pretrain_imitation.py [19:87]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
)
from exploring_exploration.utils.common import process_image
from exploring_exploration.models import RGBEncoder, MapRGBEncoder, Policy
from exploring_exploration.utils.eval import evaluate_visitation
from exploring_exploration.utils.storage import RolloutStorageImitation
from exploring_exploration.algo import Imitation
from tensorboardX import SummaryWriter
from collections import defaultdict, deque

args = get_args()

num_updates = (args.num_episodes // args.num_processes) + 1

torch.manual_seed(args.seed)
if args.cuda:
    torch.cuda.manual_seed(args.seed)

try:
    os.makedirs(args.log_dir)
except OSError:
    pass

eval_log_dir = os.path.join(args.log_dir, "eval_monitor")

try:
    os.makedirs(eval_log_dir)
except OSError:
    pass


def main():
    torch.set_num_threads(1)
    device = torch.device("cuda:0" if args.cuda else "cpu")
    ndevices = torch.cuda.device_count()
    # Setup loggers
    tbwriter = SummaryWriter(log_dir=args.log_dir)
    logging.basicConfig(filename=f"{args.log_dir}/train_log.txt", level=logging.DEBUG)
    logging.getLogger().addHandler(logging.StreamHandler(sys.stdout))
    logging.getLogger().setLevel(logging.INFO)
    if "habitat" in args.env_name:
        devices = [int(dev) for dev in os.environ["CUDA_VISIBLE_DEVICES"].split(",")]
        # Devices need to be indexed between 0 to N-1
        devices = [dev for dev in range(len(devices))]
        if len(devices) > 2:
            devices = devices[1:]
        envs = make_vec_envs_habitat(
            args.habitat_config_file, device, devices, seed=args.seed
        )
    else:
        train_log_dir = os.path.join(args.log_dir, "train_monitor")
        try:
            os.makedirs(train_log_dir)
        except OSError:
            pass
        envs = make_vec_envs_avd(
            args.env_name,
            args.seed,
            args.num_processes,
            train_log_dir,
            device,
            True,
            num_frame_stack=1,
            split="train",
            nRef=args.num_pose_refs,
        )

    args.feat_shape_sim = (512,)
    args.feat_shape_pose = (512 * 9,)
    args.obs_shape = envs.observation_space.spaces["im"].shape
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



train_curiosity_exploration.py [30:96]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
)
from exploring_exploration.utils.eval import evaluate_visitation
from exploring_exploration.utils.storage import RolloutStoragePPO
from exploring_exploration.algo import PPO
from tensorboardX import SummaryWriter
from collections import defaultdict, deque

args = get_args()

num_updates = (args.num_episodes // args.num_processes) + 1

torch.manual_seed(args.seed)
if args.cuda:
    torch.cuda.manual_seed(args.seed)

try:
    os.makedirs(args.log_dir)
except OSError:
    pass

eval_log_dir = os.path.join(args.log_dir, "eval_monitor")

try:
    os.makedirs(eval_log_dir)
except OSError:
    pass


def main():
    torch.set_num_threads(1)
    device = torch.device("cuda:0" if args.cuda else "cpu")
    ndevices = torch.cuda.device_count()
    # Setup loggers
    tbwriter = SummaryWriter(log_dir=args.log_dir)
    logging.basicConfig(filename=f"{args.log_dir}/train_log.txt", level=logging.DEBUG)
    logging.getLogger().addHandler(logging.StreamHandler(sys.stdout))
    logging.getLogger().setLevel(logging.INFO)
    if "habitat" in args.env_name:
        devices = [int(dev) for dev in os.environ["CUDA_VISIBLE_DEVICES"].split(",")]
        # Devices need to be indexed between 0 to N-1
        devices = [dev for dev in range(len(devices))]
        if len(devices) > 2:
            devices = devices[1:]
        envs = make_vec_envs_habitat(
            args.habitat_config_file, device, devices, seed=args.seed
        )
    else:
        train_log_dir = os.path.join(args.log_dir, "train_monitor")
        try:
            os.makedirs(train_log_dir)
        except OSError:
            pass
        envs = make_vec_envs_avd(
            args.env_name,
            args.seed,
            args.num_processes,
            train_log_dir,
            device,
            True,
            num_frame_stack=1,
            split="train",
            nRef=args.num_pose_refs,
        )

    args.feat_shape_sim = (512,)
    args.feat_shape_pose = (512 * 9,)
    args.obs_shape = envs.observation_space.spaces["im"].shape
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -