"Cannot connect to GPU backend" when I runtime the Wk4 prog. assignment

Course Q&A TensorFlow Developer Professional Convolutional Neural Networks in TensorFlow
1

I have been trying to do the Week 4 programming assignment since the last night, but every time I try to runtime the very first loading code in Colab I get the below message. I had completed Week3 programming assignment on the same day (yesterday). Is there any daily limit which I may have maxed out. When will I get the access to the GPU runtime back?

" Cannot connect to GPU backend

You cannot currently connect to a GPU due to usage limits in Colab. Learn more

To get more access to GPUs, consider purchasing Colab compute units with [Pay As You Go]

CloseConnect without GPU"

2

Google colab credits are not offered for this lab. You have 2 options:

  1. Pick a provider like google colab and use their paid plan.
  2. Download the notebook and complete the assignment on your laptop.
3

Sorry, I don’t understand it. I have completed the prog. assignments in Courses 1 and the 3 prog. assignments in Course 2 and all were running fine. I only got that message when I attempted the last prog. assignment in this course! There was no indication anywhere that there is a “catch” in this course - that I have to pay extra. I can’t run it on my computer! Is there a limit for free-use in Colab that I have maxed out?

5

Isn’t Colab the default environment for this course? When I click on the assignment it automatically opens in Colab! If it is not the default, how do I change it to Jupyter? I noticed that in Course 3 the default environment is Jupyter Notebook.

6

This specialization uses 2 environments:

  1. Coursera jupyter lab: There are no usage limits AFAIK. Your monthly specialization subscription fees covers usage costs.
  2. Google colab: Usage limits exist. You’ll need to pay for gpu access for more experimentation. One way to get around this is to download the notebooks to your local machine to build models.
7

Hi Robert! As another workaround, do you have another Google account where you can try this Colab? You might have reached the resource limits with the current one you’re using. These limits fluctuate as described here. In my experience, I am able to use the GPU runtime again with the same account after 24 hours of seeing that message. Hope it will be the same in your case. Let us know how it goes!

1 Like
8

Thanks, Chris! It works now.:+1:

1 Like
9

Awesome! Glad it works now!

10

As another workaround you can use kaggle to run your notebook with access to their GPU’s for free.
This requires a kaggle account, but it works very well and compiles fairly quickly, if you do not want to wait the 24 hrs for colab!

11

It’s definitely a good idea. But, using kaggle gpus for solving your assignments is likely to get the attention of kaggle admins since many notebooks will be similar. This will most likely result in converting free access to a paid plan like google colab.

Please rent a gpu instance on the cloud and have fun.