QNAP QTS 實測各家AI Framework 建議搭配 @ 觀察筆記

QNAP QTS 實測各家AI docker image 使先要注意的是QTS 怎麼使用GPU
從官網的QNAP Nvidia GPU Driver release note 發布公告主要分兩個版本：

QTS 4.3.x 系列可以用： NVIDIA GPU Driver 1.2.x

[Libraries update]
- Nvidia Driver updated to 384.111
- Cuda updated to 9.0 ,
- ffmpeg updated to 3.4
- Added 10bit transcode.

建議搭配組合：

Docker name	Tag
tensorflow/tensorflow	1.11.0-gpu
pytorch/pytorch	0.4.1-cuda9-cudnn7-devel
bvlc/caffe	gpu
mxnet/python	1.3.1_gpu_cu90_mkl_py3
microsoft/cntk	2.6-gpu-python3.5-cuda9.0-cudnn7.0
chainer/chainer	v5.0.0b4-python3

QTS 4.4.x 系列可以用： NVIDIA GPU Driver 2.0.x

[Libraries update]
- Nvidia Driver updated to 410.78
- Cuda updated to 10.0 ,
- ffmpeg updated to 34.1

建議搭配組合：

Docker name	Tag
tensorflow/tensorflow	1.11.0-gpu
pytorch/pytorch	1.0.1-cuda10.0-cudnn7-devel
bvlc/caffe	gpu
mxnet/python	1.3.1_gpu_cu92_mkl_py3
microsoft/cntk	2.6-gpu-python3.5-cuda9.0-cudnn7.0
chainer/chainer	v6.0.0b1