def make_env(batch_size=8):
"""Creates the env."""
# No resizing needed, so let's be on the normal EnvProblem.
if not FLAGS.resize: # None or False
return env_problem.EnvProblem(base_env_name=FLAGS.env_problem_name,
batch_size=batch_size,
reward_range=(-1, 1))
max_timestep = None
try:
max_timestep = int(FLAGS.max_timestep)
except Exception: # pylint: disable=broad-except
pass
wrapper_fn = functools.partial(
gym_utils.gym_env_wrapper, **{
"rl_env_max_episode_steps": max_timestep,
"maxskip_env": True,
"rendered_env": True,
"rendered_env_resize_to":
(FLAGS.resized_height, FLAGS.resized_width),
"sticky_actions": False,
"output_dtype": onp.int32 if FLAGS.use_tpu else None,
})
return rendered_env_problem.RenderedEnvProblem(
base_env_name=FLAGS.env_problem_name,
batch_size=batch_size,
env_wrapper_fn=wrapper_fn,
reward_range=(-1, 1))
def make_env(name, batch_size, max_timestep, clip_rewards, rendered_env,
resize_dims, **env_kwargs):
"""Creates the env."""
if clip_rewards:
env_kwargs.update({"reward_range": (-1, 1), "discrete_rewards": True})
else:
env_kwargs.update({"discrete_rewards": False})
# TODO(afrozm): Should we leave out some cores?
parallelism = multiprocessing.cpu_count() if FLAGS.parallelize_envs else 1
# No resizing needed, so let's be on the normal EnvProblem.
if not rendered_env:
return gym_env_problem.GymEnvProblem(base_env_name=name,
batch_size=batch_size,
parallelism=parallelism,
**env_kwargs)
wrapper_fn = functools.partial(
gym_utils.gym_env_wrapper, **{
"rl_env_max_episode_steps": max_timestep,
"maxskip_env": True,
"rendered_env": True,
"rendered_env_resize_to": resize_dims,
"sticky_actions": False,
"output_dtype": onp.int32 if FLAGS.use_tpu else None,
})
return rendered_env_problem.RenderedEnvProblem(base_env_name=name,
batch_size=batch_size,
parallelism=parallelism,
env_wrapper_fn=wrapper_fn,
**env_kwargs)
def make_env():
"""Creates the env."""
if FLAGS.env_name:
return gym.make(FLAGS.env_name)
assert FLAGS.env_problem_name
# No resizing needed, so let's be on the normal EnvProblem.
if not FLAGS.resize: # None or False
return env_problem.EnvProblem(base_env_name=FLAGS.env_problem_name,
batch_size=FLAGS.batch_size,
reward_range=(-1, 1))
wrapper_fn = functools.partial(
gym_utils.gym_env_wrapper, **{
"rl_env_max_episode_steps": FLAGS.max_timestep,
"maxskip_env": True,
"rendered_env": True,
"rendered_env_resize_to":
(FLAGS.resized_height, FLAGS.resized_width),
"sticky_actions": False
})
return rendered_env_problem.RenderedEnvProblem(
base_env_name=FLAGS.env_problem_name,
batch_size=FLAGS.batch_size,
env_wrapper_fn=wrapper_fn,
reward_range=(-1, 1))
def make_env(batch_size=1,
env_problem_name="",
resize=True,
resize_dims=(105, 80),
max_timestep="None",
clip_rewards=True,
parallelism=1,
use_tpu=False,
num_actions=None,
rendered_env=True,
**env_kwargs):
"""Creates the env."""
if clip_rewards:
env_kwargs.update({"reward_range": (-1, 1), "discrete_rewards": True})
else:
env_kwargs.update({"discrete_rewards": False})
# TODO(henrykm) - below someone linked "resize" with "abnormality"
# Probably we need more nuanced concept of "abnormality"
# decoupled from "resize". Currently the resize flag implies
# that we switch from a generic env to a wrapped env.
# Overall this file and gym_utils.py look like good candidates
# for a refactor.
# No resizing needed, so let's be on the normal EnvProblem.
if not resize: # None or False
return gym_env_problem.GymEnvProblem(base_env_name=env_problem_name,
batch_size=batch_size,
parallelism=parallelism,
**env_kwargs)
try:
max_timestep = int(max_timestep)
except Exception: # pylint: disable=broad-except
max_timestep = None
wrapper_fn = functools.partial(
gym_utils.gym_env_wrapper, **{
"rl_env_max_episode_steps": max_timestep,
"maxskip_env": True,
"rendered_env": rendered_env,
"rendered_env_resize_to": resize_dims,
"sticky_actions": False,
"output_dtype": np.int32 if use_tpu else None,
"num_actions": num_actions,
})
return rendered_env_problem.RenderedEnvProblem(
base_env_name=env_problem_name,
batch_size=batch_size,
parallelism=parallelism,
rendered_env=rendered_env,
env_wrapper_fn=wrapper_fn,
**env_kwargs)
def make_env(batch_size=1,
env_problem_name="",
resize=True,
resized_height=105,
resized_width=80,
max_timestep="None",
clip_rewards=True,
parallelism=1,
use_tpu=False,
**env_kwargs):
"""Creates the env."""
if clip_rewards:
env_kwargs.update({"reward_range": (-1, 1), "discrete_rewards": True})
else:
env_kwargs.update({"discrete_rewards": False})
# No resizing needed, so let's be on the normal EnvProblem.
if not resize: # None or False
return gym_env_problem.GymEnvProblem(
base_env_name=env_problem_name,
batch_size=batch_size,
parallelism=parallelism,
**env_kwargs)
try:
max_timestep = int(max_timestep)
except Exception: # pylint: disable=broad-except
max_timestep = None
wrapper_fn = functools.partial(
gym_utils.gym_env_wrapper, **{
"rl_env_max_episode_steps": max_timestep,
"maxskip_env": True,
"rendered_env": True,
"rendered_env_resize_to": (resized_height, resized_width),
"sticky_actions": False,
"output_dtype": np.int32 if use_tpu else None,
})
return rendered_env_problem.RenderedEnvProblem(
base_env_name=env_problem_name,
batch_size=batch_size,
parallelism=parallelism,
env_wrapper_fn=wrapper_fn,
**env_kwargs)
def make_env(batch_size=8, **env_kwargs):
"""Creates the env."""
if FLAGS.clip_rewards:
env_kwargs.update({"reward_range": (-1, 1), "discrete_rewards": True})
else:
env_kwargs.update({"discrete_rewards": False})
# TODO(afrozm): Should we leave out some cores?
parallelism = multiprocessing.cpu_count() if FLAGS.parallelize_envs else 1
# No resizing needed, so let's be on the normal EnvProblem.
if not FLAGS.resize: # None or False
return gym_env_problem.GymEnvProblem(
base_env_name=FLAGS.env_problem_name,
batch_size=batch_size,
parallelism=parallelism,
**env_kwargs)
max_timestep = None
try:
max_timestep = int(FLAGS.max_timestep)
except Exception: # pylint: disable=broad-except
pass
wrapper_fn = functools.partial(
gym_utils.gym_env_wrapper, **{
"rl_env_max_episode_steps": max_timestep,
"maxskip_env": True,
"rendered_env": True,
"rendered_env_resize_to":
(FLAGS.resized_height, FLAGS.resized_width),
"sticky_actions": False,
"output_dtype": onp.int32 if FLAGS.use_tpu else None,
})
return rendered_env_problem.RenderedEnvProblem(
base_env_name=FLAGS.env_problem_name,
batch_size=batch_size,
parallelism=parallelism,
env_wrapper_fn=wrapper_fn,
**env_kwargs)
