From fdcaad4c223d12bbd9680509f8932444d0e90b1d Mon Sep 17 00:00:00 2001 From: meng_chunyang Date: Mon, 31 Aug 2020 09:55:00 +0800 Subject: [PATCH] update README for mindspore lite --- RELEASE.md | 1 - mindspore/lite/README.md | 11 +++++++++++ mindspore/lite/README_CN.md | 11 +++++++++++ 3 files changed, 22 insertions(+), 1 deletion(-) diff --git a/RELEASE.md b/RELEASE.md index dde0d0f1e1e..0634c313364 100644 --- a/RELEASE.md +++ b/RELEASE.md @@ -57,7 +57,6 @@ * Add 93 TFLite op. * Add 24 Caffe op. * Add 62 ONNX op. - * Add support for windows. * Add 11 optimized passes, include fusion/const fold. * Support aware-training and Post-training quantization. * CPU diff --git a/mindspore/lite/README.md b/mindspore/lite/README.md index 1f0b6932ee1..3a8ae9b5995 100644 --- a/mindspore/lite/README.md +++ b/mindspore/lite/README.md @@ -54,3 +54,14 @@ For more details please check out our [MindSpore Lite Architecture Guide](https: Load the model and perform inference. [Inference](https://www.mindspore.cn/lite/tutorial/en/master/use/runtime.html) is the process of running input data through the model to get output. MindSpore provides a series of pre-trained models that can be deployed on mobile device [example](#TODO). + +## MindSpore Lite benchmark test result +Base on MindSpore r0.7, we test a couple of networks on HUAWEI Mate30 (Hisilicon Kirin990) mobile phone, and get the test results below for your reference. + +| NetWork | Thread Number | Average Run Time(ms) | +| ------------------- | ------------- | -------------------- | +| basic_squeezenet | 4 | 9.10 | +| inception_v3 | 4 | 69.361 | +| mobilenet_v1_10_224 | 4 | 7.137 | +| mobilenet_v2_10_224 | 4 | 5.569 | +| resnet_v2_50 | 4 | 48.691 | diff --git a/mindspore/lite/README_CN.md b/mindspore/lite/README_CN.md index 75b880580ae..30474c1186e 100644 --- a/mindspore/lite/README_CN.md +++ b/mindspore/lite/README_CN.md @@ -64,3 +64,14 @@ MindSpore Lite是MindSpore推出的端云协同的、轻量化、高性能AI推 主要完成模型推理工作，即加载模型，完成模型相关的所有计算。[推理](https://www.mindspore.cn/lite/tutorial/zh-CN/master/use/runtime.html)是通过模型运行输入数据，获取预测的过程。 MindSpore提供了一系列预训练模型部署在智能终端的[样例](#TODO)。 + +## MindSpore Lite性能参考数据 +我们在HUAWEI Mate30（Hisilicon Kirin990）手机上，基于MindSpore r0.7，测试了一组端侧常见网络的性能数据，供您参考： + + | 网络 | 线程数 | 平均推理时间(毫秒) | + | ------------------- | ------ | ------------------ | + | basic_squeezenet | 4 | 9.10 | + | inception_v3 | 4 | 69.361 | + | mobilenet_v1_10_224 | 4 | 7.137 | + | mobilenet_v2_10_224 | 4 | 5.569 | + | resnet_v2_50 | 4 | 48.691 |