signature_inputs, signature_outputs,
tf.saved_model.signature_constants.CLASSIFY_METHOD_NAME)
# Save model.
builder = tf.saved_model.builder.SavedModelBuilder(export_dir)
builder.add_meta_graph_and_variables(
sess, [tf.saved_model.tag_constants.SERVING],
signature_def_map={
tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY:
signature_def
},
main_op=tf.tables_initializer(),
strip_default_attrs=True)
builder.save()
tf.logging.info('Model has been exported to %s' % export_dir)
return signature_def
def main(unused_argv):
if os.path.exists(FLAGS.export_dir):
raise ValueError('export_dir already exists: exporting will fail')
export_model(FLAGS.export_dir, FLAGS.export_for_serving, FLAGS.batch_size,
FLAGS.image_width, FLAGS.image_height)
if __name__ == '__main__':
flags.mark_flag_as_required('dataset_name')
flags.mark_flag_as_required('export_dir')
app.run(main)
def testStringRequired(self):
res = FLAGS.string_foo_required
self.assertEqual(res, "default_val")
FLAGS.string_foo_required = "bar"
self.assertEqual("bar", FLAGS.string_foo_required)
def testNoneStringRequired(self):
res = FLAGS.none_string_foo_required
self.assertEqual(res, "default_val")
FLAGS.none_string_foo_required = "bar"
self.assertEqual("bar", FLAGS.none_string_foo_required)
def main(_):
# unittest.main() tries to interpret the unknown flags, so use the
# direct functions instead.
runner = unittest.TextTestRunner()
itersuite = unittest.TestLoader().loadTestsFromTestCase(FlagsTest)
runner.run(itersuite)
if __name__ == "__main__":
# Test command lines
sys.argv.extend([
"--bool_a", "--nobool_negation", "--bool_c=True", "--bool_d=False",
"and_argument", "--none_string_foo_required=default_val"
])
flags.mark_flag_as_required('string_foo_required')
flags.mark_flags_as_required(['none_string_foo_required'])
app.run()
res = FLAGS.string_foo_required
self.assertEqual(res, "default_val")
FLAGS.string_foo_required = "bar"
self.assertEqual("bar", FLAGS.string_foo_required)
def testNoneStringRequired(self):
res = FLAGS.none_string_foo_required
self.assertEqual(res, "default_val")
FLAGS.none_string_foo_required = "bar"
self.assertEqual("bar", FLAGS.none_string_foo_required)
def main(_):
# unittest.main() tries to interpret the unknown flags, so use the
# direct functions instead.
runner = unittest.TextTestRunner()
itersuite = unittest.TestLoader().loadTestsFromTestCase(FlagsTest)
runner.run(itersuite)
if __name__ == "__main__":
# Test command lines
sys.argv.extend([
"--bool_a", "--nobool_negation", "--bool_c=True", "--bool_d=False",
"--none_string_foo_required=default_val",
"and_argument"
])
flags.mark_flag_as_required('string_foo_required')
flags.mark_flags_as_required(['none_string_foo_required'])
app.run()
tf.saved_model.CLASSIFY_METHOD_NAME)
# Save model.
builder = tf.compat.v1.saved_model.builder.SavedModelBuilder(export_dir)
builder.add_meta_graph_and_variables(
sess, [tf.saved_model.SERVING],
signature_def_map={
tf.saved_model.DEFAULT_SERVING_SIGNATURE_DEF_KEY:
signature_def
},
main_op=tf.compat.v1.tables_initializer(),
strip_default_attrs=True)
builder.save()
tf.compat.v1.logging.info('Model has been exported to %s' % export_dir)
return signature_def
def main(unused_argv):
if os.path.exists(FLAGS.export_dir):
raise ValueError('export_dir already exists: exporting will fail')
export_model(FLAGS.export_dir, FLAGS.export_for_serving, FLAGS.batch_size,
FLAGS.image_width, FLAGS.image_height)
if __name__ == '__main__':
#flags.mark_flag_as_required('dataset_name')
flags.mark_flag_as_required('export_dir')
app.run(main)
from absl import app
from tensorflow.compiler.mlir.tfr.python.composite import Composite
from tensorflow.compiler.mlir.tfr.python.op_reg_gen import gen_register_op
from tensorflow.compiler.mlir.tfr.python.tfr_gen import tfr_gen_from_module
from tensorflow.python.platform import flags
FLAGS = flags.FLAGS
flags.DEFINE_string(
'output', None,
'Path to write the genereated register op file and MLIR file.')
flags.DEFINE_bool('gen_register_op', True,
'Generate register op cc file or tfr mlir file.')
flags.mark_flag_as_required('output')
@Composite('TestRandom', derived_attrs=['T: numbertype'], outputs=['o: T'])
def _composite_random_op():
pass
def main(_):
if FLAGS.gen_register_op:
assert FLAGS.output.endswith('.cc')
generated_code = gen_register_op(sys.modules[__name__], '_composite_')
else:
assert FLAGS.output.endswith('.mlir')
generated_code = tfr_gen_from_module(sys.modules[__name__],
'_composite_')
import os
import exporter
input_type = ["image_tensor","inputs"]
types = ["uint","float"]
flags.DEFINE_string('input_type','image_tensor','Type of input_node type,you can choice from '
'[image_tensor,inputs]')
flags.DEFINE_string('data_type','uint','you can choice from [uint,float]')
flags.DEFINE_string('input_shape',None,"The inpnut_node's shape,if dimensions's values is -1,it's shape is unkown,"
"but also you can specify for an input_node'shapes,default shape is [None,None,None,3]")
flags.DEFINE_string('pipline_config_path',None,"It's your model's config file's path")
flags.DEFINE_string('trained_checkpoint_prefix',None,"The path your */.ckpt")
flags.DEFINE_string('output_dict',None,"pb models output path")
flags.DEFINE_string('preprocess',None,"Wheather to add preprocess op into the final graph")
flags.DEFINE_string('postprocess',None,"Wheather to add a new output_nodes into the final graph")
flags.mark_flag_as_required('pipline_config_path')
flags.mark_flag_as_required('trained_checkpoint_prefix')
flags.mark_flag_as_required('output_dict')
FLAGS = flags.FLAGS
def main(_):
if FLAGS.input_type not in input_type:
raise ValueError("The input_type must be one from 'image_tensor' or 'inputs':{}".format(FLAGS.input_type))
if FLAGS.data_type not in types:
raise ValueError("The data_type must be one from 'uint' or float:{}".format(FLAGS.data_type))
if not os.path.exists(FLAGS.pipline_config_path):
raise FileNotFoundError("The file {} is not found".format(FLAGS.pipline_config_path))
if not os.path.exists(FLAGS.output_dict):
os.mkdir(FLAGS.output_dict)