- Do I have to do this?
- No, this is only needed if you want to take advantage of GPU computing
- Will this work on any computer
- No, it only works if you have a GPU that can be used by CNTK.
- According to Microsoft (source): You can check whether your card is CUDA-compatible here and here (for older cards). Your GPU card Compute Capability (CC) must be 3.0 or more.
- How large is it?
- CNTK will download several GB of data during installation.
- Any prerequisites?
- CNTK comes with its own Anaconda, if you already have an Anaconda installation you may want to remove it. CNTK might be able to use your existing Anaconda, but the installation instructions are surprisingly specific about the Anaconda version CNTK needs.
- Download the zip file from here (click "I accept" at the bottom) - https://cntk.ai/dlwg-2.6.html
- Unzip the file (e.g. into C:\ProgramData\CNTK)
- cd to cntk\Scripts\install\windows\
- Run
install.bat -AnacondaBasePath C:\ProgramData\Anaconda3 -PyVersion 36- Replace C:\ProgramData\Anaconda3 with the path where you want Anaconda to be installed
- When done, the script will display some info, the useful part is the location of the activation script:
- \cntk\scripts\cntkpy36.bat
- Open command line (the usual windows cmd)
- Run the bat faile generated by CNTK
- Start ipython
import keras- Check if it says during the import that it's using cntk as the backend and if it found the GPU
Try running the following to see if your GPU is seen by CNTK (related post )
import cntk.device
cntk.device.gpu(0)