torch, nn = try_import_torch()
def env_creator(env_name):
if env_name == "StocksEnv-v0":
from lib.environ import StocksEnv as env
else:
raise NotImplementedError
return env
# register the env
BARS_COUNT = 30
STOCKS = '/Users/user/Desktop/Market_Research/stock_data/stock_prices__min_train_NET.csv'
stock_data = {"NIO": data.load_relative(STOCKS)}
env = env_creator("StocksEnv-v0")
tune.register_env('myEnv', lambda config: env(stock_data, bars_count=BARS_COUNT, state_1d=False))
config_model = sac.DEFAULT_CONFIG.copy()
config_model["policy_model"] = sac.DEFAULT_CONFIG["policy_model"].copy()
config_model["env"] = "myEnv"
config_model["gamma"] = 1.0
config_model["no_done_at_end"] = True
config_model["tau"] = 3e-3
config_model["target_network_update_freq"] = 32
config_model["num_workers"] = 1 # Run locally.
config_model["twin_q"] = True
config_model["clip_actions"] = True
config_model["normalize_actions"] = True
config_model["learning_starts"] = 0
config_model["prioritized_replay"] = True
config_model["train_batch_size"] = 32
def env_creator(env_name):
if env_name == "StocksEnv-v0":
from lib.environ import StocksEnv as env
else:
raise NotImplementedError
return env
# register the env
BARS_COUNT = 30
STOCKS = '/Users/user/Desktop/Market_Research/stock_data/stock_prices__min_train_NET.csv'
stock_data = {"NIO": data.load_relative(STOCKS)}
env = env_creator("StocksEnv-v0")
tune.register_env(
'myEnv',
lambda config: env(stock_data, bars_count=BARS_COUNT, state_1d=False))
#ray.shutdown()
#ray.init(num_cpus=16, num_gpus=0, ignore_reinit_error=True)
ModelCatalog.register_custom_model("my_model", TorchCustomModel)
config_model = {
"env": "myEnv",
# Use GPUs iff `RLLIB_NUM_GPUS` env var set to > 0.
"num_gpus": int(os.environ.get("RLLIB_NUM_GPUS", "0")),
"model": {
"custom_model": "my_model",
"vf_share_layers": False,
},
"batch_mode": "truncate_episodes",
"sgd_minibatch_size": 32,