def test_multistart_monte_carlo_expected_improvement_optimization(self):
"""Check that multistart optimization (gradient descent) can find the optimum point to sample (using 2-EI)."""
numpy.random.seed(7858) # TODO(271): Monte Carlo only works for this seed
index = numpy.argmax(numpy.greater_equal(self.num_sampled_list, 20))
domain, gaussian_process = self.gp_test_environments[index]
max_num_steps = 75 # this is *too few* steps; we configure it this way so the test will run quickly
max_num_restarts = 5
num_steps_averaged = 50
gamma = 0.2
pre_mult = 1.5
max_relative_change = 1.0
tolerance = 3.0e-2 # really large tolerance b/c converging with monte-carlo (esp in Python) is expensive
gd_parameters = GradientDescentParameters(
max_num_steps,
max_num_restarts,
num_steps_averaged,
gamma,
pre_mult,
max_relative_change,
tolerance,
)
num_multistarts = 2
# Expand the domain so that we are definitely not doing constrained optimization
expanded_domain = TensorProductDomain([ClosedInterval(-4.0, 2.0)] * self.dim)
num_to_sample = 2
repeated_domain = RepeatedDomain(num_to_sample, expanded_domain)
num_mc_iterations = 10000
# Just any random point that won't be optimal
points_to_sample = repeated_domain.generate_random_point_in_domain()
ei_eval = ExpectedImprovement(gaussian_process, points_to_sample, num_mc_iterations=num_mc_iterations)
# Compute EI and its gradient for the sake of comparison
ei_initial = ei_eval.compute_expected_improvement(force_monte_carlo=True) # TODO(271) Monte Carlo only works for this seed
grad_ei_initial = ei_eval.compute_grad_expected_improvement()
ei_optimizer = GradientDescentOptimizer(repeated_domain, ei_eval, gd_parameters)
best_point = multistart_expected_improvement_optimization(ei_optimizer, num_multistarts, num_to_sample)
# Check that gradients are "small"
ei_eval.current_point = best_point
ei_final = ei_eval.compute_expected_improvement(force_monte_carlo=True) # TODO(271) Monte Carlo only works for this seed
grad_ei_final = ei_eval.compute_grad_expected_improvement()
self.assert_vector_within_relative(grad_ei_final, numpy.zeros(grad_ei_final.shape), tolerance)
# Check that output is in the domain
T.assert_equal(repeated_domain.check_point_inside(best_point), True)
# Since we didn't really converge to the optimal EI (too costly), do some other sanity checks
# EI should have improved
T.assert_gt(ei_final, ei_initial)
# grad EI should have improved
for index in numpy.ndindex(grad_ei_final.shape):
T.assert_lt(numpy.fabs(grad_ei_final[index]), numpy.fabs(grad_ei_initial[index]))
def test_big_fork(self):
"""Tests that we can fork a large number of processes, each of which
will wait for a few milliseconds, and return.
NOTE: currently fails if you bump 70 up to 200. We're going to fix this very soon.
"""
time_sleep_s = 0.2
test_time = self.run_big_fork_test(time_sleep_s, 70, 70, 3)
print("Big fork performance test: {0:.2f} s (nominal: {1:.2f} s)".format(
test_time, time_sleep_s))
T.assert_lt(test_time, time_sleep_s * 2)
def _assert_range(self, x, lower, upper):
assert_gt(x, lower)
assert_lt(x, upper)
def test__cmp__(self):
other_node = node.Node('mocalhost', 'mocal', self.ssh_options)
assert_lt(self.node, 'thename')
assert_lt(self.node, other_node)
def _test_end_to_end(self, args=()):
# read from STDIN, a local file, and a remote file
stdin = StringIO('foo\nbar\n')
local_input_path = os.path.join(self.tmp_dir, 'input')
with open(local_input_path, 'w') as local_input_file:
local_input_file.write('bar\nqux\n')
input_to_upload = os.path.join(self.tmp_dir, 'remote_input')
with open(input_to_upload, 'w') as input_to_upload_file:
input_to_upload_file.write('foo\n')
remote_input_path = 'hdfs:///data/foo'
check_call([self.hadoop_bin,
'fs', '-put', input_to_upload, remote_input_path])
# doesn't matter what the intermediate output is; just has to exist.
add_mock_hadoop_output([''])
add_mock_hadoop_output(['1\t"qux"\n2\t"bar"\n', '2\t"foo"\n5\tnull\n'])
mr_job = MRTwoStepJob(['-r', 'hadoop', '-v',
'--no-conf', '--hadoop-arg', '-libjar',
'--hadoop-arg', 'containsJars.jar'] + list(args)
+ ['-', local_input_path, remote_input_path]
+ ['--hadoop-input-format', 'FooFormat']
+ ['--hadoop-output-format', 'BarFormat']
+ ['--jobconf', 'x=y'])
mr_job.sandbox(stdin=stdin)
local_tmp_dir = None
results = []
# don't care that --hadoop-*-format is deprecated
with logger_disabled('mrjob.job'):
runner = mr_job.make_runner()
with runner as runner: # i.e. call cleanup when we're done
assert isinstance(runner, HadoopJobRunner)
runner.run()
for line in runner.stream_output():
key, value = mr_job.parse_output_line(line)
results.append((key, value))
local_tmp_dir = runner._get_local_tmp_dir()
# make sure cleanup hasn't happened yet
assert os.path.exists(local_tmp_dir)
assert any(runner.ls(runner.get_output_dir()))
# make sure we're writing to the correct path in HDFS
hdfs_root = os.environ['MOCK_HDFS_ROOT']
assert_equal(sorted(os.listdir(hdfs_root)), ['data', 'user'])
home_dir = os.path.join(hdfs_root, 'user', getpass.getuser())
assert_equal(os.listdir(home_dir), ['tmp'])
assert_equal(os.listdir(os.path.join(home_dir, 'tmp')), ['mrjob'])
assert_equal(runner._opts['hadoop_extra_args'],
['-libjar', 'containsJars.jar'])
# make sure mrjob.tar.gz is uploaded and in PYTHONPATH
assert runner._mrjob_tar_gz_path
mrjob_tar_gz_file_dicts = [
file_dict for file_dict in runner._files
if file_dict['path'] == runner._mrjob_tar_gz_path]
assert_equal(len(mrjob_tar_gz_file_dicts), 1)
mrjob_tar_gz_file_dict = mrjob_tar_gz_file_dicts[0]
assert mrjob_tar_gz_file_dict['name']
pythonpath = runner._get_cmdenv()['PYTHONPATH']
assert_in(mrjob_tar_gz_file_dict['name'],
pythonpath.split(':'))
assert_equal(sorted(results),
[(1, 'qux'), (2, 'bar'), (2, 'foo'), (5, None)])
# make sure we called hadoop the way we expected
with open(os.environ['MOCK_HADOOP_LOG']) as mock_log:
hadoop_cmd_args = [shlex.split(line) for line in mock_log]
jar_cmd_args = [args for args in hadoop_cmd_args
if args[:1] == ['jar']]
assert_equal(len(jar_cmd_args), 2)
step_0_args, step_1_args = jar_cmd_args
# check input/output format
assert_in('-inputformat', step_0_args)
assert_not_in('-outputformat', step_0_args)
assert_not_in('-inputformat', step_1_args)
assert_in('-outputformat', step_1_args)
# make sure -libjar extra arg comes before -mapper
for args in (step_0_args, step_1_args):
assert_in('-libjar', args)
assert_in('-mapper', args)
assert_lt(args.index('-libjar'), args.index('-mapper'))
# make sure -jobconf made it through
assert_in('-D', step_0_args)
# make sure cleanup happens
assert not os.path.exists(local_tmp_dir)
assert not any(runner.ls(runner.get_output_dir()))
def _test_end_to_end(self, args=()):
# read from STDIN, a local file, and a remote file
stdin = StringIO('foo\nbar\n')
local_input_path = os.path.join(self.tmp_dir, 'input')
with open(local_input_path, 'w') as local_input_file:
local_input_file.write('bar\nqux\n')
input_to_upload = os.path.join(self.tmp_dir, 'remote_input')
with open(input_to_upload, 'w') as input_to_upload_file:
input_to_upload_file.write('foo\n')
remote_input_path = 'hdfs:///data/foo'
check_call([
self.hadoop_bin, 'fs', '-put', input_to_upload, remote_input_path
])
# doesn't matter what the intermediate output is; just has to exist.
add_mock_hadoop_output([''])
add_mock_hadoop_output(['1\t"qux"\n2\t"bar"\n', '2\t"foo"\n5\tnull\n'])
mr_job = MRTwoStepJob([
'-r', 'hadoop', '-v', '--no-conf', '--hadoop-arg', '-libjar',
'--hadoop-arg', 'containsJars.jar'
] + list(args) + ['-', local_input_path, remote_input_path] +
['--hadoop-input-format', 'FooFormat'] +
['--hadoop-output-format', 'BarFormat'] +
['--jobconf', 'x=y'])
mr_job.sandbox(stdin=stdin)
local_tmp_dir = None
results = []
# don't care that --hadoop-*-format is deprecated
with logger_disabled('mrjob.job'):
runner = mr_job.make_runner()
with runner as runner: # i.e. call cleanup when we're done
assert isinstance(runner, HadoopJobRunner)
runner.run()
for line in runner.stream_output():
key, value = mr_job.parse_output_line(line)
results.append((key, value))
local_tmp_dir = runner._get_local_tmp_dir()
# make sure cleanup hasn't happened yet
assert os.path.exists(local_tmp_dir)
assert any(runner.ls(runner.get_output_dir()))
# make sure we're writing to the correct path in HDFS
hdfs_root = os.environ['MOCK_HDFS_ROOT']
assert_equal(sorted(os.listdir(hdfs_root)), ['data', 'user'])
home_dir = os.path.join(hdfs_root, 'user', getpass.getuser())
assert_equal(os.listdir(home_dir), ['tmp'])
assert_equal(os.listdir(os.path.join(home_dir, 'tmp')), ['mrjob'])
assert_equal(runner._opts['hadoop_extra_args'],
['-libjar', 'containsJars.jar'])
# make sure mrjob.tar.gz is uploaded and in PYTHONPATH
assert runner._mrjob_tar_gz_path
mrjob_tar_gz_file_dicts = [
file_dict for file_dict in runner._files
if file_dict['path'] == runner._mrjob_tar_gz_path
]
assert_equal(len(mrjob_tar_gz_file_dicts), 1)
mrjob_tar_gz_file_dict = mrjob_tar_gz_file_dicts[0]
assert mrjob_tar_gz_file_dict['name']
pythonpath = runner._get_cmdenv()['PYTHONPATH']
assert_in(mrjob_tar_gz_file_dict['name'], pythonpath.split(':'))
assert_equal(sorted(results), [(1, 'qux'), (2, 'bar'), (2, 'foo'),
(5, None)])
# make sure we called hadoop the way we expected
with open(os.environ['MOCK_HADOOP_LOG']) as mock_log:
hadoop_cmd_args = [shlex.split(line) for line in mock_log]
jar_cmd_args = [
args for args in hadoop_cmd_args if args[:1] == ['jar']
]
assert_equal(len(jar_cmd_args), 2)
step_0_args, step_1_args = jar_cmd_args
# check input/output format
assert_in('-inputformat', step_0_args)
assert_not_in('-outputformat', step_0_args)
assert_not_in('-inputformat', step_1_args)
assert_in('-outputformat', step_1_args)
# make sure -libjar extra arg comes before -mapper
for args in (step_0_args, step_1_args):
assert_in('-libjar', args)
assert_in('-mapper', args)
assert_lt(args.index('-libjar'), args.index('-mapper'))
# make sure -jobconf made it through
assert_in('-D', step_0_args)
# make sure cleanup happens
assert not os.path.exists(local_tmp_dir)
assert not any(runner.ls(runner.get_output_dir()))
def _assert_range(self, x, lower, upper):
assert_gt(x, lower)
assert_lt(x, upper)
def test_end_to_end(self):
# read from STDIN, a local file, and a remote file
stdin = StringIO("foo\nbar\n")
local_input_path = os.path.join(self.tmp_dir, "input")
with open(local_input_path, "w") as local_input_file:
local_input_file.write("bar\nqux\n")
input_to_upload = os.path.join(self.tmp_dir, "remote_input")
with open(input_to_upload, "w") as input_to_upload_file:
input_to_upload_file.write("foo\n")
remote_input_path = "hdfs:///data/foo"
check_call([self.hadoop_bin, "fs", "-put", input_to_upload, remote_input_path])
# doesn't matter what the intermediate output is; just has to exist.
add_mock_hadoop_output([""])
add_mock_hadoop_output(['1\t"qux"\n2\t"bar"\n', '2\t"foo"\n5\tnull\n'])
mr_job = MRTwoStepJob(
[
"-r",
"hadoop",
"-v",
"--no-conf",
"--hadoop-arg",
"-libjar",
"--hadoop-arg",
"containsJars.jar",
"-",
local_input_path,
remote_input_path,
"--hadoop-input-format",
"FooFormat",
"--hadoop-output-format",
"BarFormat",
]
)
mr_job.sandbox(stdin=stdin)
local_tmp_dir = None
results = []
with mr_job.make_runner() as runner:
assert isinstance(runner, HadoopJobRunner)
runner.run()
for line in runner.stream_output():
key, value = mr_job.parse_output_line(line)
results.append((key, value))
local_tmp_dir = runner._get_local_tmp_dir()
# make sure cleanup hasn't happened yet
assert os.path.exists(local_tmp_dir)
assert any(runner.ls(runner.get_output_dir()))
# make sure we're writing to the correct path in HDFS
hdfs_root = os.environ["MOCK_HDFS_ROOT"]
assert_equal(sorted(os.listdir(hdfs_root)), ["data", "user"])
home_dir = os.path.join(hdfs_root, "user", getpass.getuser())
assert_equal(os.listdir(home_dir), ["tmp"])
assert_equal(os.listdir(os.path.join(home_dir, "tmp")), ["mrjob"])
assert_equal(runner._opts["hadoop_extra_args"], ["-libjar", "containsJars.jar"])
# make sure mrjob.tar.gz is uploaded and in PYTHONPATH
assert runner._mrjob_tar_gz_path
mrjob_tar_gz_file_dicts = [
file_dict for file_dict in runner._files if file_dict["path"] == runner._mrjob_tar_gz_path
]
assert_equal(len(mrjob_tar_gz_file_dicts), 1)
mrjob_tar_gz_file_dict = mrjob_tar_gz_file_dicts[0]
assert mrjob_tar_gz_file_dict["name"]
pythonpath = runner._get_cmdenv()["PYTHONPATH"]
assert_in(mrjob_tar_gz_file_dict["name"], pythonpath.split(":"))
assert_equal(sorted(results), [(1, "qux"), (2, "bar"), (2, "foo"), (5, None)])
# make sure we called hadoop the way we expected
with open(os.environ["MOCK_HADOOP_LOG"]) as mock_log:
hadoop_cmd_args = [shlex.split(line) for line in mock_log]
jar_cmd_args = [args for args in hadoop_cmd_args if args[:1] == ["jar"]]
assert_equal(len(jar_cmd_args), 2)
step_0_args, step_1_args = jar_cmd_args
# check input/output format
assert_in("-inputformat", step_0_args)
assert_not_in("-outputformat", step_0_args)
assert_not_in("-inputformat", step_1_args)
assert_in("-outputformat", step_1_args)
# make sure -libjar extra arg comes before -mapper
for args in (step_0_args, step_1_args):
assert_in("-libjar", args)
assert_in("-mapper", args)
assert_lt(args.index("-libjar"), args.index("-mapper"))
# make sure cleanup happens
assert not os.path.exists(local_tmp_dir)
assert not any(runner.ls(runner.get_output_dir()))
