Colab run time stays “Busy” state after restarting the run time

Question

I have legacy code which uses tensorflow 1.0.1 version in my ipython notebook. I have a requirements.txt file to install dependencies. So I run the cell - .pip3 install -r requirements.txt in my Colab notebook to run this legacy code.

requirements.txt has following dependencies,

• tensorflow-gpu==1.0.1
• nltk==3.2.5
• matplotlib==2.0.2
• scipy==0.19.1
• scikit-learn==0.19.0
• jupyter==1.0.0

After running the above cell in my notebook I get a message to restart the runtime to setup the environment for TensorFlow 1.0.1. After restarting the kernel I always get the runtime in the "Busy" state.

Even though I tried terminating the session to start again. It does not help because I have to restart the kernel in new session too.

Can anyone clarify what should I do to avoid "staying Busy state" after kernel restart?

Answer 1

Try dropping the jupyter install. It's likely that's installing deps that are incompatible with Colab, which results in your backend becoming wedged.

When in this state, you can reset your runtime using the Runtime -> Factory reset runtime menu.