def factory(Model, model_args, env_args, optim_args, memory_args, learner_args,
name):
model = Model(**model_args)
env = CartPole(**env_args)
optim = torch.optim.SGD(model.parameters(), **optim_args)
crit = torch.nn.MSELoss()
l_args = dict(learner_args)
l_args['name'] = f"{learner_args['name']}_{name}"
return pg.QLearner(env=env,
model=model,
optimizer=optim,
crit=crit,
action_selector=sp.QActionSelection(temperature=.3),
callbacks=[rcb.MemoryUpdater(1.)],
**l_args)
def factory(Model, model_args, env_args, optim_args, memory_args, learner_args,
crit_args, temp, name):
model = Model(**model_args)
env = TorchGym(env_args['env_name'])
optim = torch.optim.SGD(model.parameters(), **optim_args)
crit = torch.nn.MSELoss(**crit_args)
l_args = dict(learner_args)
l_args['name'] = f"{learner_args['name']}_{name}"
return rl.QLearner(env=env,
model=model,
optimizer=optim,
crit=crit,
action_selector=sp.QActionSelection(temperature=temp),
callbacks=[],
**l_args)
def get_selection_strategy(key, params):
if key == 'AdaptiveQSelection':
return sp.AdaptiveQActionSelectionEntropy(
post_pipeline=[hat.EntropyHat()], **params)
if key == 'QSelectionCertainty':
return sp.QActionSelectionCertainty(post_pipeline=[hat.EntropyHat()],
**params)
if key == 'QSelection':
return sp.QActionSelection(post_pipeline=[hat.EnsembleHat()], **params)
if key == 'EpsilonGreedy':
return sp.EpsilonGreedyActionSelection(
post_pipeline=[hat.EnsembleHat()], **params)
if key == 'AdaptiveEpsilonGreedy':
return sp.AdaptiveEpsilonGreedyActionSelection(
post_pipeline=[hat.EntropyHat()], **params)
if key == 'Greedy':
return sp.GreedyValueSelection(post_pipeline=[hat.EnsembleHat()],
**params)
raise ValueError('Unknown selection strategy')
def factory(Model, core, model_args, env_args, optim_args, memory_args,
learner_args, name):
model = Model(core, **model_args)
env = CartPole(**env_args)
memory_updater = MemoryUpdater(**memory_args)
optim = torch.optim.SGD(model.parameters(), **optim_args)
crit = torch.nn.MSELoss()
learner_args['name'] = name
return pg.QLearner(env=env,
model=model,
optimizer=optim,
memory_updater=memory_updater,
crit=crit,
action_selector=sp.QActionSelection(temperature=.3),
callbacks=[
cb.Checkpointer(),
rcb.EnvironmentEvaluator(env=env,
n_evaluations=10,
frequency=1),
],
**learner_args)
rcb.UncertaintyUpdater(),
rcb.EnvironmentEvaluator(
env=TorchGym(**params['factory_args']['env_args']),
n_evaluations=10,
action_selector=sp.GreedyValueSelection(
post_pipeline=[EnsembleHat()]),
metrics={
'det_val_reward_mean': np.mean,
'deter_val_reward_std': np.std
},
frequency=1,
epoch_name='det_val_epoch'),
rcb.EnvironmentEvaluator(
env=TorchGym(**params['factory_args']['env_args']),
n_evaluations=10,
action_selector=sp.QActionSelection(temperature=params['temp'],
post_pipeline=[EnsembleHat()]),
metrics={
'prob_val_reward_mean': np.mean,
'prob_val_reward_std': np.std
},
frequency=1,
epoch_name='prob_val_epoch'),
rcb.EnsembleRewardPlotter(
metrics={
'det_val_reward_mean': 'det_val_epoch',
'prob_val_reward_mean': 'prob_val_epoch',
}),
])
learner.fit(**params['fit'])