VS2017下AI工具的安装
以下根据教程进行：https://www.cnblogs.com/ms-uap/p/9182530.html
-
卸载了所有可以卸载的已安装文件（cntk，tensorflow,...)
-
Path下所有路径清除
-
VS里Python工具安装失败（找不到python36.dll)，在VS installer中修复后OK
-
按照文章说明进行安装VS的AI工具插件
-
安装CUDA9.0，失败（CUDA VS插件安装失败）：拷贝CUDA VS插件安装文件，只安装其它，然后再单独安装CUDA插件
-
按照文章说明把CuDNN的dll文件拷到安装目录的bin文件夹内
-
按照文章说明git clone安装脚本（我的默认Python路径与文章给出的不一样，可能是注册表里的信息）
git clone https://github.com/Microsoft/samples-for-ai.git cd samples-for-ai -
运行安装脚本（需要在admin模式，python需要可用）
-
执行 intall.py
python installer/install.pyD:\repos\MSAI\samples-for-ai>python installer/install.py 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Detecting system information ... 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] OS: Windows-10, 64bit 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Python: 3.6.5, 64bit 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] NVIDIA GPU: Find 1 GPU device(s) that meet the compute capability requirement. 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Git: True 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Visual Studio: VS2017 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] CUDA: 9.0 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Cudnn: 7 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Begin to install scipy(numpy, scipy) ... 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install numpy 1.14.3 ... 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] Begin to install CNTK(BrainScript) ... 06:22:37 [INFO] [Microsoft Visual Studio Tools for AI] CNTK(BrainScript)-2.5.1 is already installed. 06:22:38 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install numpy 1.14.3 successfully! 06:22:38 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install scipy 1.1.0 ... 06:22:38 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install scipy 1.1.0 successfully! 06:22:38 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install cntk-gpu 2.5.1 ... 06:22:38 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install cntk-gpu 2.5.1 successfully! 06:22:38 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install tensorflow-gpu 1.5.0 ... 06:22:39 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install tensorflow-gpu 1.5.0 successfully! 06:22:39 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install torch 0.4.0 ... 06:22:39 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install torch 0.4.0 successfully! 06:22:39 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install torchvision 0.2.1 ... 06:22:40 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install torchvision 0.2.1 successfully! 06:22:40 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install mxnet-cu90 1.2.0 ... 06:22:40 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install mxnet-cu90 1.2.0 successfully! 06:22:40 [INFO] [Microsoft Visual Studio Tools for AI] Begin to install chainer(cupy, chainer) ... 06:22:40 [WARNING] [Microsoft Visual Studio Tools for AI] On windows, please manully install cupy. You can reference thi s link https://github.com/Microsoft/vs-tools-for-ai/blob/master/docs/prepare-localmachine.md#chainer. 06:22:40 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install chainer 4.1.0 ... 06:22:41 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install chainer 4.1.0 successfully! 06:22:41 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install Theano 1.0.2 ... 06:22:41 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install Theano 1.0.2 successfully! 06:22:41 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install Keras 2.1.6 ... 06:22:42 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install Keras 2.1.6 successfully! 06:22:42 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install caffe2 0.8.1 ... 06:22:42 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install caffe2 0.8.1 successfully! 06:22:42 [INFO] [Microsoft Visual Studio Tools for AI] Begin to install ml software(scikit-learn, xgboost and libsvm) .. . 06:22:42 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install scikit-learn 0.19.1 ... 06:22:43 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install scikit-learn 0.19.1 successfully! 06:22:43 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install xgboost 0.71 ... 06:22:43 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install xgboost 0.71 successfully! 06:22:43 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install libsvm 3.22 ... 06:22:44 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install libsvm 3.22 successfully! 06:22:44 [INFO] [Microsoft Visual Studio Tools for AI] Begin to install converter(coremltools, onnx, tf2onnx, onnxmltool s and winmltools) ... 06:22:44 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install coremltools 0.8 ... 06:22:55 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install coremltools 0.8 successfully! 06:22:55 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install onnx 1.1.2 ... 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install onnx 1.1.2 successfully! 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] onnxmltools is already installed. 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] winmltools is already installed. 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] tf2onnx is already installed. We will uninstall it and upgrade to the latest version. 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-uninstall tf2onnx 0.0.0.1 ... 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] Pip-uninstall tf2onnx 0.0.0.1 successfully! 06:22:56 [INFO] [Microsoft Visual Studio Tools for AI] Begin to pip-install tf2onnx 0.0.0.1 ... 06:23:14 [INFO] [Microsoft Visual Studio Tools for AI] Pip-install tf2onnx 0.0.0.1 successfully! 06:23:14 [INFO] [Microsoft Visual Studio Tools for AI] Begin to install extra software(jupyter, matplotlib, and pandas) ... 06:23:14 [INFO] [Microsoft Visual Studio Tools for AI] jupyter is already installed. 06:23:14 [INFO] [Microsoft Visual Studio Tools for AI] matplotlib is already installed. 06:23:14 [INFO] [Microsoft Visual Studio Tools for AI] pandas is already installed. 06:23:14 [INFO] [Microsoft Visual Studio Tools for AI] Setup finishes. Press enter to exit. -
训练第一个模型MINST
D:\repos\MSAI\samples-for-ai>python examples/tensorflow/mnist/mnist.pySuccessfully downloaded train-images-idx3-ubyte.gz 9912422 bytes. Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes. Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes. Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes. Extracting input\train-images-idx3-ubyte.gz Extracting input\train-labels-idx1-ubyte.gz Extracting input\t10k-images-idx3-ubyte.gz Extracting input\t10k-labels-idx1-ubyte.gz 2018-08-09 06:34:46.101979: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\36\tensorflow\core\platform\cpu_feature_g uard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2 2018-08-09 06:34:46.359657: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\36\tensorflow\core\common_runtime\gpu\gpu _device.cc:1105] Found device 0 with properties: name: GeForce GTX 1080 Ti major: 6 minor: 1 memoryClockRate(GHz): 1.582 pciBusID: 0000:01:00.0 totalMemory: 11.00GiB freeMemory: 9.10GiB 2018-08-09 06:34:46.376253: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\36\tensorflow\core\common_runtime\gpu\gpu _device.cc:1195] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 1080 Ti, pci bus id: 0000:0 1:00.0, compute capability: 6.1) Initialized! Step 0 (epoch 0.00), 19.4 ms Minibatch loss: 8.334, learning rate: 0.010000 Minibatch error: 85.9% Validation error: 84.6% Step 100 (epoch 0.12), 5.2 ms Minibatch loss: 3.237, learning rate: 0.010000 Minibatch error: 4.7% Validation error: 7.8% ... ... ... Step 8500 (epoch 9.89), 4.6 ms Minibatch loss: 1.622, learning rate: 0.006302 Minibatch error: 1.6% Validation error: 0.9% Test error: 0.8% -
文件输出的模型位于 D:\repos\MSAI\samples-for-ai\export,和文章的不一致，用everything找到的
-
按照文章说明可以拷贝一个手写数字识别的VS工程sln
git clone https://github.com/squirrelsc/sample-mnist.git start sample-mnist\MnistDemo.sln -
nuget包自动加载ML.Scoring
-
导入训练好的Mnist模型(.pd文件)，工具自动生成了class 供导入
-
Compile&Run