scisharp
/
TensorFlow.NET
Not watched
1
0
Fork 0
Code
Releases 21 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
2184 Commits
27 Branches
392 MB
0 Download
 
 
 
 
Tree: 634860d755
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '634860d755'
${ noResults }
TensorFlow.NET/tools/tensorflowlib
Yaohui Liu b26c37ab20
build: add native library splitter and adjust directory structure. 		2 years ago
..
README.md build: add native library splitter and adjust directory structure. 2 years ago

README.md

TensorFlow.NET pack all required libraries in architecture-specific assemblies folders per NuGet standard.

PM> Install-Package TensorFlow.NET
PM> Install-Package SciSharp.TensorFlow.Redist

Add <RuntimeIdentifier>win-x64</RuntimeIdentifier> to a PropertyGroup in your .csproj when targeting .NET 472.

Run in Linux

Download Linux pre-built library and unzip libtensorflow.so and libtensorflow_framework.so into current running directory.

To run image recognition in Linux, please ensure some prerequisite libraries is install.

sudo apt install libc6-dev 
sudo apt install libgdiplus

More information about System.Drawing on Linux.

Run TensorFlow with GPU

Before running verify you installed CUDA and cuDNN (TensorFlow v1.15 is compatible with CUDA v10.0 and cuDNN v7.4 , TensorFlow v2.x is compatible with CUDA v10.2 and cuDNN v7.65), and make sure the corresponding cuda version is compatible.

Mac OS

There is no GPU support for macOS, in the future TensorFlow will support Apple M1 chip.

GPU for Windows

PM> Install-Package SciSharp.TensorFlow.Redist-Windows-GPU

GPU for Linux

PM> Install-Package SciSharp.TensorFlow.Redist-Linux-GPU

Since NuGet limits file size for 250M, we can't ship Linux GPU version as NuGet, you can download the library from Google TensorFlow Storage.

Download prebuild binary manually

TensorFlow packages are built nightly and uploaded to GCS for all supported platforms. They are uploaded to the libtensorflow-nightly GCS bucket and are indexed by operating system and date built.

Build from source for Windows

https://www.tensorflow.org/install/source_windows

Download Bazel 2.0.0 to build tensorflow2.x. We build customized binary to export c_api from this fork.

Set ENV BAZEL_VC=C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC.

pacman -S git patch unzip

  1. Build static library

bazel build --output_base=C:/tmp/tfcompilation --config=opt //tensorflow:tensorflow

  1. Build pip package

bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package

  1. Generate pip installation file

bazel-bin\tensorflow\tools\pip_package\build_pip_package C:/tmp/tensorflow_pkg

  1. Install from local wheel file.

pip install C:/tmp/tensorflow_pkg/tensorflow-1.15.0-cp36-cp36m-win_amd64.whl

Build from source for MacOS

$ cd /usr/local/lib/bazel/bin
$ curl -LO https://release.bazel.build/3.7.2/release/bazel-3.7.2-darwin-x86_64
$ chmod +x bazel-3.7.2-darwin-x86_64
$ cd ~/Projects/tensorflow
$ bazel build --config=opt //tensorflow:tensorflow

Build specific version for tf.net

https://github.com/SciSharp/tensorflow

For Linux version, these APIs symbols should also be put into tensorflow/c/version_script.lds to be exported.
Please refer to commit https://github.com/SciSharp/tensorflow/commit/58122da06be3e7707500ad889dfd5c760a3e0424

Delete task

Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.

cancel
confirm

tensorflow框架的.NET版本，提供了丰富的特性和API，可以借此很方便地在.NET平台下搭建深度学习训练与推理流程。

C# Unity3D Asset PureBasic Python