def get_test_lines(test_case, test_params):
"""
Create a list of strings corresponding to the lines in a single test case.
Uses TensorFlow to compute the expected results for the given parameters,
and provides the code to call the test fixture to run the test.
"""
output, max_input_val = get_result_and_size(test_case, test_params)
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(test_type=camel_case_type,
window=test_case.window,
stride=test_case.stride,
direction=helpers.to_camel_case(
test_case.direction))
test_name = TEST_NAME_TPL.format(padding=test_params.padding,
in_s=test_params.in_shape)
in_shape_init = IN_SHAPE_INIT_TPL.format(test_params.in_shape)
test_lines = [
"TYPED_TEST({}, {}) {{".format(test_case_name, test_name),
" using DataType = typename TestFixture::DataType;",
" const std::vector<DataType> exp_out = {};".format(
helpers.format_tensor(output)),
" const std::array<int, 4> in_shape = {};".format(in_shape_init),
" const auto padding = PaddingMode::{};".format(test_params.padding),
" const auto params = getPoolingParams<{}, {}>(in_shape, padding);".
format(test_case.window, test_case.stride),
" const DataType max_input_val = {:.1f};".format(max_input_val),
" this->test_pool(exp_out, params, max_input_val);",
"}",
]
return test_lines
def output_for_test_case(test_case):
"""
Create a list of strings corresponding to separate lines in the full test
case. The output contains headers, includes, setup and all the tests for
the test case.
"""
scriptname = os.path.basename(__file__)
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(
test_type=camel_case_type,
direction=helpers.to_camel_case(test_case.direction),
operation=helpers.to_camel_case(test_case.operation))
output = [
helpers.get_license(),
helpers.get_dont_modify_comment(scriptname=scriptname),
INCLUDES,
TYPED_TEST_CASE_DECL_TPL.format(
test_case=test_case_name,
direction=DIRECTION_MAP[test_case.direction],
operation=OPERATION_MAP[test_case.operation]),
]
for test_params in test_params_for_test_case(test_case):
output.extend(get_test_lines(test_case, test_params))
output.append("\n")
return output
def get_test_lines(test_case, test_params):
"""
Create a list of strings corresponding to the lines in a single test case.
Uses TensorFlow to compute the expected results for the given parameters,
and provides the code to call the test fixture to run the test.
"""
if test_case.direction == 'forward':
output, mean, variance, max_input_val = get_result(
test_case, test_params)
else:
output, mean, variance, max_input_val, grad_scale, grad_offset = get_result(
test_case, test_params)
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(
test_type=camel_case_type,
direction=helpers.to_camel_case(test_case.direction),
operation=helpers.to_camel_case(test_case.operation))
test_name = TEST_NAME_TPL.format(in_s=test_params.in_shape)
in_shape_init = IN_SHAPE_INIT_TPL.format(test_params.in_shape)
if test_case.direction == 'forward':
test_lines = [
"TYPED_TEST({}, {}) {{".format(test_case_name, test_name),
" using DataType = typename TestFixture::DataType;",
" const std::vector<DataType> exp_out = {};".format(
helpers.format_tensor(output)),
" const std::vector<DataType> mean = {};".format(
helpers.format_tensor(mean)),
" const std::vector<DataType> variance = {};".format(
helpers.format_tensor(variance)),
" const std::array<int, 4> in_shape = {};".format(in_shape_init),
" const auto params = getBatchNormParams(in_shape, DataFormat::{});"
.format(test_params.data_format),
" const DataType max_input_val = {:.1f};".format(max_input_val),
" this->test_batchnorm(exp_out, mean, variance, params, max_input_val);",
"}",
]
return test_lines
else:
test_lines = [
"TYPED_TEST({}, {}) {{".format(test_case_name, test_name),
" using DataType = typename TestFixture::DataType;",
" const std::vector<DataType> exp_grad = {};".format(
helpers.format_tensor(output)),
" const std::vector<DataType> mean = {};".format(
helpers.format_tensor(mean)),
" const std::vector<DataType> variance = {};".format(
helpers.format_tensor(variance)),
" const std::vector<DataType> grad_scale = {};".format(
helpers.format_tensor(grad_scale)),
" const std::vector<DataType> grad_offset = {};".format(
helpers.format_tensor(grad_offset)),
" const std::array<int, 4> in_shape = {};".format(in_shape_init),
" const auto params = getBatchNormParams(in_shape, DataFormat::{});"
.format(test_params.data_format),
" const DataType max_input_val = {:.1f};".format(max_input_val),
" this->test_batchnorm(exp_grad, mean, variance, grad_scale, grad_offset, params, max_input_val);",
"}",
]
return test_lines
def logger_from_dict(config, use_visdom=True):
"""
Create a logger from a dictionary
:param config: Dictionary containing the experiment hyper-parameters
:param use_visdom: Wether to use Visdom
:return: Logger
"""
experiment_name = datetime.datetime.now().strftime("%Y-%m-%d-")
ordered_config = collections.OrderedDict(sorted(config.items()))
for key, value in ordered_config.items():
key = helpers.to_camel_case(key.replace(' ', '_'))
if value is str:
value = helpers.to_camel_case(value.replace(' ', '_'))
if key == 'Transforms' or key == 'trainType':
continue
if value is not None and key is not None:
experiment_name += "{}_{}-".format(key, value)
# Delete last dash
experiment_name = experiment_name[:-1]
# Create logger
log = logger.Experiment(name=experiment_name,
use_visdom=use_visdom,
visdom_opts={
'server': 'http://localhost',
'port': 8097
},
time_indexing=False,
xlabel='Epoch')
log.log_config(config)
# create parent metric for training metrics (easier interface)
log.ParentWrapper(
tag='train',
name='parent',
children=[log.AvgMetric(name='loss'),
log.AvgMetric(name='acc')])
# same for validation metrics (note all children inherit tag from parent)
log.ParentWrapper(
tag='val',
name='parent',
children=[log.AvgMetric(name='loss'),
log.AvgMetric(name='acc')])
# Add a best metric for the validation accuracy
log.ParentWrapper(tag='best',
name='parent',
children=[log.BestMetric(name='acc')])
return log
def get_test_lines(test_case, test_params):
"""
Create a list of strings corresponding to the lines in a single test case.
Uses TensorFlow to compute the expected results for the given parameters,
and provides the code to call the test fixture to run the test.
"""
output = get_result(test_case, test_params)
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(test_type=camel_case_type,
direction=helpers.to_camel_case(
test_case.direction))
test_name = TEST_NAME_TPL.format(in_s=test_params.in_size)
test_lines = [
"TYPED_TEST({}, {}) {{".format(test_case_name, test_name),
" using DataType = typename TestFixture::DataType;",
" const std::vector<DataType> exp_out = {};".format(
helpers.format_tensor(output)),
" this->test_pointwise(exp_out);",
"}",
]
return test_lines
def output_for_test_case(test_case):
"""
Create a list of strings corresponding to separate lines in the full test
case. The output contains headers, includes, setup and all the tests for
the test case.
"""
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(test_type=camel_case_type,
window=test_case.window,
stride=test_case.stride,
direction=helpers.to_camel_case(
test_case.direction))
output = [
TYPED_TEST_SUITE_DECL_TPL.format(
test_case=test_case_name,
operation=OPERATOR_MAP[test_case.test_type],
direction=DIRECTION_MAP[test_case.direction])
]
for test_params in test_case.param_gen(test_case):
output.extend(get_test_lines(test_case, test_params))
output.append("\n")
return output
def output_for_test_case(test_case):
"""
Create a list of strings corresponding to separate lines in the full test
case. The output contains headers, includes, setup and all the tests for
the test case.
"""
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(test_type=camel_case_type)
output = [
TYPED_TEST_SUITE_DECL_TPL.format(
test_case=test_case_name)
]
for test_params in test_case.param_gen(test_case):
output.extend(get_test_lines(test_case, test_params))
output.append("\n")
return output
def output_for_test_case(test_case):
"""
Create a list of strings corresponding to separate lines in the full test
case. The output contains headers, includes, setup and all the tests for
the test case.
"""
scriptname = os.path.basename(__file__)
camel_case_type = helpers.to_camel_case(test_case.test_type)
test_case_name = TEST_CASE_TPL.format(test_type=camel_case_type,
window=test_case.window,
stride=test_case.stride)
output = [
helpers.get_license(),
helpers.get_dont_modify_comment(scriptname=scriptname), INCLUDES,
DATA_TYPES,
TYPED_TEST_SUITE_DECL_TPL.format(test_case=test_case_name,
window=test_case.window,
stride=test_case.stride)
]
for test_params in test_params_for_test_case(test_case):
output.extend(get_test_lines(test_case, test_params))
return output
