keras runing cnn on python3.8
P ython 3.8.0 released in 2020/01/17 .Several days later,anaconda began support python3.8 on multiple system.We yangwaiyang tech test python 3.8 on ubuntu,win10 accord to install virtual environment of python3.8 by anaconda.We test and explore deep learning and gpu speed up accelerated computing. We test framework of deep learning by mxnet and keras.We test framework of gpuspeed up accelerated computing by cupy.Now we released training models in python3.8 on nvidia gpu.
We test it on ubuntu 18.04,run an east cnn of mnist dataset.The answer is the picture above.Obviously it successfully called gpu.
cupy speed up accelerated computing in python3.8,
Cupy is a library of speed up accelerated computing by cuda in python,which similar with numpy. support mult version of python,cuda,syterm.But not released in python3.8， (until 2019/12/05,cupy and chainer support python 3.8.0 )so we build it by ourself,base cuda10.0 on ubuntu.The answer is the picture above.Obviously it successfully called gpu.
Postscript:2019/12/25 while we build tensorflow we found it support python 3.8 since tensorflow version 2.1.0rc0,but should you build itself,no prebuild whl gived. so we built it ourself and test it,then we found whl built by ubuntu 18.04 can't work in ubuntu 16.04,so we build tensorflow in ubuntu 16.04 and cuda10.0 with gcc,amd rocm 2.8.0( downloads links).You can use in ubuntu 16.04 or later you can get code here《anaconda pypi国内镜像和深度学习 cuda加速 图像处理框架测试》in article center
If want have a try or install you can visit our taobao online shop
python3.8深度学习和gpu加速：python3.8深度学习cuda加速mxnet keras cupy gpu租用 安装调试
if any error pleasr contact us,contact information is in abous us
about us
yangwaiyang tech： yangwaiyang tech
share happy
