TF_Adapter is committed to providing the outstanding computing power of Ascend AI Processors to developers who use the TensorFlow framework. Developers only need to install the TF_Adapter plug-in and add a small amount of configuration to the existing TensorFlow script to accelerate their training jobs on the Ascend AI Processors.
You can read TF_Adapter Interface for more details。
You can build the TF_Adapter software package from the source code and install it on the Ascend AI Processor environment.
The TF_Adapter plug-in has a strict matching relationship with TensorFlow. Before building from source code, you need to ensure that it has been installed correctly TensorFlow v1.15.0 Version.
You may also build GraphEngine from the source. To build GraphEngine, please make sure that you have access to an Ascend 910 environment as compiling environment, and make sure that the following software requirements are fulfilled.
git clone https://gitee.com/ascend/tensorflow.git
cd tensorflow
chmod +x build.sh
./build.sh
After the script is successfully executed, a compressed file of tfadapter.tar will be generated in the output directory.
Unzip the tfadapter.tar file to generate npu_bridge-1.15.0-py3-none-any.whl. Then you can install the TF_Adapter plug-in using pip.
pip install ./dist/python/dist/npu_bridge-1.15.0-py3-none-any.whl
It should be noted that you should ensure that the installation path is the same as the Python you specified when compiling The interpreter search path is consistent.
Welcome to contribute.
For Release Notes, see our RELEASE.
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。
Activity
Community
Health
Trend
Influence
:Code submit frequency
:React/respond to issue & PR etc.
:Well-balanced team members and collaboration
:Recent popularity of project
:Star counts, download counts etc.