def main(argv):
try:
if (len(argv) < 9):
eprint("usage: PythonRunner test_type module_path input_filepath " +\
"output_dir minimum_measurable_time nruns_F nruns_J time_limit [-rep]\n")
return 1
test_type = argv[1]
module_path = os.path.normpath(argv[2])
input_filepath = os.path.normpath(argv[3])
output_prefix = os.path.normpath(argv[4])
minimum_measurable_time = float(argv[5])
nruns_F = int(argv[6])
nruns_J = int(argv[7])
time_limit = float(argv[8])
# read only 1 point and replicate it?
replicate_point = (len(argv) > 9 and str(argv[9]) == "-rep")
# If the given prefix is a directory then add a separator to its end
# thus we can just use concatenation further
if os.path.isdir(output_prefix):
output_prefix += os.path.sep
if test_type == "GMM":
# read gmm input
_input = input_utils.read_gmm_instance(input_filepath, replicate_point)
elif test_type == "BA":
# read ba input
_input = input_utils.read_ba_instance(input_filepath)
elif test_type == "HAND":
model_dir = os.path.join(filepath_to_dirname(input_filepath), "model")
# read hand input
_input = input_utils.read_hand_instance(model_dir, input_filepath, False)
elif test_type == "HAND-COMPLICATED":
model_dir = os.path.join(filepath_to_dirname(input_filepath), "model")
# read hand complicated input
_input = input_utils.read_hand_instance(model_dir, input_filepath, True)
elif test_type == "LSTM":
_input = input_utils.read_lstm_instance(input_filepath)
else:
raise RuntimeError("Python runner doesn't support tests of " + test_type + " type")
run_benchmark(
module_path,
input_filepath,
_input,
output_prefix,
minimum_measurable_time,
nruns_F,
nruns_J,
time_limit
)
except RuntimeError as ex:
eprint("Runtime exception caught: ", ex)
except Exception as ex:
eprint("An exception caught: ", ex)
return 0
def test_jacobian_runs_multiple_times(self):
'''Checks if jacobian can be calculated multiple times.'''
input = read_lstm_instance(TEST_INPUT_FILE_NAME)
self.test.prepare(input)
func = self.test.calculate_jacobian
self.assertTrue(utils.can_objective_run_multiple_times(func))
def objective_calculation_correctness(self, times):
'''Checks objective calculation correctness running calculation
several times.'''
input = read_lstm_instance(TEST_INPUT_FILE_NAME)
self.test.prepare(input)
self.test.calculate_objective(times)
output = self.test.output()
self.assertFloatEqual(0.66666518, output.objective,
self.params["tolerance"])
def jacobian_calculation_correctness(self, times):
'''Checks jacobian calculation correctness running calculation
several times.'''
input = read_lstm_instance(TEST_INPUT_FILE_NAME)
self.test.prepare(input)
self.test.calculate_jacobian(times)
output = self.test.output()
expected_gradient = [
3.14387024996101187e-06, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
-3.67840807336070475e-05, -1.54848612683180522e-04,
5.48717267869773910e-06, -4.04527413533285253e-06,
-8.77353908686236070e-07, -1.81973718092835457e-06,
-6.61404620766666418e-07, 1.25432314145999742e-04,
5.26459549995064566e-06, 3.42833197850334681e-06,
2.79076684010150125e-05, 4.45603055235501376e-05,
4.04185405141412220e-05, 1.13487138666022078e-05,
-7.36119414163935681e-05, -3.92318596883106641e-04,
6.03607595779890245e-06, -2.50749024433336479e-05,
-4.83955404587018941e-06, -4.94175793354897013e-06,
-1.49848451794559008e-06, 4.01789457601843220e-06, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, -6.16501771215585166e-05, -2.14471058586458561e-04,
5.56661202545761896e-06, -5.22021372686648383e-06,
-1.95401365542668346e-06, -4.68776054707821880e-06,
-2.67879854428802963e-06, 8.49896476968742872e-04,
2.14480696808520274e-04, 1.27240956373907973e-04,
1.03380278076147588e-04, 3.26983387274766885e-05,
1.47321495437521366e-04, 2.03767485364521309e-05,
-1.71027241883012373e-04, -9.97102906947628905e-04,
1.05573445256271496e-04, -4.10813160791978987e-05,
-4.66363145389210376e-06, -5.61458724856520127e-06,
-1.72719535440704236e-06, 7.41562400488615767e-04,
1.35905892236421730e-04, 9.18319427682989299e-05,
1.14039861417365726e-04, 1.31037490514354042e-04,
1.89958003915723621e-04, 6.66667089111214168e-05,
-2.99753093683376040e-04, -1.39428488543169235e-03,
7.81997486731591678e-05, -7.11003463557882681e-05,
-1.56689414959509608e-05, -1.59132673852719420e-05,
-5.88701856095873723e-06, 3.65214776221949554e-04,
1.08440867176087962e-04, 6.75251197117822909e-05,
5.07904844905876160e-05, 6.94964603070496247e-05,
9.71513382958139477e-05, 2.40897410354102865e-05,
-1.31891358419510490e-04, -7.36772740642002493e-04,
4.48757608883305795e-05, -4.51743229379305454e-05,
-7.63715831354149397e-06, -8.36900345795258848e-06,
-2.41858043904271068e-06, 8.00930878140701909e-04,
8.05738754618121506e-05, 7.04205639535801850e-05,
1.34615500556859374e-04, 2.07383290113708586e-04,
2.94203848388590759e-04, 2.08296283404346797e-04,
-5.00619564307732703e-04, -1.90384053275097134e-03,
6.83889421204583172e-05, -8.85415990620916452e-05,
-4.01278522469193820e-05, -3.19293895829058307e-05,
-2.48575295654621768e-05, 2.47822472805507175e-03,
5.07502578529807780e-03, 2.53859920727141910e-03,
1.88350319998056696e-04, 5.10157230919509725e-05,
3.54335963584264193e-04, 4.32702650224108475e-05,
-3.06537980830570744e-04, -1.87256106103047193e-03,
7.86527299400905083e-04, -7.40280147056494078e-05,
-7.36222276800657044e-06, -9.51938152613299431e-06,
-2.78936845972344488e-06, 5.38571851736894072e-04,
4.12509633131798620e-05, 6.47788903296268593e-05,
3.76860255482900047e-04, 1.00145332877384172e-03,
2.29657803047558681e-04, 4.44627297218581029e-04,
-9.73952035222110210e-04, -4.51211367568013846e-03,
1.41925554887538418e-04, -1.66675086094364628e-05,
-2.28915871203058926e-05, -6.90861176975165640e-05,
-2.81513790989548166e-04, 2.53531909024735675e-04,
2.91530880231834222e-04, 2.87839568382086426e-04,
6.61417346851010207e-05, 6.11690024226054017e-04,
1.08165251111585849e-04, 9.94833840089197921e-05,
-5.39732825546851164e-04, -1.95313692267145384e-03,
2.87262803065020395e-04, -6.42028361615228470e-06,
-6.81225636233755266e-06, -3.58438396937832646e-05,
-1.41607977576401395e-04, 9.90063334626659942e-04,
6.71457692678222891e-05, 9.64171460882922153e-05,
2.83657185457934097e-04, 1.41431021522050158e-03,
7.42601545855336369e-04, 1.63135992873617838e-04,
-1.80905573085369175e-03, -5.78938583485935632e-03,
1.08933179139588087e-04, -2.82053334001221257e-04,
-1.58448932394055300e-04, -1.62053197372598410e-04,
-4.44062656277905885e-04, 1.23784405389406552e-03,
3.70797204635395672e-04, 8.95872992208362269e-04,
9.54650103551505725e-04, 7.94492683033071450e-04,
1.33045599071493839e-04, 1.90972131022078425e-03,
-1.27723370300947075e-03, -7.98761497515965146e-03,
1.31804902759735014e-03, -1.18749502523366873e-05,
-1.17265333639628955e-05, -5.72852084344175575e-05,
-1.76267731147033947e-04, 1.57111882911252715e-03,
9.63872765804848195e-04, 1.29692708705586861e-03,
6.85018812295237929e-04, 1.56084730076406770e-03,
5.51507203835322228e-04, 9.44664300904442429e-04,
-1.74595527620669054e-03, -8.47272040239691569e-03,
1.06146398161773513e-03, -3.01166484114288631e-05,
-3.61150963335128990e-05, -1.17040277318999298e-04,
-4.53714566010890040e-04, 4.83883598142451641e-04,
5.36532886336024261e-04, 7.10491025781043260e-04,
4.03370948422198858e-04, 9.21362107223384799e-04,
1.83544588672272390e-04, 5.84842906913008364e-04,
-9.12226698645542925e-04, -3.71928760349100754e-03,
5.24341732641706525e-04, -8.78104574502159225e-06,
-9.40356235117202346e-06, -6.16753351894278713e-05,
-2.59729129852274867e-04, 2.88643409216300618e-03,
1.56939967321338843e-03, 1.92928469338899343e-03,
5.15538874537588353e-04, 2.20418224227594915e-03,
1.78306235870001779e-03, 3.46590677977152962e-04,
-3.24323971790235277e-03, -1.08700028019174302e-02,
8.13601709951329772e-04, -5.09026696834261072e-04,
-2.50224828500900109e-04, -2.74229202243154427e-04,
-7.15862244954266883e-04, 2.36221822129175068e-03,
6.82455302403955932e-04, 2.21184971904590886e-03,
5.81458374769466737e-03, 1.19699412544996883e-03,
2.26008787507052335e-04, 1.12207195875873196e-02,
-2.15943584477549123e-03, -1.52105422913329149e-02,
2.40692059621100754e-03, -1.62908509623434489e-05,
-1.62307954324598644e-05, -9.93149651364052892e-05,
-3.23349485361925845e-04, 8.95056656284422921e-06, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, -1.73683965955792591e-04, -3.21979564391077671e-04,
7.71925547683623476e-06, -2.98807224102069401e-05,
-1.11738854533001905e-05, -2.92505411923477579e-06,
-2.04187281292674588e-06, 1.14042748169088035e-02,
9.93947389065250562e-03, 4.53876797372907911e-03,
1.46753585074672422e-03, 1.59578501610211192e-02,
7.13215996147348839e-03, 2.52120644391976836e-03,
-1.79990239825079217e-02, -2.62659614088943671e-02,
2.44293992080160194e-03, -3.67074714937877696e-03,
-1.31444481103537205e-02, -4.54999892021307108e-03,
-1.09270983790279372e-02, 2.17348662792538061e-02,
1.82857375496154884e-02, 1.11887441899222798e-02,
8.97672919990902263e-03, 2.39853893219980456e-02,
1.20800375241284126e-02, 1.48638788575309582e-02,
-3.02324690101714345e-02, -5.00116775406580991e-02,
4.46620942840666828e-03, -5.23602406001136191e-03,
-1.79316285235904226e-02, -7.73167106472800580e-03,
-2.00225690895980656e-02
]
self.assertFloatArrayEqual(expected_gradient, output.gradient,
self.params["tolerance"])