
vis_encode_type: simple
reward_signals:
extrinsic:
gamma: 0.99
strength: 1.0
keep_checkpoints: 5
checkpoint_interval: 500000
max_steps: 5000000
time_horizon: 128
summary_freq: 10000
threaded: true
Yml file courtesy of Point_Nemo on Unity Github
When correct the unity logo will be displayed.
One you see go to Unity and press play, you should do this immediately.
If correct you will begin to see data downloading from the hummingbird
It will take time to see it but the bird will begin to learn about its environment and the flowers within it as it starts to look for nectar.
Use tensorboard—logdir results instead
Represents the same data as a graph for ease of reference
Here it begins to show the hummingbird learning how to successfully locate the nectar.
There is a lot of information here !
And even more here !
The hummingbird is successfully drinking the flowers dry of nectar as they turn from pink to purple and move on.
This is where the data is stored on the drive.
I have left the the training to run to continue the learning process so that I can begin the next part tomorrow.
Unity Repository
