We can use microsoft nni tool to tune hyper-parameters in deep learning. In order to start a experiment, we can use nnictl create command. However, you may find /bin/sh: 1: nnictl: not found error on ubuntu. In this tutorial, we will introduce you how to fix it.
Preliminary
We have rent a gpu server from a cloud gup platform and have installed microsoft neural network intelligence (NNI) using jupyter notebook, then we will use nnictl command to start a trial using jupyter notebook, however we find this error: nnictl: not found.
First, jupyter notebook can allow us to run command line. Here is tutorial:
Jupyter Notebook Run Python Script (.py) File and Command Line: A Simple Guide
Second, we have installed microsoft neural network intelligence successfully. Why does this error occur?
How to fix this error to start a trial?
We have not found the reason why this error occur. In order to fix this error and start a trial for tuning hyper-parameters, we can do like this:
1.Use Xshell to connect gpu server.
Xshell is a powerful terminal emulator that supports SSH, SFTP, TELNET, RLOGIN and SERIAL. We can use it to connect our gpu server and run some commands.
You should download and install it.
2.Run nnictl command
After having connected our gpu server using xshell, we can run nnictl create command to start a trial. Here is an example.
Then you will find this error is fixed.