def _test_parametric_dqn_workflow(self, use_gpu=False, use_all_avail_gpus=False):
"""Run Parametric DQN workflow to ensure no crashes, algorithm correctness
not tested here."""
with tempfile.TemporaryDirectory() as tmpdirname:
params = {
"training_data_path": os.path.join(
curr_dir, "test_data/parametric_action/cartpole_training.json.bz2"
),
"eval_data_path": os.path.join(
curr_dir, "test_data/parametric_action/cartpole_eval.json.bz2"
),
"state_norm_data_path": os.path.join(
curr_dir, "test_data/parametric_action/state_features_norm.json"
),
"action_norm_data_path": os.path.join(
curr_dir, "test_data/parametric_action/action_norm.json"
),
"model_output_path": tmpdirname,
"use_gpu": use_gpu,
"use_all_avail_gpus": use_all_avail_gpus,
"epochs": 1,
"rl": {},
"rainbow": {},
"training": {"minibatch_size": 128},
}
predictor = parametric_dqn_workflow.train_network(params)
test_float_state_features = [{"0": 1.0, "1": 1.0, "2": 1.0, "3": 1.0}]
test_int_state_features = [{}]
test_action_features = [{"4": 0.0, "5": 1.0}]
q_values = predictor.predict(
test_float_state_features, test_int_state_features, test_action_features
)
assert len(q_values[0].keys()) == 1
def test_parametric_dqn_workflow(self):
"""Run Parametric DQN workflow to ensure no crashes, algorithm correctness
not tested here."""
params = {
"training_data_path":
os.path.join(
curr_dir,
"test_data/parametric_action/cartpole_training_data.json"),
"state_norm_data_path":
os.path.join(
curr_dir,
"test_data/parametric_action/state_features_norm.json"),
"action_norm_data_path":
os.path.join(curr_dir,
"test_data/parametric_action/action_norm.json"),
"model_output_path":
None,
"use_gpu":
False,
"epochs":
1,
"rl": {},
"rainbow": {},
"training": {
"minibatch_size": 16
},
"in_training_cpe":
None,
}
predictor = parametric_dqn_workflow.train_network(params)
test_float_state_features = [{"0": 1.0, "1": 1.0, "2": 1.0, "3": 1.0}]
test_int_state_features = [{}]
test_action_features = [{"4": 0.0, "5": 1.0}]
q_values = predictor.predict(test_float_state_features,
test_int_state_features,
test_action_features)
assert len(q_values[0].keys()) == 1
