An index repository of prebuilt Windows binaries of TensorFlow (Files stored at Google Drive and Cow Transfer).
一个预编译的Windows下的TensorFlow索引仓库（文件主体在谷歌云盘和奶牛快传）。
| TensorFlow | CUDA | cuDNN | VC | LLVM/Clang | MKL | Python | Google Drive | 奶牛快传 | File Type | Size |
|---|---|---|---|---|---|---|---|---|---|---|
| 2.18.0-rc0 | No | No | 2022 | 18.1.7 | No | 3.12.7 | ⬇️ | ⬇️ | wheel (rar) | 234.81MB |
| 2.5.0 | 11.3 | 8.2.0 | 2019 | No | Yes | 3.9.5 | ⬇️ | ⬇️ | wheel (rar) | 249.58MB |
| - | ⬇️ | ⬇️ | C++ Library (rar) | 217.57MB | ||||||
| No | 3.9.5 | ⬇️ | ⬇️ | wheel (rar) | 250.15MB | |||||
| - | ⬇️ | ⬇️ | C++ Library (rar) | 218.10MB | ||||||
| 2.4.1 | 11.2 | 8.1.0 | 2019 | No | Yes | 3.9.1 | ⬇️ | ⬇️ | wheel (rar) | 294.55MB |
| - | ⬇️ | ⬇️ | C++ Library (rar) | 282.68MB | ||||||
| No | 3.9.1 | ⬇️ | ⬇️ | wheel (rar) | 292.47MB | |||||
| - | ⬇️ | ⬇️ | C++ Library (rar) | 279.57MB | ||||||
| 2.3 | 11.0.3 | 8.0.2 | 2019 | No | Yes | 3.8.5 | ⬇️ | ⬇️ | wheel (rar) | 246.55MB |
| - | ⬇️ | ⬇️ | C Library (rar) | 181.82MB | ||||||
| No | 3.8.5 | ⬇️ | ⬇️ | wheel (rar) | 210.11MB | |||||
| - | ⬇️ | ⬇️ | C Library (rar) | 175.60MB | ||||||
| 2.2 | 10.2 | 7.6.5 | 2019 | No | No | 3.8.2 | ⬇️ | ⬇️ | wheel (rar) | 153.37MB |
| - | ⬇️ | ⬇️ | C Library (rar) | 110.90MB |
-
You can directly download whl file to install with command like
pip install tensorflow-x.x.x-x-x-x.whl.For someone with bad network connection, you could download rar file, then extract into a directory, then pack with ZIP and change extension to "whl" to install.
可以直接下载whl文件，使用
pip install tensorflow-x.x.x-x-x-x.whl命令安装。对于网络不好的人，也可以下载rar文件，解压到一个目录下，再对其使用ZIP打包后更改文件扩展名为whl进行安装。 -
For 2.3 and lower, MKL-enabled libraries need mklml.dll and libiomp5md.dll in system Path, or just in the same path as Python.exe.
对于2.3及以下版本, 启用MKL的库需要把mklml.dll和libiomp5md.dll放在系统环境变量的路径（Path）中，或者直接把他们放到Python.exe同目录下。
TensorFlow Google Drive 奶牛快传 2.3.0 mklml-libiomp5md-tensorflow-2.3.0.rar mklml-libiomp5md-tensorflow-2.3.0.rar -
For 2.4.1 and higher, MKL-enabled libraries need libomp.dll in system Path, or just in the same path as Python.exe.
对于2.4.1及更高版本, 启用MKL的库需要把libomp.dll放在系统环境变量的路径（Path）中，或者直接把他们放到Python.exe同目录下。
TensorFlow Google Drive 奶牛快传 2.4.1 libomp-tensorflow-2.4.1.rar libomp-tensorflow-2.4.1.rar 2.5.0 libomp-tensorflow-2.5.0.rar libomp-tensorflow-2.5.0.rar
To save space, wheels of old version would be removed, please download compressed RAR files, the method of which is as Usage.1 said.
为了节约网盘空间，旧版的wheel安装包会被移除，请下载压缩的RAR文件，使用方法如Usage.1所述。