def test_train(self):
model_config = core.ModelConfig(_lineworld_name)
tc = core.PpoTrainContext()
ppo_agent = tfagents.TfPpoAgent(model_config=model_config)
ppo_agent.train(train_context=tc,
callbacks=[duration.Fast(),
log.Iteration()])
def test_train(self):
for backend in get_backends(SacAgent):
sac_agent: SacAgent = SacAgent('MountainCarContinuous-v0', backend=backend)
sac_agent.train([log.Duration(), log.Iteration(), log.Agent(), duration.Fast()],
num_iterations=10,
max_steps_per_episode=200,
default_plots=False)
def test_train(self):
model_config = core.ModelConfig("CartPole-v0")
tc = core.StepsTrainContext()
dqn_agent = tfagents.TfDqnAgent(model_config=model_config)
dqn_agent.train(train_context=tc,
callbacks=[duration.Fast(),
log.Iteration()])
def test_train(self):
model_config = core.ModelConfig(_lineworld_name)
tc = core.TrainContext()
random_agent = tfagents.TfRandomAgent(model_config=model_config)
random_agent.train(train_context=tc,
callbacks=[duration.Fast(),
log.Iteration()])
assert tc.episodes_done_in_iteration == 1
def test_reinforce_train(self):
model_config = core.ModelConfig("CartPole-v0")
tc = core.EpisodesTrainContext()
reinforceAgent = tforce.TforceReinforceAgent(model_config=model_config)
reinforceAgent.train(
train_context=tc,
callbacks=[log.Iteration(),
log.Agent(),
duration.Fast()])
def test_ppo_train(self):
model_config = core.ModelConfig("CartPole-v0")
tc = core.PpoTrainContext()
ppoAgent = tforce.TforcePpoAgent(model_config=model_config)
ppoAgent.train(
train_context=tc,
callbacks=[log.Iteration(),
log.Agent(),
duration.Fast()])
def test_train(self):
model_config = core.ModelConfig("CartPole-v0")
tc = core.EpisodesTrainContext()
reinforceAgent = tfagents.TfReinforceAgent(model_config=model_config)
reinforceAgent.train(train_context=tc, callbacks=[duration.Fast(), log.Iteration()])
assert tc.episodes_done_in_iteration == tc.num_episodes_per_iteration > 0
assert tc.iterations_done_in_training == tc.num_iterations > 0
rmin, ravg, rmax = tc.eval_rewards[tc.episodes_done_in_training]
assert rmax >= 10
def test_train(self):
model_config = core.ModelConfig(_lineworld_name)
tc = core.EpisodesTrainContext()
reinforce_agent = tfagents.TfReinforceAgent(model_config=model_config)
reinforce_agent.train(train_context=tc,
callbacks=[duration.Fast(),
log.Iteration()])
assert tc.episodes_done_in_iteration == tc.num_episodes_per_iteration > 0
assert tc.iterations_done_in_training == tc.num_iterations > 0
def test_train(self):
model_config = core.ModelConfig(_mountaincar_continuous_name)
tc = core.StepsTrainContext()
dqn_agent = tfagents.TfSacAgent(model_config=model_config)
dqn_agent.train(
train_context=tc,
callbacks=[duration.Fast(),
log.Iteration(),
log.Agent()])
def test_dqn(self):
easyagents.agents.seed = 0
model_config = core.ModelConfig("CartPole-v0")
tc = core.StepsTrainContext()
dqnAgent = kerasrl.KerasRlDqnAgent(model_config=model_config)
dqnAgent.train(train_context=tc,
callbacks=[
duration.Fast(),
log.Agent(),
log.Step(),
log.Iteration()
])
def test_log_agent(self):
agent = agents.PpoAgent(_step_count_name)
agent.train([log.Agent(), duration.Fast()])
def test_train_plotsteprewards(self):
ppoAgent = agents.PpoAgent('CartPole-v0')
ppoAgent.train([plot.StepRewards(), duration.Fast()])
def test_default_plots_None_durationcallback(self):
agent = agents.PpoAgent("CartPole-v0")
p = plot.Loss()
c = agent._add_plot_callbacks([duration.Fast()], None, [p])
assert p in c
def test_train(self):
for backend in get_backends(SacAgent):
sac_agent: SacAgent = SacAgent(_mountaincart_continuous_name, backend=backend)
tc: core.TrainContext = sac_agent.train([log.Duration(), log.Iteration(eval_only=True), duration.Fast()],
default_plots=False)
r = max_avg_rewards(tc)
assert r >= -1