from runner import Runner from common.arguments import get_args from common.utils import make_env import numpy as np import random import torch if __name__ == '__main__': # get the params args = get_args() env, args = make_env(args) runner = Runner(args, env) if args.evaluate: returns = runner.evaluate() print('Average returns is', returns) else: runner.run()
import torch import pickle import matplotlib.pyplot as plt from common.plot import plot from common.train_test import train, test from common.arguments import get_args from common.data import save_model from common.data import load_data from common.data import load_model from algos.DQN import DQNAgent # ------------ environment ------------ env_id = "CartPole-v0" env = gym.make(env_id) config = get_args() seed = config.seed def seed_torch(seed): torch.manual_seed(seed) if torch.backends.cudnn.enabled: torch.backends.cudnn.benchmark = False torch.backends.cudnn.deterministic = True np.random.seed(seed) seed_torch(seed) env.seed(seed)
if _args.aicrowd_challenge: from aicrowd import utils_pytorch as pyu, aicrowd_helpers # Export the representation extractor path_to_saved = pyu.export_model(pyu.RepresentationExtractor(model.model.encoder, 'mean'), input_shape=(1, model.num_channels, model.image_size, model.image_size)) logging.info(f'A copy of the model saved in {path_to_saved}') if on_aicrowd_server: # AICrowd will handle the evaluation aicrowd_helpers.register_progress(1.0) aicrowd_helpers.submit() else: # Run evaluation locally # The local_evaluation is implemented by aicrowd in the global namespace, so importing it suffices. # todo: implement a modular version of local_evaluation # noinspection PyUnresolvedReferences from aicrowd import local_evaluation if __name__ == "__main__": _args = get_args(sys.argv[1:]) setup_logging(_args.verbose) initialize_seeds(_args.seed) # set the environment variables for dataset directory and name, and check if the root dataset directory exists. set_environment_variables(_args.dset_dir, _args.dset_name) assert os.path.exists(os.environ.get('DISENTANGLEMENT_LIB_DATA', '')), \ 'Root dataset directory does not exist at: \"{}\"'.format(_args.dset_dir) main(_args)
import common.arguments as arguments from common.utils import * check_exp_folder() args = arguments.get_args() seed = args.seed output_data_folder = get_output_data_folder() input_data_folder = get_input_data_folder() current_folder = get_current_folder()