# Stepsize of SGD.
"lr": 0.0005,
# PPO clip parameter.
"clip_param": 0.03,
# Clip param for the value function. Note that this is sensitive to the
# scale of the rewards. If your expected V is large, increase this.
"vf_clip_param": 10.0,
# If specified, clip the global norm of gradients by this amount.
"grad_clip": None,
# Target value for KL divergence.
"kl_target": 0.001,
}
tune.run(
run_or_experiment=trainer_class,
name=experiment_name,
metric="br_reward_mean",
config=hyperparams,
num_samples=2,
search_alg=None,
mode="max",
local_dir=data_dir(),
stop={"timesteps_total": int(3e6)},
loggers=[
get_trainer_logger_creator(base_dir=data_dir(),
scenario_name=experiment_name,
should_log_result_fn=lambda result:
result["training_iteration"] % 20 == 0)
],
)
# if async_updates is set, then each worker returns gradients for a
# batch of this size.
"train_batch_size": choice([4096, 2048, 1024]),
# Whether to compute priorities on workers.
"worker_side_prioritization": False,
# Prevent iterations from going lower than this time span
"min_iter_time_s": 0,
# Minimum env steps to optimize for per train call. This value does
# not affect learning (JB: this is a lie!), only the length of train iterations.
"timesteps_per_iteration": 0,
}
search = HyperOptSearch(metric="br_reward_mean", mode="max", n_initial_points=20)
tune.run(run_or_experiment=trainer_class,
name=experiment_name,
metric="br_reward_mean",
config=hyperparams,
num_samples=200000000,
search_alg=search,
mode="max",
local_dir=data_dir(),
stop={"timesteps_total": int(3e6)},
loggers=[get_trainer_logger_creator(
base_dir=data_dir(),
scenario_name=experiment_name,
should_log_result_fn=lambda result: result["training_iteration"] % 20 == 0)],
)