def run_rl_example(env, agent, params, mode="visualize"):
env.ml_behavior = agent
runner = SACRunner(params=params,
environment=env,
agent=agent)
if mode == "train":
runner.SetupSummaryWriter()
runner.Train()
elif mode == "visualize":
runner.Visualize(5)
elif mode == "evaluate":
runner.Evaluate()
def run_configuration(argv):
# Uncomment one of the following default parameter filename definitions,
# depending on which GNN library you'd like to use.
# File with standard parameters for tf2_gnn use:
# param_filename = "examples/example_params/tfa_sac_gnn_tf2_gnn_default.json"
# File with standard parameters for spektral use:
param_filename = "examples/example_params/tfa_sac_gnn_spektral_default.json"
params = ParameterServer(filename=param_filename)
# NOTE: Modify these paths to specify your preferred path for checkpoints and summaries
# params["ML"]["BehaviorTFAAgents"]["CheckpointPath"] = "YOUR_PATH"
# params["ML"]["TFARunner"]["SummaryPath"] = "YOUR_PATH"
#viewer = MPViewer(
# params=params,
# x_range=[-35, 35],
# y_range=[-35, 35],
# follow_agent_id=True)
#viewer = VideoRenderer(
# renderer=viewer,
# world_step_time=0.2,
# fig_path="/your_path_here/training/video/")
# create environment
bp = ContinuousHighwayBlueprint(params,
number_of_senarios=2500,
random_seed=0)
observer = GraphObserver(params=params)
env = SingleAgentRuntime(blueprint=bp, observer=observer, render=False)
sac_agent = BehaviorGraphSACAgent(environment=env,
observer=observer,
params=params)
env.ml_behavior = sac_agent
runner = SACRunner(params=params, environment=env, agent=sac_agent)
if FLAGS.mode == "train":
runner.SetupSummaryWriter()
runner.Train()
elif FLAGS.mode == "visualize":
runner.Visualize(5)
elif FLAGS.mode == "evaluate":
runner.Evaluate()