78 lines
2.8 KiB
YAML
78 lines
2.8 KiB
YAML
# Copyright (c) 2021, salesforce.com, inc.
|
|
# All rights reserved.
|
|
# SPDX-License-Identifier: BSD-3-Clause
|
|
# For full license text, see the LICENSE file in the repo root
|
|
# or https://opensource.org/licenses/BSD-3-Clause
|
|
|
|
# YAML configuration for the tag continuous environment
|
|
name: "covid_and_economy_environment"
|
|
# Environment settings
|
|
env:
|
|
collate_agent_step_and_reset_data: True
|
|
components:
|
|
- ControlUSStateOpenCloseStatus:
|
|
action_cooldown_period: 28
|
|
- FederalGovernmentSubsidy:
|
|
num_subsidy_levels: 20
|
|
subsidy_interval: 90
|
|
max_annual_subsidy_per_person: 20000
|
|
- VaccinationCampaign:
|
|
daily_vaccines_per_million_people: 3000
|
|
delivery_interval: 1
|
|
vaccine_delivery_start_date: "2021-01-12"
|
|
economic_reward_crra_eta: 2
|
|
episode_length: 540
|
|
flatten_masks: True
|
|
flatten_observations: False
|
|
health_priority_scaling_agents: 0.3
|
|
health_priority_scaling_planner: 0.45
|
|
infection_too_sick_to_work_rate: 0.1
|
|
multi_action_mode_agents: False
|
|
multi_action_mode_planner: False
|
|
n_agents: 51
|
|
path_to_data_and_fitted_params: ""
|
|
pop_between_age_18_65: 0.6
|
|
risk_free_interest_rate: 0.03
|
|
world_size: [1, 1]
|
|
start_date: "2020-03-22"
|
|
use_real_world_data: False
|
|
use_real_world_policies: False
|
|
# Trainer settings
|
|
trainer:
|
|
num_envs: 60 # number of environment replicas
|
|
num_episodes: 1000 # number of episodes to run the training for
|
|
train_batch_size: 5400 # total batch size used for training per iteration (across all the environments)
|
|
# Policy network settings
|
|
policy: # list all the policies below
|
|
a:
|
|
to_train: True # flag indicating whether the model needs to be trained
|
|
algorithm: "PPO" # algorithm used to train the policy
|
|
vf_loss_coeff: 1 # loss coefficient schedule for the value function loss
|
|
entropy_coeff: 0.05 # loss coefficient schedule for the entropy loss
|
|
gamma: 0.98 # discount factor
|
|
lr: 0.0001 # learning rate
|
|
model:
|
|
type: "fully_connected"
|
|
fc_dims: [256, 256]
|
|
model_ckpt_filepath: ""
|
|
p:
|
|
to_train: True
|
|
algorithm: "PPO"
|
|
vf_loss_coeff: 1
|
|
entropy_coeff: # annealing entropy over time
|
|
- [0, 0.5]
|
|
- [50000000, 0.05]
|
|
gamma: 0.98
|
|
lr: 0.0001
|
|
model:
|
|
type: "fully_connected"
|
|
fc_dims: [256, 256]
|
|
model_ckpt_filepath: ""
|
|
# Checkpoint saving setting
|
|
saving:
|
|
metrics_log_freq: 100 # How often (in iterations) to print the metrics
|
|
model_params_save_freq: 500 # How often (in iterations) to save the model parameters
|
|
basedir: "/tmp" # base folder used for saving
|
|
name: "covid19_and_economy" # experiment name
|
|
tag: "experiments" # experiment tag
|