Question Behavior Name DOES match the one in the trainer configuration file but default config is still used?

Sorry for the weird phrasing on the title, had to keep it short.

I want to change the max_step parameter in the configurations, and despite the Behavior Name in Unity matching the one specified in configuration.yaml (that name being "WarshAgent"), MLAgents still insists on hitting me with this:

[WARNING] Behavior name WarshAgent does not match any behaviors specified in the trainer configuration file. A default configuration will be used.

And then uses a default configuration, so everytime I changed the max_step, it would get changed back to the default value.

I did try using the command "python -m mlagents.trainers.upgrade_config configuration.yaml
WarshAgent.yaml" but all that did was copy the content of configuration.yaml to WarshAgent.yaml, even when I deleted configuration.yaml it still created a new one and completely ignored WarshAgent.yaml.

Here are more details:

Behavior Parameters:


configuration.yaml:

Spoiler

default_settings: null
behaviors:
WarshAgent:
trainer_type: ppo
hyperparameters:
batch_size: 1024
buffer_size: 10240
learning_rate: 0.0003
beta: 0.005
epsilon: 0.2
lambd: 0.95
num_epoch: 3
learning_rate_schedule: linear
beta_schedule: linear
epsilon_schedule: linear
network_settings:
normalize: false
hidden_units: 128
num_layers: 2
vis_encode_type: simple
memory: null
goal_conditioning_type: hyper
deterministic: false
reward_signals:
extrinsic:
gamma: 0.99
strength: 1.0
network_settings:
normalize: false
hidden_units: 128
num_layers: 2
vis_encode_type: simple
memory: null
goal_conditioning_type: hyper
deterministic: false
init_path: null
keep_checkpoints: 5
checkpoint_interval: 500000
max_steps: 500000
time_horizon: 64
summary_freq: 50000
threaded: false
self_play: null
behavioral_cloning: null
env_settings:
env_path: null
env_args: null
base_port: 5005
num_envs: 1
num_areas: 1
seed: -1
max_lifetime_restarts: 10
restarts_rate_limit_n: 1
restarts_rate_limit_period_s: 60
engine_settings:
width: 84
height: 84
quality_level: 5
time_scale: 20
target_frame_rate: -1
capture_frame_rate: 60
no_graphics: false
environment_parameters: null
checkpoint_settings:
run_id: ppo
initialize_from: null
load_model: false
resume: true
force: false
train_model: false
inference: false
results_dir: results
torch_settings:
device: null
debug: false

WarshAgent.yaml:

Spoiler

behaviors:
WarshAgent:
trainer_type: ppo
hyperparameters:
batch_size: 1024
buffer_size: 10240
learning_rate: 0.0003
beta: 0.005
epsilon: 0.2
lambd: 0.95
num_epoch: 3
learning_rate_schedule: linear
beta_schedule: linear
epsilon_schedule: linear
network_settings:
normalize: false
hidden_units: 128
num_layers: 2
vis_encode_type: simple
goal_conditioning_type: hyper
deterministic: false
reward_signals:
extrinsic:
gamma: 0.99
strength: 1.0
network_settings:
normalize: false
hidden_units: 128
num_layers: 2
vis_encode_type: simple
goal_conditioning_type: hyper
deterministic: false
keep_checkpoints: 5
checkpoint_interval: 500000
max_steps: 900000000
time_horizon: 64
summary_freq: 50000
threaded: false
env_settings:
base_port: 5005
num_envs: 1
num_areas: 1
seed: -1
max_lifetime_restarts: 10
restarts_rate_limit_n: 1
restarts_rate_limit_period_s: 60
engine_settings:
width: 84
height: 84
quality_level: 5
time_scale: 20
target_frame_rate: -1
capture_frame_rate: 60
no_graphics: false
checkpoint_settings:
run_id: ppo
load_model: false
resume: true
force: false
train_model: false
inference: false
results_dir: results
torch_settings: {}
debug: false

What am I doing wrong? Or am is my understanding of how this works is just wrong?

 

Im having the same issue

 

bump

 

I found out that typing the command "mlagents-learn <path_to_your_config_file> --run-id=YourSimulationID" works


For example, for me, it was : "mlagents-learn results/Test/MoveToTarget.yaml --run-id=Test --force" (added "force" to reset the neural network)