If you use python neural network intelligence (nni) for tuning hyper-parameters, you may find the status of trial jobs may be failed. How to fix it? In this tutorial, we will discuss some tips to fix this error.
The status of trial jobs
The status of nni trial jobs may be:
To fix it, you can try these steps:
1. View nni logfiles
You can check nni logfiles on the top right. Here is an example.
Clicking Logfiles, we will find Dispatcher Log.
We find python dill library should be installed.
We can use pip install dill to install it.
If the stauts of nni trial jobs is still failed, you can view step 2.
2.View nni stderr
You can use command nnictl log stderr to view nni stderr. You can find some errors in this file.
Or you can use everything to find and view these stderr files.
View this file, we may find some errors. As to us, we find the powershell can not load WindowsPowerShell\profile.ps1. You can fix this error by following the tutorial below:
Fix PowerShell cannot Load WindowsPowerShell\profile.ps1 on Windows 10 – PowerShell Tutorial
If the status is also failed, you can view step 3.
3.Check the python version is correct?
You should make sure the python nni is run in the right python version. For example, if you have only installed tensorflow in python 3.6. However, you run python nni in python 3.7, which may cuase you can not run nni. To fix this problem, you can read this tutorial.
Run Neural Network Intelligence (NNI) in Multiple Python Version Environments – Python NNI Tutorial