train_submission_code.py

import logging
import os

import numpy as np
import aicrowd_helper
import gym
import minerl
from utility.parser import Parser
from basalt_baselines.bc import bc_baseline

from kairos_minerl import state_classifier
from kairos_minerl import behavior_cloner
import kairos_minerl.data_processing
import torch as th


import coloredlogs
coloredlogs.install(logging.DEBUG)


# You need to ensure that your submission is trained by launching less than MINERL_TRAINING_MAX_INSTANCES instances
MINERL_TRAINING_MAX_INSTANCES = int(os.getenv('MINERL_TRAINING_MAX_INSTANCES', 5))
# The dataset is available in data/ directory from repository root.
MINERL_DATA_ROOT = os.getenv('MINERL_DATA_ROOT', 'data/')
# You need to ensure that your submission is trained within allowed training time.
MINERL_TRAINING_TIMEOUT = int(os.getenv('MINERL_TRAINING_TIMEOUT_MINUTES', 4 * 24 * 60))
# You need to ensure that your submission is trained by launching less than MINERL_TRAINING_MAX_INSTANCES instances
MINERL_TRAINING_MAX_INSTANCES = int(os.getenv('MINERL_TRAINING_MAX_INSTANCES', 5))

BASALT_GYM_ENV = os.getenv('MINERL_GYM_ENV', 'MineRLBasaltFindCave-v0')

PREPROCESS_DATASET_AND_RETRAIN = True
TRAIN_BASELINES = True


# Optional: You can view best effort status of your instances with the help of parser.py
# This will give you current state like number of steps completed, instances launched and so on.
# Make your you keep a tap on the numbers to avoid breaching any limits.
parser = Parser(
    'performance/',
    maximum_instances=MINERL_TRAINING_MAX_INSTANCES,
    raise_on_error=False,
    no_entry_poll_timeout=600,
    submission_timeout=MINERL_TRAINING_TIMEOUT * 60,
    initial_poll_timeout=600
)


def basic_train():
    """
    This function will be called for training phase.
    This should produce and save same files you upload during your submission.
    """
    # How to sample minerl data is document here:
    # http://minerl.io/docs/tutorials/data_sampling.html
    data = minerl.data.make('MineRLBasaltFindCave-v0', data_dir=MINERL_DATA_ROOT)

    # Sample code for illustration, add your training code below
    env = gym.make('MineRLBasaltFindCave-v0')

    # For an example, lets just run one episode of MineRL for training
    obs = env.reset()
    done = False
    while not done:
        obs, reward, done, info = env.step(env.action_space.sample())
        # Do your training here

        # To get better view in your training phase, it is suggested
        # to register progress continuously, example when 54% completed
        # aicrowd_helper.register_progress(0.54)

        # To fetch latest information from instance manager, you can run below when you want to know the state
        #>> parser.update_information()
        #>> print(parser.payload)

    # Save trained model to train/ directory
    # For a demonstration, we save some dummy data.
    np.save("./train/parameters.npy", np.random.random((10,)))

    # Training 100% Completed
    aicrowd_helper.register_progress(1)
    env.close()


def main():
    if PREPROCESS_DATASET_AND_RETRAIN:
        # # preprocess original dataset in the format we need to train our classifiers and models
        # kairos_minerl.data_processing.extract_dataset()
        
        # # train state classifier
        # state_classifier.train()

        # # train behavior cloner navigation policy for all tasks
        # behavior_cloner.train(model_type="find_cave", experiment_name="bc_model_navigation_FINAL")
        # behavior_cloner.train(model_type="animal_pen", experiment_name="bc_model_navigation_FINAL")
        # behavior_cloner.train(model_type="make_waterfall", experiment_name="bc_model_navigation_FINAL")
        # behavior_cloner.train(model_type="village_house", experiment_name="bc_model_navigation_FINAL")

        # train end-to-end behavior clonerfor all tasks (baseline only)
        if TRAIN_BASELINES:
            behavior_cloner.train(model_type="find_cave", navigation=False, num_epochs = 20, experiment_name="bc_baseline")
            # behavior_cloner.train(model_type="animal_pen", navigation=False, experiment_name="bc_baseline")
            # behavior_cloner.train(model_type="make_waterfall", navigation=False, experiment_name="bc_baseline")
            # behavior_cloner.train(model_type="village_house", navigation=False, experiment_name="bc_baseline")



if __name__ == "__main__":
    main()