Trouble Running the Ungraded Labs (always leads to errors)

AchintyaGahalaut · August 11, 2024, 7:56pm

The Course 2 Week 3 ungraded lab 1:

https-deeplearning-ai/tensorflow-1-public/blob/main/C2/W3/ungraded_lab/C2_W3_Lab_1_transfer_learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<a href=\"https://colab.research.google.com/github/https-deeplearning-ai/tensorflow-1-public/blob/master/C2/W3/ungraded_lab/C2_W3_Lab_1_transfer_learning.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bT0to3TL2q7H"
   },
   "source": [
    "# Ungraded Lab: Transfer Learning\n",
    "\n",
    "In this lab, you will see how you can use a pre-trained model to achieve good results even with a small training dataset. This is called _transfer learning_ and you do this by leveraging the trained layers of an existing model and adding your own layers to fit your application. For example, you can:\n",
    "\n",
    "1. just get the convolution layers of one model\n",

This file has been truncated. show original

I have been having this issue for the duration of the entire certificate, so I wanted to bring this up. I can never run the ungraded labs without some cell throwing an error. An example screenshot is attached. Laurence’s output is also attached below. When I try to fix it with the assistance of the built-in Gemini chatbot, it still doesn’t work. It’s weird because Laurence can run it in the videos without any issues. Any tips? Thanks!

Mine:

Laurence’s (in video " Exploring Transfer Learning with Inception", timestamp: 0:28):

TMosh · August 12, 2024, 3:06am

Chatbots are not programming experts.

balaji.ambresh · August 12, 2024, 3:38am

Coursera jupyter labs run on tensorflow 2.7.x
Google colab uses tensorflow 2.17 which doesn’t play well with code written for 2.7.x (see colab output below)

The staff are aware of this.

AchintyaGahalaut · August 12, 2024, 3:50am

But the error above happened in the google colab notebook, not the jupyter notebook. All of the 2nd course assignments were administered through Google Colab. Is there a way to resolve this? I would really like to be able to play with the ungraded labs on my own. Thanks!

balaji.ambresh · August 12, 2024, 3:52am

Sure.

Use docker to get hold of legacy images. Start here

AchintyaGahalaut · August 12, 2024, 4:04am

I see. Thanks for this resource. I’m kind of new to docker, so how exactly can I use it? Like at a high level, how can Docker help?

balaji.ambresh · August 12, 2024, 4:11am

Install docker on your machine and ensure that dockerd is running.
Run the following command to map the current working directory on the host machine to /tf/notebooks and launch the notebook server:

docker container run --rm -it --network host -v $PWD:/tf/notebooks tensorflow/tensorflow:2.7.4-gpu-jupyter

Pick an image that matches your hardware (like check for gpu)

chris.favila · August 12, 2024, 4:20am

Hi Achintya! Thank you for reporting this! For this lab, please downgrade tensorflow to v2.15 if you want to keep using Colab. You can run this in the first cell.

!pip install tensorflow==2.15.0
!pip install keras==2.15.0

TF2.17 uses Keras 3 by default which can indeed cause bugs if the code expects Keras 2. We’re working on patching the labs so other learners won’t run into the same issue. Thank you!