# Copyright 2020 The TensorFlow Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== """Test configs for batchmatmul.""" import tensorflow as tf from tensorflow.lite.testing.zip_test_utils import create_tensor_data from tensorflow.lite.testing.zip_test_utils import make_zip_of_tests from tensorflow.lite.testing.zip_test_utils import register_make_test_function @register_make_test_function("make_batchmatmul_tests") def make_batchmatmul_tests(options): """Make a set of tests to do basic batch matrix multiply.""" test_parameters = [ { "dtype": [tf.float32], "shapes": [((3, 4, 7), (7, 9), (3, 4, 7), (7, 9)), ((None, 4, 5), (None, 5, 6), (3, 4, 5), (3, 5, 6)), ((None, 1, 3, 4), (None, 4, 2), (2, 1, 3, 4), (5, 4, 2)), ((None, None, None, 3, 4), (None, None, None, 4, 3), (2, 2, 2, 3, 4), (2, 2, 2, 4, 3))], "adjoint_b": [False, True], "adjoint_a": [False, True], "rhs_constant": [False], "fully_quantize": [False, True], }, ] def swap_last_two_dims(*args): """Return a tuple with the last two dimensions swapped.""" return args[:-2] + (args[-1],) + (args[-2],) def build_graph(parameters): """Build a simple graph with BatchMatMul.""" placeholder0_shape = parameters["shapes"][0] adj_a = parameters["adjoint_a"] adj_b = parameters["adjoint_b"] rhs_constant = parameters["rhs_constant"] if adj_a: placeholder0_shape = swap_last_two_dims(*placeholder0_shape) input0_tensor = tf.compat.v1.placeholder( dtype=parameters["dtype"], shape=placeholder0_shape) if rhs_constant: if adj_b: constant1_shape = swap_last_two_dims(*parameters["shapes"][3]) else: constant1_shape = parameters["shapes"][3] data = create_tensor_data( parameters["dtype"], constant1_shape, min_value=-1.0, max_value=1.0) input1_constant = tf.constant( data, shape=constant1_shape, dtype=parameters["dtype"]) out = tf.matmul( input0_tensor, input1_constant, adjoint_a=adj_a, adjoint_b=adj_b) return [input0_tensor], [out] else: if adj_b: placeholder1_shape = swap_last_two_dims(*parameters["shapes"][1]) else: placeholder1_shape = parameters["shapes"][1] input1_tensor = tf.compat.v1.placeholder( dtype=parameters["dtype"], shape=placeholder1_shape) out = tf.matmul( input0_tensor, input1_tensor, adjoint_a=adj_a, adjoint_b=adj_b) return [input0_tensor, input1_tensor], [out] def build_inputs(parameters, sess, inputs, outputs): """Feed inputs, assign variables, and freeze graph.""" input0_shape = parameters["shapes"][2] adj_a = parameters["adjoint_a"] adj_b = parameters["adjoint_b"] rhs_constant = parameters["rhs_constant"] if adj_a: input0_shape = swap_last_two_dims(*input0_shape) input0_value = create_tensor_data( parameters["dtype"], input0_shape, min_value=-1.0, max_value=1.0) if rhs_constant: output_values = sess.run( outputs, feed_dict=dict(zip(inputs, [input0_value]))) return [input0_value], output_values else: input1_shape = parameters["shapes"][ 3] if not adj_b else swap_last_two_dims(*parameters["shapes"][3]) input1_value = create_tensor_data( parameters["dtype"], input1_shape, min_value=-1.0, max_value=1.0) output_values = sess.run( outputs, feed_dict=dict(zip(inputs, [input0_value, input1_value]))) return [input0_value, input1_value], output_values options.disable_batchmatmul_unfold = True make_zip_of_tests( options, test_parameters, build_graph, build_inputs, expected_tf_failures=0)