def test_get_epsilon(self):
tp = TrainingParam()
tp.final_epsilon = None
eps = tp.get_next_epsilon(1)
assert eps == 0.
tp.final_epsilon = 0.01
tp.initial_epsilon = None
eps = tp.get_next_epsilon(1)
assert eps == 0.
tp.initial_epsilon = 0.01
tp.final_epsilon = 0.01
eps = tp.get_next_epsilon(1)
assert eps == 0.01
tp.min_iter = None
tp.update_nb_iter = None # once 100 scenarios are solved, increase of "step_increase_nb_iter"
# oversampling hard scenarios
tp.oversampling_rate = None # None to deactivate it
# experience replay
tp.buffer_size = 1000000
# just observe the data for a while
tp.min_observe = None # int(10000)
# e greedy
tp.min_observation = 128
tp.initial_epsilon = 0.2
tp.final_epsilon = 1./(288.)
tp.step_for_final_epsilon = int(1e5)
# TODO add the "i dont do anything for a few time steps at the beginning of the training"
# don't start always at the same hour (if not None) otherwise random sampling, see docs
tp.random_sample_datetime_start = None
# saving, logging etc.
tp.save_model_each = 10000
tp.update_tensorboard_freq = 256
# which actions i keep
if env.name == "l2rpn_case14_sandbox":
kwargs_converters = {"all_actions": None,
"set_line_status": False,
"change_line_status": True,
# limit the number of time steps played per scenarios
tp.step_increase_nb_iter = 100 # None to deactivate it
tp.min_iter = 10
tp.update_nb_iter = 100 # once 100 scenarios are solved, increase of "step_increase_nb_iter"
# oversampling hard scenarios
tp.oversampling_rate = 3 # None to deactivate it
# experience replay
tp.buffer_size = 1000000
# e greedy
tp.min_observation = 10000
tp.initial_epsilon = 0.2
tp.final_epsilon = 1. / (7 * 288.)
tp.step_for_final_epsilon = int(1e5)
# don't start always at the same hour (if not None) otherwise random sampling, see docs
tp.random_sample_datetime_start = None
# saving, logging etc.
tp.save_model_each = 10000
tp.update_tensorboard_freq = 256
# which actions i keep
kwargs_converters = {
"all_actions": None,
"set_line_status": False,
"change_bus_vect": True,
"set_topo_vect": False