def test_on_task_switch_is_called_multi_task():
setting = MultiTaskSetting(
dataset="mnist",
nb_tasks=5,
# steps_per_task=100,
# max_steps=500,
# test_steps_per_task=100,
train_transforms=[],
test_transforms=[],
val_transforms=[],
)
method = DummyMethod()
results = setting.apply(method)
assert method.n_task_switches == 0
assert method.received_task_ids == []
assert method.received_while_training == []
def test_fit_and_on_task_switch_calls():
setting = ContinualRLSetting(
dataset=DummyEnvironment,
nb_tasks=5,
steps_per_task=100,
max_steps=500,
test_steps_per_task=100,
train_transforms=[],
test_transforms=[],
val_transforms=[],
)
method = DummyMethod()
results = setting.apply(method)
# == 30 task switches in total.
assert method.n_task_switches == 0
assert method.n_fit_calls == 1 # TODO: Add something like this.
assert not method.received_task_ids
assert not method.received_while_training
def test_on_task_switch_is_called_incremental_rl():
setting = IncrementalRLSetting(
dataset=DummyEnvironment,
nb_tasks=5,
steps_per_task=100,
max_steps=500,
test_steps_per_task=100,
train_transforms=[],
test_transforms=[],
val_transforms=[],
)
method = DummyMethod()
_ = setting.apply(method)
# 5 after learning task 0
# 5 after learning task 1
# 5 after learning task 2
# 5 after learning task 3
# 5 after learning task 4
# == 30 task switches in total.
assert method.n_task_switches == 30
assert method.received_task_ids == [
0,
*[None for _ in range(5)],
1,
*[None for _ in range(5)],
2,
*[None for _ in range(5)],
3,
*[None for _ in range(5)],
4,
*[None for _ in range(5)],
]
assert method.received_while_training == [
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
]
def test_metaworld_auto_task_schedule(
pass_env_id_instead_of_env_instance: bool):
""" Test that when passing just an env id from metaworld and a number of tasks,
the task schedule is created automatically.
"""
import metaworld
from metaworld import MetaWorldEnv
benchmark = metaworld.ML10() # Construct the benchmark, sampling tasks
env_name = "reach-v1"
env_type: Type[MetaWorldEnv] = benchmark.train_classes[env_name]
env = env_type()
# TODO: When not passing a nb_tasks, the number of available tasks for that env
# is used.
# setting = TaskIncrementalRLSetting(
# dataset=env_name if pass_env_id_instead_of_env_instance else env,
# steps_per_task=1000,
# )
# assert setting.nb_tasks == 50
# assert setting.steps_per_task == 1000
# assert sorted(setting.train_task_schedule.keys()) == list(range(0, 50_000, 1000))
# Test passing a number of tasks:
setting = TaskIncrementalRLSetting(
dataset=env_name if pass_env_id_instead_of_env_instance else env,
steps_per_task=1000,
nb_tasks=2,
test_steps_per_task=1000,
transforms=[],
)
assert setting.nb_tasks == 2
assert setting.steps_per_task == 1000
assert sorted(setting.train_task_schedule.keys()) == list(
range(0, 2000, 1000))
from sequoia.common.metrics.rl_metrics import EpisodeMetrics
method = DummyMethod()
results: IncrementalRLSetting.Results[EpisodeMetrics] = setting.apply(
method)
def test_on_task_switch_is_called_task_incremental_rl():
setting = IncrementalRLSetting(
dataset=DummyEnvironment,
nb_tasks=5,
steps_per_task=100,
test_steps_per_task=100,
max_steps=500,
train_transforms=[],
test_transforms=[],
val_transforms=[],
task_labels_at_test_time=True,
)
method = DummyMethod()
_ = setting.apply(method)
assert method.n_task_switches == 30
assert method.received_task_ids == [
0,
*list(range(5)),
1,
*list(range(5)),
2,
*list(range(5)),
3,
*list(range(5)),
4,
*list(range(5)),
]
assert method.received_while_training == [
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
]
def test_monsterkong(task_labels_at_test_time: bool, state: bool):
""" checks that the MonsterKong env works fine with monsterkong and state input. """
setting = IncrementalRLSetting(
dataset="monsterkong",
observe_state_directly=state,
nb_tasks=5,
steps_per_task=100,
test_steps_per_task=100,
train_transforms=[],
test_transforms=[],
val_transforms=[],
task_labels_at_test_time=task_labels_at_test_time,
max_episode_steps=10,
)
if state:
# State-based monsterkong: We observe a flattened version of the game state
# (20 x 20 grid + player cell and goal cell, IIRC.)
assert setting.observation_space.x == spaces.Box(
0, 292, (402, ), np.int16)
else:
assert setting.observation_space.x == Image(0, 255, (64, 64, 3),
np.uint8)
if task_labels_at_test_time:
assert setting.observation_space.task_labels == spaces.Discrete(5)
else:
assert setting.observation_space.task_labels == Sparse(
spaces.Discrete(5), sparsity=0.0)
assert setting.test_steps == 500
with setting.train_dataloader() as env:
obs = env.reset()
assert obs in setting.observation_space
method = DummyMethod()
_ = setting.apply(method)
assert method.n_task_switches == 30
if task_labels_at_test_time:
assert method.received_task_ids == [
0,
*list(range(5)),
1,
*list(range(5)),
2,
*list(range(5)),
3,
*list(range(5)),
4,
*list(range(5)),
]
else:
assert method.received_task_ids == [
0,
*[None for _ in range(5)],
1,
*[None for _ in range(5)],
2,
*[None for _ in range(5)],
3,
*[None for _ in range(5)],
4,
*[None for _ in range(5)],
]
assert method.received_while_training == [
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
]
def test_monsterkong_pixels(task_labels_at_test_time: bool):
""" checks that the MonsterKong env works fine with monsterkong and state input. """
setting = IncrementalRLSetting(
dataset="monsterkong",
observe_state_directly=False,
nb_tasks=5,
steps_per_task=100,
test_steps_per_task=100,
train_transforms=[],
test_transforms=[],
val_transforms=[],
task_labels_at_test_time=task_labels_at_test_time,
max_episode_steps=10,
)
assert setting.test_steps == 500
assert setting.observation_space.x == Image(0, 255, (64, 64, 3), np.uint8)
with setting.train_dataloader() as env:
obs = env.reset()
assert obs in setting.observation_space
method = DummyMethod()
results = setting.apply(method)
assert method.n_task_switches == 30
if task_labels_at_test_time:
assert method.received_task_ids == [
0,
*list(range(5)),
1,
*list(range(5)),
2,
*list(range(5)),
3,
*list(range(5)),
4,
*list(range(5)),
]
else:
assert method.received_task_ids == [
0,
*[None for _ in range(5)],
1,
*[None for _ in range(5)],
2,
*[None for _ in range(5)],
3,
*[None for _ in range(5)],
4,
*[None for _ in range(5)],
]
assert method.received_while_training == [
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
True,
*[False for _ in range(5)],
]