Richard Burton | dc0c6ed | 2020-04-08 16:39:05 +0100 | [diff] [blame] | 1 | # Copyright © 2020 Arm Ltd. All rights reserved. |
| 2 | # SPDX-License-Identifier: MIT |
| 3 | import os |
| 4 | |
| 5 | import pytest |
| 6 | import pyarmnn as ann |
| 7 | import numpy as np |
| 8 | |
| 9 | |
| 10 | @pytest.fixture(scope="function") |
| 11 | def get_tensor_info_input(shared_data_folder): |
| 12 | """ |
| 13 | Sample input tensor information. |
| 14 | """ |
| 15 | parser = ann.ITfLiteParser() |
| 16 | parser.CreateNetworkFromBinaryFile(os.path.join(shared_data_folder, 'mock_model.tflite')) |
| 17 | graph_id = 0 |
| 18 | |
| 19 | input_binding_info = [parser.GetNetworkInputBindingInfo(graph_id, 'input_1')] |
| 20 | |
| 21 | yield input_binding_info |
| 22 | |
| 23 | |
| 24 | @pytest.fixture(scope="function") |
| 25 | def get_tensor_info_output(shared_data_folder): |
| 26 | """ |
| 27 | Sample output tensor information. |
| 28 | """ |
| 29 | parser = ann.ITfLiteParser() |
| 30 | parser.CreateNetworkFromBinaryFile(os.path.join(shared_data_folder, 'mock_model.tflite')) |
| 31 | graph_id = 0 |
| 32 | |
| 33 | output_names = parser.GetSubgraphOutputTensorNames(graph_id) |
| 34 | outputs_binding_info = [] |
| 35 | |
| 36 | for output_name in output_names: |
| 37 | outputs_binding_info.append(parser.GetNetworkOutputBindingInfo(graph_id, output_name)) |
| 38 | |
| 39 | yield outputs_binding_info |
| 40 | |
| 41 | |
| 42 | def test_make_input_tensors(get_tensor_info_input): |
| 43 | input_tensor_info = get_tensor_info_input |
| 44 | input_data = [] |
| 45 | |
| 46 | for tensor_id, tensor_info in input_tensor_info: |
| 47 | input_data.append(np.random.randint(0, 255, size=(1, tensor_info.GetNumElements())).astype(np.uint8)) |
| 48 | |
| 49 | input_tensors = ann.make_input_tensors(input_tensor_info, input_data) |
| 50 | assert len(input_tensors) == 1 |
| 51 | |
| 52 | for tensor, tensor_info in zip(input_tensors, input_tensor_info): |
| 53 | # Because we created ConstTensor function, we cannot check type directly. |
| 54 | assert type(tensor[1]).__name__ == 'ConstTensor' |
| 55 | assert str(tensor[1].GetInfo()) == str(tensor_info[1]) |
| 56 | |
| 57 | |
| 58 | def test_make_output_tensors(get_tensor_info_output): |
| 59 | output_binding_info = get_tensor_info_output |
| 60 | |
| 61 | output_tensors = ann.make_output_tensors(output_binding_info) |
| 62 | assert len(output_tensors) == 1 |
| 63 | |
| 64 | for tensor, tensor_info in zip(output_tensors, output_binding_info): |
| 65 | assert type(tensor[1]) == ann.Tensor |
| 66 | assert str(tensor[1].GetInfo()) == str(tensor_info[1]) |
| 67 | |
| 68 | |
| 69 | def test_workload_tensors_to_ndarray(get_tensor_info_output): |
| 70 | # Check shape and size of output from workload_tensors_to_ndarray matches expected. |
| 71 | output_binding_info = get_tensor_info_output |
| 72 | output_tensors = ann.make_output_tensors(output_binding_info) |
| 73 | |
| 74 | data = ann.workload_tensors_to_ndarray(output_tensors) |
| 75 | |
| 76 | for i in range(0, len(output_tensors)): |
| 77 | assert data[i].shape == tuple(output_tensors[i][1].GetShape()) |
| 78 | assert data[i].size == output_tensors[i][1].GetNumElements() |
| 79 | |
| 80 | |
| 81 | def test_make_input_tensors_fp16(get_tensor_info_input): |
| 82 | # Check ConstTensor with float16 |
| 83 | input_tensor_info = get_tensor_info_input |
| 84 | input_data = [] |
| 85 | |
| 86 | for tensor_id, tensor_info in input_tensor_info: |
| 87 | input_data.append(np.random.randint(0, 255, size=(1, tensor_info.GetNumElements())).astype(np.float16)) |
| 88 | tensor_info.SetDataType(ann.DataType_Float16) # set datatype to float16 |
| 89 | |
| 90 | input_tensors = ann.make_input_tensors(input_tensor_info, input_data) |
| 91 | assert len(input_tensors) == 1 |
| 92 | |
| 93 | for tensor, tensor_info in zip(input_tensors, input_tensor_info): |
| 94 | # Because we created ConstTensor function, we cannot check type directly. |
| 95 | assert type(tensor[1]).__name__ == 'ConstTensor' |
| 96 | assert str(tensor[1].GetInfo()) == str(tensor_info[1]) |
| 97 | assert tensor[1].GetDataType() == ann.DataType_Float16 |
| 98 | assert tensor[1].GetNumElements() == 28*28*1 |
| 99 | assert tensor[1].GetNumBytes() == (28*28*1)*2 # check each element is two byte |