All course materials for the Zero to Mastery Deep Learning with TensorFlow course.

Overview

Zero to Mastery Deep Learning with TensorFlow

All of the course materials for the Zero to Mastery Deep Learning with TensorFlow course.

This course will teach you foundations of deep learning and TensorFlow as well as prepare you to pass the TensorFlow Developer Certification exam (optional).

Before signing up to the full course, you can watch the first 14-hours of videos on YouTube in a two part series:

  • Part 1 contains notebooks 00, 01 and some of 02 (see below)
  • Part 2 starts where part 1 left off and finishes the rest of 02

Videos going through the rest of the notebooks (03 - 10) are available in the full course.

Got questions about the course?

Check out the livestream Q&A celebrating the course launch on YouTube.

Otherwise, many of them might be answered below.

Contents of this page

Course materials

This table is the ground truth for course materials. All the links you need for everything will be here.

Key:

  • Number: The number of the target notebook (this may not match the video section of the course but it ties together all of the materials in the table)
  • Notebook: The notebook for a particular module with lots of code and text annotations (notebooks from the videos are based on these)
  • Data/model: Links to datasets/pre-trained models for the assosciated notebook
  • Exercises & Extra-curriculum: Each module comes with a set of exercises and extra-curriculum to help practice your skills and learn more, I suggest going through these before you move onto the next module
  • Slides: Although we focus on writing TensorFlow code, we sometimes use pretty slides to describe different concepts, you'll find them here

Note: You can get all of the notebook code created during the videos in the video_notebooks directory.

Number Notebook Data/Model Exercises & Extra-curriculum Slides
00 TensorFlow Fundamentals Go to exercises & extra-curriculum Go to slides
01 TensorFlow Regression Go to exercises & extra-curriculum Go to slides
02 TensorFlow Classification Go to exercises & extra-curriculum Go to slides
03 TensorFlow Computer Vision pizza_steak, 10_food_classes_all_data Go to exercises & extra-curriculum Go to slides
04 Transfer Learning Part 1: Feature extraction 10_food_classes_10_percent Go to exercises & extra-curriculum Go to slides
05 Transfer Learning Part 2: Fine-tuning 10_food_classes_10_percent, 10_food_classes_1_percent, 10_food_classes_all_data Go to exercises & extra-curriculum Go to slides
06 Transfer Learning Part 3: Scaling up 101_food_classes_10_percent, custom_food_images, fine_tuned_efficientnet_model Go to exercises & extra-curriculum Go to slides
07 Milestone Project 1: Food Vision πŸ” πŸ‘ , Template (your challenge) feature_extraction_mixed_precision_efficientnet_model, fine_tuned_mixed_precision_efficientnet_model Go to exercises & extra-curriculum Go to slides
08 TensorFlow NLP Fundamentals diaster_or_no_diaster_tweets, USE_feature_extractor_model Go to exercises & extra-curriculum Go to slides
09 Milestone Project 2: SkimLit πŸ“„ πŸ”₯ pubmed_RCT_200k_dataset, skimlit_tribrid_model Go to exercises & extra-curriculum Go to slides
10 TensorFlow Time Series Fundamentals & Milestone Project 3: BitPredict πŸ’° πŸ“ˆ (videos coming soon)

Course structure

This course is code first. The goal is to get you writing deep learning code as soon as possible.

It is taught with the following mantra:

Code -> Concept -> Code -> Concept -> Code -> Concept

This means we write code first then step through the concepts behind it.

If you've got 6-months experience writing Python code and a willingness to learn (most important), you'll be able to do the course.

Should you do this course?

Do you have 1+ years experience with deep learning and writing TensorFlow code?

If yes, no you shouldn't, use your skills to build something.

If no, move onto the next question.

Have you done at least one beginner machine learning course and would like to learn about deep learning/pass the TensorFlow Developer Certification?

If yes, this course is for you.

If no, go and do a beginner machine learning course and if you decide you want to learn TensorFlow, this page will still be here.

Prerequisites

What do I need to know to go through this course?

  • 6+ months writing Python code. Can you write a Python function which accepts and uses parameters? That’s good enough. If you don’t know what that means, spend another month or two writing Python code and then come back here.
  • At least one beginner machine learning course. Are you familiar with the idea of training, validation and test sets? Do you know what supervised learning is? Have you used pandas, NumPy or Matplotlib before? If no to any of these, I’d going through at least one machine learning course which teaches these first and then coming back.
  • Comfortable using Google Colab/Jupyter Notebooks. This course uses Google Colab throughout. If you have never used Google Colab before, it works very similar to Jupyter Notebooks with a few extra features. If you’re not familiar with Google Colab notebooks, I’d suggest going through the Introduction to Google Colab notebook.
  • Plug: The Zero to Mastery beginner-friendly machine learning course (I also teach this) teaches all of the above (and this course is designed as a follow on).

πŸ›  Exercises & πŸ“– Extra-curriculum

To prevent the course from being 100+ hours (deep learning is a broad field), various external resources for different sections are recommended to puruse under your own discrestion.

(solutions to come after the course is released... try the exercises out for yourself first!)


πŸ›  00 TensorFlow Fundamentals Exercises

  1. Create a vector, scalar, matrix and tensor with values of your choosing using tf.constant().
  2. Find the shape, rank and size of the tensors you created in 1.
  3. Create two tensors containing random values between 0 and 1 with shape [5, 300].
  4. Multiply the two tensors you created in 3 using matrix multiplication.
  5. Multiply the two tensors you created in 3 using dot product.
  6. Create a tensor with random values between 0 and 1 with shape [224, 224, 3].
  7. Find the min and max values of the tensor you created in 6 along the first axis.
  8. Created a tensor with random values of shape [1, 224, 224, 3] then squeeze it to change the shape to [224, 224, 3].
  9. Create a tensor with shape [10] using your own choice of values, then find the index which has the maximum value.
  10. One-hot encode the tensor you created in 9.

πŸ“– 00 TensorFlow Fundamentals Extra-curriculum

  • Read through the list of TensorFlow Python APIs, pick one we haven't gone through in this notebook, reverse engineer it (write out the documentation code for yourself) and figure out what it does.
  • Try to create a series of tensor functions to calculate your most recent grocery bill (it's okay if you don't use the names of the items, just the price in numerical form).
    • How would you calculate your grocery bill for the month and for the year using tensors?
  • Go through the TensorFlow 2.x quick start for beginners tutorial (be sure to type out all of the code yourself, even if you don't understand it).
    • Are there any functions we used in here that match what's used in there? Which are the same? Which haven't you seen before?
  • Watch the video "What's a tensor?" - a great visual introduction to many of the concepts we've covered in this notebook.

πŸ›  01 Neural network regression with TensorFlow Exercises

  1. Create your own regression dataset (or make the one we created in "Create data to view and fit" bigger) and build fit a model to it.
  2. Try building a neural network with 4 Dense layers and fitting it to your own regression dataset, how does it perform?
  3. Try and improve the results we got on the insurance dataset, some things you might want to try include:
  • Building a larger model (how does one with 4 dense layers go?).
  • Increasing the number of units in each layer.
  • Lookup the documentation of Adam and find out what the first parameter is, what happens if you increase it by 10x?
  • What happens if you train for longer (say 300 epochs instead of 200)?
  1. Import the Boston pricing dataset from TensorFlow tf.keras.datasets and model it.

πŸ“– 01 Neural network regression with TensorFlow Extra-curriculum


πŸ›  02 Neural network classification with TensorFlow Exercises

  1. Play with neural networks in the TensorFlow Playground for 10-minutes. Especially try different values of the learning, what happens when you decrease it? What happens when you increase it?
  2. Replicate the model pictured in the TensorFlow Playground diagram below using TensorFlow code. Compile it using the Adam optimizer, binary crossentropy loss and accuracy metric. Once it's compiled check a summary of the model. tensorflow playground example neural network Try this network out for yourself on the TensorFlow Playground website. Hint: there are 5 hidden layers but the output layer isn't pictured, you'll have to decide what the output layer should be based on the input data.
  3. Create a classification dataset using Scikit-Learn's make_moons() function, visualize it and then build a model to fit it at over 85% accuracy.
  4. Create a function (or write code) to visualize multiple image predictions for the fashion MNIST at the same time. Plot at least three different images and their prediciton labels at the same time. Hint: see the classifcation tutorial in the TensorFlow documentation for ideas.
  5. Recreate TensorFlow's softmax activation function in your own code. Make sure it can accept a tensor and return that tensor after having the softmax function applied to it.
  6. Train a model to get 88%+ accuracy on the fashion MNIST test set. Plot a confusion matrix to see the results after.
  7. Make a function to show an image of a certain class of the fashion MNIST dataset and make a prediction on it. For example, plot 3 images of the T-shirt class with their predictions.

πŸ“– 02 Neural network classification with TensorFlow Extra-curriculum


πŸ›  03 Computer vision & convolutional neural networks in TensorFlow Exercises

  1. Spend 20-minutes reading and interacting with the CNN explainer website.
  • What are the key terms? e.g. explain convolution in your own words, pooling in your own words
  1. Play around with the "understanding hyperparameters" section in the CNN explainer website for 10-minutes.
  • What is the kernel size?
  • What is the stride?
  • How could you adjust each of these in TensorFlow code?
  1. Take 10 photos of two different things and build your own CNN image classifier using the techniques we've built here.
  2. Find an ideal learning rate for a simple convolutional neural network model on your the 10 class dataset.

πŸ“– 03 Computer vision & convolutional neural networks in TensorFlow Extra-curriculum


πŸ›  04 Transfer Learning in TensorFlow Part 1: Feature Extraction Exercises

  1. Build and fit a model using the same data we have here but with the MobileNetV2 architecture feature extraction (mobilenet_v2_100_224/feature_vector) from TensorFlow Hub, how does it perform compared to our other models?
  2. Name 3 different image classification models on TensorFlow Hub that we haven't used.
  3. Build a model to classify images of two different things you've taken photos of.
  • You can use any feature extraction layer from TensorFlow Hub you like for this.
  • You should aim to have at least 10 images of each class, for example to build a fridge versus oven classifier, you'll want 10 images of fridges and 10 images of ovens.
  1. What is the current best performing model on ImageNet?

πŸ“– 04 Transfer Learning in TensorFlow Part 1: Feature Extraction Extra-curriculum


πŸ›  05 Transfer Learning in TensorFlow Part 2: Fine-tuning Exercises

  1. Write a function to visualize an image from any dataset (train or test file) and any class (e.g. "steak", "pizza"... etc), visualize it and make a prediction on it using a trained model.
  2. Use feature-extraction to train a transfer learning model on 10% of the Food Vision data for 10 epochs using tf.keras.applications.EfficientNetB0 as the base model. Use the ModelCheckpoint callback to save the weights to file.
  3. Fine-tune the last 20 layers of the base model you trained in 2 for another 10 epochs. How did it go?
  4. Fine-tune the last 30 layers of the base model you trained in 2 for another 10 epochs. How did it go?

πŸ“– 05 Transfer Learning in TensorFlow Part 2: Fine-tuning Extra-curriculum

  • Read the documentation on data augmentation in TensorFlow.
  • Read the ULMFit paper (technical) for an introduction to the concept of freezing and unfreezing different layers.
  • Read up on learning rate scheduling (there's a TensorFlow callback for this), how could this influence our model training?
    • If you're training for longer, you probably want to reduce the learning rate as you go... the closer you get to the bottom of the hill, the smaller steps you want to take. Imagine it like finding a coin at the bottom of your couch. In the beginning your arm movements are going to be large and the closer you get, the smaller your movements become.

πŸ›  06 Transfer Learning in TensorFlow Part 3: Scaling-up Exercises

  1. Take 3 of your own photos of food and use the trained model to make predictions on them, share your predictions with the other students in Discord and show off your Food Vision model πŸ” πŸ‘ .
  2. Train a feature-extraction transfer learning model for 10 epochs on the same data and compare its performance versus a model which used feature extraction for 5 epochs and fine-tuning for 5 epochs (like we've used in this notebook). Which method is better?
  3. Recreate the first model (the feature extraction model) with mixed_precision turned on.
  • Does it make the model train faster?
  • Does it effect the accuracy or performance of our model?
  • What's the advatanges of using mixed_precision training?

πŸ“– 06 Transfer Learning in TensorFlow Part 3: Scaling-up Extra-curriculum

  • Spend 15-minutes reading up on the EarlyStopping callback. What does it do? How could we use it in our model training?
  • Spend an hour reading about Streamlit. What does it do? How might you integrate some of the things we've done in this notebook in a Streamlit app?

πŸ›  07 Milestone Project 1: πŸ” πŸ‘ Food Vision Bigβ„’ Exercises

Note: The chief exercise for Milestone Project 1 is to finish the "TODO" sections in the Milestone Project 1 Template notebook. After doing so, move onto the following.

  1. Use the same evaluation techniques on the large-scale Food Vision model as you did in the previous notebook (Transfer Learning Part 3: Scaling up). More specifically, it would be good to see:
  • A confusion matrix between all of the model's predictions and true labels.
  • A graph showing the f1-scores of each class.
  • A visualization of the model making predictions on various images and comparing the predictions to the ground truth.
    • For example, plot a sample image from the test dataset and have the title of the plot show the prediction, the prediction probability and the ground truth label.
  1. Take 3 of your own photos of food and use the Food Vision model to make predictions on them. How does it go? Share your images/predictions with the other students.
  2. Retrain the model (feature extraction and fine-tuning) we trained in this notebook, except this time use EfficientNetB4 as the base model instead of EfficientNetB0. Do you notice an improvement in performance? Does it take longer to train? Are there any tradeoffs to consider?
  3. Name one important benefit of mixed precision training, how does this benefit take place?

πŸ“– 07 Milestone Project 1: πŸ” πŸ‘ Food Vision Bigβ„’ Extra-curriculum


πŸ›  08 Introduction to NLP (Natural Language Processing) in TensorFlow Exercises

  1. Rebuild, compile and train model_1, model_2 and model_5 using the Keras Sequential API instead of the Functional API.
  2. Retrain the baseline model with 10% of the training data. How does perform compared to the Universal Sentence Encoder model with 10% of the training data?
  3. Try fine-tuning the TF Hub Universal Sentence Encoder model by setting training=True when instantiating it as a Keras layer.
# We can use this encoding layer in place of our text_vectorizer and embedding layer
sentence_encoder_layer = hub.KerasLayer("https://tfhub.dev/google/universal-sentence-encoder/4",
                                        input_shape=[],
                                        dtype=tf.string,
                                        trainable=True) # turn training on to fine-tune the TensorFlow Hub model
  1. Retrain the best model you've got so far on the whole training set (no validation split). Then use this trained model to make predictions on the test dataset and format the predictions into the same format as the sample_submission.csv file from Kaggle (see the Files tab in Colab for what the sample_submission.csv file looks like). Once you've done this, make a submission to the Kaggle competition, how did your model perform?
  2. Combine the ensemble predictions using the majority vote (mode), how does this perform compare to averaging the prediction probabilities of each model?
  3. Make a confusion matrix with the best performing model's predictions on the validation set and the validation ground truth labels.

πŸ“– 08 Introduction to NLP (Natural Language Processing) in TensorFlow Extra-curriculum

To practice what you've learned, a good idea would be to spend an hour on 3 of the following (3-hours total, you could through them all if you want) and then write a blog post about what you've learned.


πŸ›  09 Milestone Project 2: SkimLit πŸ“„ πŸ”₯ Exercises

  1. Train model_5 on all of the data in the training dataset for as many epochs until it stops improving. Since this might take a while, you might want to use:
  1. Checkout the Keras guide on using pretrained GloVe embeddings. Can you get this working with one of our models?
  • Hint: You'll want to incorporate it with a custom token Embedding layer.
  • It's up to you whether or not you fine-tune the GloVe embeddings or leave them frozen.
  1. Try replacing the TensorFlow Hub Universal Sentence Encoder pretrained embedding for the TensorFlow Hub BERT PubMed expert (a language model pretrained on PubMed texts) pretrained embedding. Does this effect results?
  1. What happens if you were to merge our line_number and total_lines features for each sequence? For example, created a X_of_Y feature instead? Does this effect model performance?
  • Another example: line_number=1 and total_lines=11 turns into line_of_X=1_of_11.
  1. Write a function (or series of functions) to take a sample abstract string, preprocess it (in the same way our model has been trained), make a prediction on each sequence in the abstract and return the abstract in the format:

πŸ“– 09 Milestone Project 2: SkimLit πŸ“„ πŸ”₯ Extra-curriculum

What this course is missing

Deep learning is a broad topic. So this course doesn't cover it all.

Here are some of the main topics you might want to look into next:

  • Transformers (the neural network architecture taking the NLP world by storm)
  • Multi-modal models (models which use more than one data source such as text & images)
  • Reinforcement learning
  • Unsupervised learning

Extensions (possible places to go after the course)

Ask questions

Contact Daniel Bourke or add a discussion (preferred).

Status

As of: 28 May 2021 - time series materials inbound!

TODO

geez... I forgot how much there was still to go... classic project planning

  • βœ… Make a GitHub Project for course (see "Projects" tab)
  • βœ… Make a GitHub Discussions for course (thank you Alvaro)
  • βœ… Polish GitHub readme (what you're reading now) with extra resources:
    • data links used in course
    • extra resources & curriculum
  • πŸ”œ Upload slides for each section, done for: 00, 01, 02, 03, 04, 05, 06, 07, 08, 09 (see course materials)
  • πŸ”œ Upload video notebooks for each section, done for: 00, 01, 02, 03, 04, 05, 06, 07, 08, 09 (see .video_notebooks/)
  • Make Colab overview video (Colab is the tool we'll be using for the whole course)
  • Make course resource overview video (e.g. how to use this GitHub, Discussions page, exercises, extra-curriculum etc)
  • Upload solutions for exercises (probably livestream the creation of these after course launch)

Log

  • 28 May 2021 - added future forecasts, added ensemble model, added prediction intervals to time series notebook
  • 25 May 2021 - added multivariate time series to time series notebook, fix LSTM model, next we add TensorFlow windowing/experimenting with window sizes
  • 24 May 2021 - fixed broken preprocessing function in time series notebook, LSTM model is broken, more material to come
  • 20 May 2021 - more time series material creation
  • 19 May 2021 - more time series material creation, streaming much of it live on Twitch - https://twitch.tv/mrdbourke
  • 18 May 2021 - added time series forecasting notebook outline (notebook 10), going to really start ramping up the materials here
  • 12 May 2021 - all videos for 09 have now been released on Udemy & ZTM!!! enjoy build SkimLit πŸ“„ πŸ”₯
  • 11 May 2021 - 40+ section 08 & 09 videos released on Udemy & ZTM!!!
  • 10 May 2021 - time series materials research + preparation
  • 08 May 2021 - time series materials research + preparation
  • 05 May 2021 - ~20+ videos edited for 08, ~10+ videos edited for 09, time series materials in 1st draft mode
  • 04 May 2021 - fixed the remaining videos for 08 (audio missing), now onto making time series materials!
  • 03 May 2021 - rerecorded 10 videos for 08 fixing the sound isse, these are going straight to editing and should be uploaded by end of week
  • 02 May 2021 - found an issue with videos 09-20 of section 08 (no audio), going to rerecord them
  • 29 Apr 2021 - πŸš€ πŸš€ πŸš€ launched on Udemy!!! πŸš€ πŸš€ πŸš€
  • 22 Apr 2021 - finished recording videos for 09! added slides and video notebook 09
  • 21 Apr 2021 - recorded 14 videos for 09! biggggg day of recording! getting closer to finishing 09
  • 20 Apr 2021 - recorded 10 videos for 09
  • 19 Apr 2021 - recorded 9 videos for 09
  • 16 Apr 2021 - slides done for 09, ready to start recording!
  • 15 Apr 2021 - added slides, extra-curriculum, exercises and video notebook for 08, started making slides for 09, will finish tomorrow
  • 14 Apr 2021 - recorded 12 videos for notebook 08, finished the section! time to make slides for 09 and get into it
  • 10 Apr 2021 - recorded 4 videos for notebook 08
  • 9 Apr 2021 - recorded 6 videos for notebook 08
  • 8 Apr 2021 - recorded 10 videos for notebook 08! more coming tomorrow! home stretch baby!!!
  • 7 Apr 2021 - added a whole bunch of images to notebook 08, getting ready for recording tomorrow!
  • 1 Apr 2021 - added notebook 09: SkimLit, almost finished, a little cleaning and we'll be ready for slide making!
  • 31 Mar 2021 - added notebook 08, going to finish tomorrow, then onto 09!
  • 24 Mar 2021 - Recorded 8 videos for 07, finished! onto materials (slides/notebooks) for 08, 09
  • 23 Mar 2021 - Recorded 6 videos for 07 (finally), going to finish tomorrow
  • 22 Mar 2021 - Polished notebook 07 ready for recording, made slides for 07, added template for 07 (for a student to go through and practice), ready to record!
  • 17 Mar 2021 - 99% finished notebook 07, added links to first 14 hours of the course on YouTube (10 hours in part 1, 4 hours in part 2)
  • 11 Mar 2021 - added even more text annotations to notebook 07, finishing tomorrow, then slides
  • 10 Mar 2021 - Typed a whole bunch of explanations into notebook 07, continuing tomorrow
  • 09 Mar 2021 - fixed plenty of code in notebook 07, should run end to end very cleanly (though loading times are still a thing)
  • 05 Mar 2021 - added draft notebook 07 (heaps of data loading and model training improvements in this one!), gonna fix up over next few days
  • 01 Mar 2021 - Added slides for 06 (see them here)
  • 26 Feb 2021 - πŸš€ LAUNCHED!!!!! also finished recording videos for 06, onto 07, 08, 09 for next release
  • 24 Feb 2021 - recorded 9 videos for section 06, launch inbound!!!
  • 23 Feb 2021 - rearranged GitHub in preparation for launch πŸš€
  • 18 Feb 2021 - recorded 8 videos for 05 and... it's done! onto polishing the GitHub
  • 17 Feb 2021 - recorded 10 videos for 05! going to finish tomorrow πŸš€
  • 16 Feb 2021 - polished slides for 05 and started recording videos, got 7 videos done for 05
  • 15 Feb 2021 - finished videos for 04, now preparing to record for 05!
  • 12 Feb 2021 - recored 7 videos for section 04... wanted 10 but we'll take 7 ( πŸ€” this seems to have happened before)
  • 11 Feb 2021 - NO PROGRESS - gave a Machine Learning deployment tutorial for Stanford's CS329s (using the model code from this course!!!) - see the full tutorial materials
  • 08 Feb 2021 - recorded 10 videos for section 03... and section 03 is done! πŸš€ onto section 04
  • 30 Jan 2021 - 07 Feb 2021: NO PROGRESS (working on a ML deployment lecture for Stanford's CS329s... more on this later)
  • 29 Jan 2021 - recorded 9 videos for section 03... closer to 10 than yesterday but still not there
  • 28 Jan 2021 - recorded 7 videos for section 03... wanted 10 but we'll take 7
  • 27 Jan 2021 - recorded 10 videos for section 03
  • 26 Jan 2021 - polished GitHub README (what you're looking at) with a nice table
  • 23 Jan 2021 - finished slides of 06
  • 22 Jan 2021 - finished review of notebook 06 & started slides of 06
  • 21 Jan 2021 - finished slides for 05 & started review of 06
  • 20 Jan 2021 - finished notebook 05 & 95% slides for 05
  • 19 Jan 2021 - found a storage idea for data during course (use Google Storage in same region as Colab Notebooks, cheapest/fastest)
  • 18 Jan 2021 - reviewed notebook 05 & slides for 05
  • 17 Jan 2021 - finished notebook 04 & slides for 04
  • 16 Jan 2021 - review notebook 04 & made slides for transfer learning
  • 13 Jan 2021 - review notebook 03 again & finished slides for 03, BIGGGGG updates to the README, notebook 03 99% done, just need to figure out optimum way to transfer data (e.g. when a student downloads it, where's best to store it in the meantime? Dropbox? S3? GS (too expensive)
  • 11 Jan 2021 - reviewed notebook 03, 95% ready for recording, onto slides for 03
  • 9 Jan 2021 - I'm back baby! Finished all videos for 02, now onto slides/materials for 03, 04, 05 (then I'll get back in the lab)
  • 19 Dec 2020 - ON HOLD (family holiday until Jan 02 2021)
  • 18 Dec 2020 - recorded 75% of videos for 02
  • 17 Dec 2020 - recorded 50% of videos for 02
  • 16 Dec 2020 - recorded 100% of videos for 01
  • 15 Dec 2020 - recorded 90% of videos for 01
  • 09 Dec 2020 - finished recording videos for 00
  • 08 Dec 2020 - recorded 90% of videos for 00
  • 05 Dec 2020 - trialled recording studio for ~6 videos with notebook 00 material
  • 04 Dec 2020 - setup recording studio in closet
  • 03 Dec 2020 - finished notebook 02, finished slides for 02, time to setup recording studio
  • 02 Dec 2020 - notebook 02 95% done, slides for 02 90% done
  • 01 Dec 2020 - added notebook 02 (90% polished), start preparing slides for 02
  • 27 Nov 2020 - polished notebook 01, made slides for notebook 01
  • 26 Nov 2020 - polished notebook 00, made slides for notebook 00
Comments
  • In notebook 5, GlobalAveragePooling2D() does not give the same accuracy and loss

    In notebook 5, GlobalAveragePooling2D() does not give the same accuracy and loss

    Hey Daniel,

    [EfficientNetB0] I tried running the notebook with a slight change where I used ImageDataGenerator and flow_from_dataframe instead of image_dataset_from_directory for data preparation and when I train the same model(EfficientNetB0) as in notebook 5 which uses GlobalAveragePooling2D(). The loss and accuracy are far far very far less and infact stops improving as compared to the losses and accuracy you got while training.

    I thought I had made mistake in my data processing but then I double checked by running the different model from notebook 4(EfficientNetB0 from tf hub) with worked fine cause I saw the loss going down and accuracy improving like it should as demonstrated by you in the notebook 4. So, the data preparation is fine.

    Now, I'm wondering why loss and accuracy is note improving when using the model in notebook 5. Is there something wrong with the model? or data?

    opened by ManojKumarTiwari 6
  • First learning model doesnt work.

    First learning model doesnt work.

    On your first learning model at roughly 4:45:00 into the video. Google colab is having an issue with I believe the model.fit(X, y, epochs=5) .

    I have adjusted the code a few times, trying to debug the system. Though am still very new at this. Any advice would help.

    opened by jcr0269 5
  • 02_neural_network_classification_in_tensorflow - plot_decision_boundary function issue

    02_neural_network_classification_in_tensorflow - plot_decision_boundary function issue

    while building plot_decision_boundary it is giving error as xx and yy not define.

    Even though below is already define.

    x_min,x_max = X[:, 0].min() - 0.1,X[:, 0].max() + 0.1 y_min,y_max = X[:, 1].min() - 0.1,X[:, 1].max() + 0.1 xx, yy = np.meshgrid(np.linspace(x_min,x_max,100), np.linspace(y_min,y_max,100)) Please advise

    opened by prashantsalode 3
  • Update ImageDataGenerator to image_dataset_from_directory

    Update ImageDataGenerator to image_dataset_from_directory

    This will greatly save time for some students instead of waiting for the kernel to finish the run. As a bonus, to teach students how to pipeline the input better. Thanks to your "documentation reading" skills, I have learned to improve things because I hate to wait.

    Current behaviour

    Your code in 03_convolutional_neural_networks_in_tensorflow uses ImageDataGenerator. As you said, this method causes more delay in every epoch because it augments the data when it loaded the model.

    train_datagen_augmented = ImageDataGenerator(rescale=1/255.,
                                                 rotation_range=20, 
                                                 shear_range=0.2,
                                                 zoom_range=0.2, 
                                                 width_shift_range=0.2, 
                                                 height_shift_range=0.2, 
                                                 horizontal_flip=True
    
    
    train_data_augmented = train_datagen_augmented.flow_from_directory(train_dir,
                                                                       target_size=(224, 224),
                                                                       batch_size=32,
                                                                       class_mode='categorical') 
    

    Suggestion

    Since tf.keras.preprocessing is marked as deprecated (but still usable), maybe we can use tf.keras.utils.image_dataset_from_directory as to utilize tf.data.Dataset.

    
    from tensorflow.keras import layers, models
    tf.random.set_seed(42)
    
    train_augmented = tf.keras.utils.image_dataset_from_directory(
        train_dir,
        label_mode='categorical',
        image_size=(256, 256),
        batch_size=32,
        seed=42
    )
    
    preprocessing_layer  = models.Sequential([
        layers.Rescaling(1./255),
        layers.RandomFlip(mode="horizontal"),
        layers.RandomRotation((0.0, 0.01)),
        layers.RandomZoom((0.0, 0.1)),
        layers.RandomTranslation((0.0, 0.1), (0.0, 0.1))
    
    ])
    
    train_augmented = (
        train_augmented
        .map(
            lambda x, y: (preprocessing_layer(x), y),
            num_parallel_calls=tf.data.AUTOTUNE
            )
        .cache()
        .prefetch(buffer_size=tf.data.AUTOTUNE)
    
    )
    
    

    Benchmark

    Tested on multiclass classification with 20 classes without sampling. 15000 training files and 5000 validation files. Although this dataset is not similar to the notebook example, it shows how it can benefit smaller datasets as well.

    ImageDataGenerator takes 11 minutes and 19.4 seconds. image_dataset_from_directory takes 1 minute and 37.5 seconds. Improvement speed 6.97 times over normal run.

    System information

    • OS Platform and Distribution: Windows 10
    • TensorFlow version: v2.8.0
    • GPU model and memory: RTX 3060 Ti 8GB
    opened by nazmi 2
  • 02_neural_network_classification_with_tensorflow - wrong shape in model_3

    02_neural_network_classification_with_tensorflow - wrong shape in model_3

    We are doing binary classification, expecting the predictions to be 0 or 1. For model_2, the shape of model_2.predict(X) is (1000, 1), i.e., for each input row we get a scalar result. However, the first layer in the TF 2.7.0+ tweak to model_3,

    tf.keras.layers.Dense(100, input_shape=(None, 1))

    seems to be wrong: the shape of model_3.predict(X) is (1000, 2, 1), i.e., for each input row we now get 2 numbers. This is also indicated by the function plot_decision_boundary(), saying 'doing multiclass classification...' and showing different contours than those shown in the video.

    Based on the documentation of Dense and Input, I believe that the input_shape should be (2,), i.e., using

    tf.keras.layers.Dense(100, input_shape=(2,))

    or we can just keep the model as it was originally, with

    tf.keras.layers.Dense(100)

    This throws no errors and produces the same plot as shown in the video, with 'doing binary classification'.

    Compare the original plot to the plot corresponding to the tweaked version:

    image

    image

    opened by lenka-cizkova 2
  • plot_decision_boundary for model 7 reads

    plot_decision_boundary for model 7 reads "doing multiclass classification"

    In trying to visualize the metrics for model_7, I ran plot_boundary_decision(model_7, X, y) and it prints doing multiclass classification. I've tried adding the input_shape parameter to the first layer, as well as expand_dims in the model_7.fit function call, and these result in errors that logits and labels have to be the same dimension.

    The plot drawn is all orange and doesn't look like the video, but the data in the model is the same as the video.

    opened by pekasus 2
  • Incorrect behaviour of plot_decision_boundary() in 02_neural_network_classification_with_tensorflow.ipynb

    Incorrect behaviour of plot_decision_boundary() in 02_neural_network_classification_with_tensorflow.ipynb

    Hello,

    I am working through video 76 on the "TensorFlow Developer Certificate in 2022: Zero to Mastery" course. The function plot_decision_boundary() has the following logic:

    if len(y_pred[0]) > 1:
        print("doing multiclass classification...")
        # We have to reshape our predictions to get them ready for plotting
        y_pred = np.argmax(y_pred, axis=1).reshape(xx.shape)
      else:
        print("doing binary classifcation...")
        y_pred = np.round(y_pred).reshape(xx.shape) #rounding to give us a 0 or 1
    

    When I check len(y_pred[0]) which comes from y_pred = model.predict(x_in), I get a length of 2. So the current function then follows the logic for multi-class classification, although the problem being worked on is binary. Can this be fixed?

    Cheers! Donal

    opened by donalhill 2
  • plot_and_pred function

    plot_and_pred function

    In the explanation for binary image classification https://github.com/mrdbourke/tensorflow-deep-learning/blob/main/03_convolutional_neural_networks_in_tensorflow.ipynb you defined class_names which were later used in pred_and_plot function to round the probability >0.5 to 1 and <0.5 to 0 to refer to steak and pizza, respectively. But unfortunately the references are 1 for steak and 2 for pizza due to ".DS_store" file at index 0.

    If you look at the prediction (pizza_dad.jpeg) right before the start of the section on multi class classification, it predicts ".DS_store" where it should've been pizza, and the prediction above that is given pizza where it should be steak - The model is predicting right but comes out wrong due to class_names reference. Perhaps class_names = class_names[1:] will help.

    opened by Ishticode 2
  • mixed precision with TPU

    mixed precision with TPU

    Hello,

    On colab, I don't have the GPU with minimum required compute capability for mixed precision. So I decided to use TPU and set the policy as 'mixed_bfloat16'. However, while fitting the feature extraction model, I get the following error. I am not able to find solutions online as well. I am also not sure if this is because I am using TPU or if there is something I am missing. Any help would be great.This is the error:

    /usr/local/lib/python3.7/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument. category=CustomMaskWarning) Epoch 1/2

    InvalidArgumentError Traceback (most recent call last) in () ----> 1 hist1 = model.fit(train_data,steps_per_epoch=len(train_data),epochs=2, validation_data=test_data,validation_steps=int(0.15*len(test_data)),callbacks = [callback1,callback2])

    6 frames /usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name) 58 ctx.ensure_initialized() 59 tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name, ---> 60 inputs, attrs, num_outputs) 61 except core._NotOkStatusException as e: 62 if name is not None:

    InvalidArgumentError: No OpKernel was registered to support Op 'DepthwiseConv2dNativeBackpropInput' used by {{node gradient_tape/model_7/efficientnetb0/block7a_dwconv/depthwise/DepthwiseConv2dNativeBackpropInput}} with these attrs: [dilations=[1, 1, 1, 1], T=DT_BFLOAT16, data_format="NHWC", strides=[1, 1, 1, 1], explicit_paddings=[], padding="SAME"] Registered devices: [CPU] Registered kernels: device='XLA_CPU_JIT'; T in [DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF] device='XLA_GPU_JIT'; T in [DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF] device='GPU'; T in [DT_DOUBLE] device='GPU'; T in [DT_FLOAT] device='GPU'; T in [DT_HALF] device='CPU'; T in [DT_DOUBLE] device='CPU'; T in [DT_FLOAT] device='CPU'; T in [DT_HALF] device='GPU'; label='cudnn_grouped_convolution'; T in [DT_DOUBLE] device='GPU'; label='cudnn_grouped_convolution'; T in [DT_FLOAT] device='GPU'; label='cudnn_grouped_convolution'; T in [DT_HALF]

     [[gradient_tape/model_7/efficientnetb0/block7a_dwconv/depthwise/DepthwiseConv2dNativeBackpropInput]] [Op:__inference_train_function_197064]
    

    I get the same error with different versions of EfficientNet.

    opened by gprs1809 2
  • Update πŸ›  00_TensorFlow_Fundamentals_πŸ› _Exercises_&_πŸ“–_Extra_curriculum.…

    Update πŸ›  00_TensorFlow_Fundamentals_πŸ› _Exercises_&_πŸ“–_Extra_curriculum.…

    …ipynb

    Dimensions correction for scalar and tensor objects According to this video "14. Creating your first tensors with TensorFlow and tf.constant()" scalar dimension is 0 and vector dimension is 1 and matrix dimension is 2 and tensor dimension is equal or greater than 3

    opened by MahmoudHassan 2
  • Data augmentation - rotation_range is an integer not a float

    Data augmentation - rotation_range is an integer not a float

    In the Computer Vision and Convolutional Neural Network and data augmentation sections

    https://github.com/mrdbourke/tensorflow-deep-learning/blob/main/03_convolutional_neural_networks_in_tensorflow.ipynb

    we should emphasize that rotation_range is an int. Putting 0.2 equates finally to zero

    opened by vincestorm 2
  •  ValueError: logits and labels must have the same shape, received ((None, 15, 1) vs (None,)).

    ValueError: logits and labels must have the same shape, received ((None, 15, 1) vs (None,)).

    inputs = tf.keras.layers.Input(shape=(1,), dtype=tf.string)
    x = text_vectorizer(inputs)
    x = embedding(x)  # Create an embed
    outputs = tf.keras.layers.Dense(1, activation="sigmoid")(x)
    model_1 = tf.keras.Model(inputs, outputs, name="model_1_dense")
    model_1.compile(loss="binary_crossentropy",
                    optimizer=tf.keras.optimizers.Adam(), metrics=["accuracy"])
    
    model_1_history = model_1.fit(x=train_sentences, y=train_labels, epochs=5, validation_data=(
        val_sentences, val_labels), callbacks=[create_tensorboard_callback(dir_name="model_logs", experiment_name="model_1_dense")])
    

    this works in the video but is broken in TensorFlow 2.90, expanding solves it, but I don't understand why. might be worth adding a note here? https://academy.zerotomastery.io/courses/learn-tensorflow/lectures/32197595

    In the later part of the video : x = tf.keras.layers.GlobalAveragePooling1D()(x) is added which will solve the problem but it's not mentioned

    opened by latheeshvm 0
  • Speed of fitting model is hours for some reason, 07_food_vision_milestone_project. Also Found a resolution for the mixed precision float16 - float32 error

    Speed of fitting model is hours for some reason, 07_food_vision_milestone_project. Also Found a resolution for the mixed precision float16 - float32 error

      Also running into a problem. It appears that the ETA for fitting the model after downgrading to 2.4.1 is running at like 100 hours even with using the Nvida T4. Even running the project directly from github https://github.com/mrdbourke/tensorflow-deep-learning/blob/main/07_food_vision_milestone_project_1.ipynb it still is getting 100+ hours
    

    Saving TensorBoard log files to: training_logs/efficientnetb0_101_classes_all_data_feature_extract/20221227-085937 Epoch 1/3 3/2368 [..............................] - ETA: 97:10:20 - loss: nan - accuracy: 0.0382

    Tue Dec 27 08:59:33 2022
    +-----------------------------------------------------------------------------+ | NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |===============================+======================+======================| | 0 Tesla T4 Off | 00000000:00:04.0 Off | 0 | | N/A 66C P0 31W / 70W | 3MiB / 15109MiB | 0% Default | | | | N/A | +-------------------------------+----------------------+----------------------+

    +-----------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=============================================================================| | No running processes found | +-----------------------------------------------------------------------------+

    Turn on mixed precision training

    from tensorflow.keras import mixed_precision mixed_precision.set_global_policy(policy="mixed_float16") # set global policy to mixed precision

    WARNING:tensorflow:Mixed precision compatibility check (mixed_float16): WARNING The dtype policy mixed_float16 may run slowly because this machine does not have a GPU. Only Nvidia GPUs with compute capability of at least 7.0 run quickly with mixed_float16. If you will use compatible GPU(s) not attached to this host, e.g. by running a multi-worker model, you can ignore this warning. This message will only be logged once

    Originally posted by @Jayms8462 in https://github.com/mrdbourke/tensorflow-deep-learning/discussions/82#discussioncomment-4501569

    opened by Jayms8462 2
  • Question about model 4 conv1d in notebook 10(time series prediction for bitcoin price)

    Question about model 4 conv1d in notebook 10(time series prediction for bitcoin price)

    This is the model summary in model 4

    Model: "model_4_conv1D"
    _________________________________________________________________
    Layer (type)                 Output Shape              Param #   
    =================================================================
    lambda_1 (Lambda)            (None, 1, 7)              0         
    _________________________________________________________________
    conv1d (Conv1D)              (None, 1, 128)            4608      
    _________________________________________________________________
    dense_6 (Dense)              (None, 1, 1)              129       
    =================================================================
    Total params: 4,737
    Trainable params: 4,737
    Non-trainable params: 0
    

    Don't we want input to conv1d layer have shape of (batch_size, timesteps(window size), input_dim) ?

    opened by tsinggggg 0
  • 07_food_vision_milestone_project_1.ipynb - In:29

    07_food_vision_milestone_project_1.ipynb - In:29

    So when I run cell 29 I get the following error. The GPU is "GPU 0: Tesla T4 (UUID: GPU-6b3d416d-69cc-a08d-bde3-f0e6c29fdd87)".


    TypeError Traceback (most recent call last)

    in 3 # Create base model 4 input_shape = (224, 224, 3) ----> 5 base_model = tf.keras.applications.EfficientNetB0(include_top=False) 6 base_model.trainable = False # freeze base model layers 7

    4 frames

    /usr/local/lib/python3.8/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs) 65 except Exception as e: # pylint: disable=broad-except 66 filtered_tb = _process_traceback_frames(e.traceback) ---> 67 raise e.with_traceback(filtered_tb) from None 68 finally: 69 del filtered_tb

    TypeError: Exception encountered when calling layer "tf.math.truediv" (type TFOpLambda).

    x and y must have the same dtype, got tf.float16 != tf.float32.

    Call arguments received by layer "tf.math.truediv" (type TFOpLambda): β€’ x=tf.Tensor(shape=(None, None, None, 3), dtype=float16) β€’ y=tf.Tensor(shape=(3,), dtype=float32) β€’ name=None

    opened by vl0 0
  • Error in number of epochs in notebook tensorflow-deep-learning/01_neural_network_regression_in_tensorflow.ipynb

    Error in number of epochs in notebook tensorflow-deep-learning/01_neural_network_regression_in_tensorflow.ipynb

    Under the markdown that has the text "Let's train the same model as above for a little longer. We can do this but calling fit on it again."

    The number of epochs should be 200 not 100

    # Try training for a little longer (100 more epochs) history_2 = insurance_model_2.fit(X_train, y_train, epochs=100, verbose=0)

    should be changed to this:

    # Try training for a little longer (100 more epochs) history_2 = insurance_model_2.fit(X_train, y_train, epochs=200, verbose=0)

    opened by jhchang 0
Owner
Daniel Bourke
Machine Learning Engineer live on YouTube.
Daniel Bourke
Pre-trained model, code, and materials from the paper "Impact of Adversarial Examples on Deep Learning Models for Biomedical Image Segmentation" (MICCAI 2019).

Adaptive Segmentation Mask Attack This repository contains the implementation of the Adaptive Segmentation Mask Attack (ASMA), a targeted adversarial

Utku Ozbulak 53 Jul 4, 2022
Useful materials and tutorials for 110-1 NTU DBME5028 (Application of Deep Learning in Medical Imaging)

Useful materials and tutorials for 110-1 NTU DBME5028 (Application of Deep Learning in Medical Imaging)

null 7 Jun 22, 2022
PyTorch implementation of 1712.06087 "Zero-Shot" Super-Resolution using Deep Internal Learning

Unofficial PyTorch implementation of "Zero-Shot" Super-Resolution using Deep Internal Learning Unofficial Implementation of 1712.06087 "Zero-Shot" Sup

Jacob Gildenblat 196 Nov 27, 2022
Free Book about Deep-Learning approaches for Chess (like AlphaZero, Leela Chess Zero and Stockfish NNUE)

Free Book about Deep-Learning approaches for Chess (like AlphaZero, Leela Chess Zero and Stockfish NNUE)

Dominik Klein 189 Dec 21, 2022
Computer Vision Script to recognize first person motion, developed as final project for the course "Machine Learning and Deep Learning"

Overview of The Code BaseColab/MLDL_FPAR.pdf: it contains the full explanation of our work Base Colab: it contains the base colab used to perform all

Simone Papicchio 4 Jul 16, 2022
This repo contains research materials released by members of the Google Brain team in Tokyo.

Brain Tokyo Workshop ?? ?? This repo contains research materials released by members of the Google Brain team in Tokyo. Past Projects Weight Agnostic

Google 1.2k Jan 2, 2023
The materials used in the SaxonJS tutorial presented at Declarative Amsterdam, 2021

SaxonJS-Tutorial-2021, version 1.0.4 Last updated on 4 November, 2021. Table of contents Background Prerequisites Starting a web server Running a Java

Saxonica 11 Oct 23, 2022
An SE(3)-invariant autoencoder for generating the periodic structure of materials

Crystal Diffusion Variational AutoEncoder This software implementes Crystal Diffusion Variational AutoEncoder (CDVAE), which generates the periodic st

Tian Xie 94 Dec 10, 2022
Workshop Materials Delivered on 28/02/2022

intro-to-cnn-p1 Repo for hosting workshop materials delivered on 28/02/2022 Questions you will answer in this workshop Learning Objectives What are co

Beginners Machine Learning 5 Feb 28, 2022
Official code for the CVPR 2022 (oral) paper "Extracting Triangular 3D Models, Materials, and Lighting From Images".

nvdiffrec Joint optimization of topology, materials and lighting from multi-view image observations as described in the paper Extracting Triangular 3D

NVIDIA Research Projects 1.4k Jan 1, 2023
TensorFlow Ranking is a library for Learning-to-Rank (LTR) techniques on the TensorFlow platform

TensorFlow Ranking is a library for Learning-to-Rank (LTR) techniques on the TensorFlow platform

null 2.6k Jan 4, 2023
All the code and files related to the MI-Lab of UE19CS305 course in sem 5

Machine-Intelligence-Lab-CS305 The compilation of all the code an drelated files from MI-Lab UE19CS305 (of batch 2019-2023) offered by PES University

Arvind Krishna 3 Nov 10, 2022
Zero-shot Synthesis with Group-Supervised Learning (ICLR 2021 paper)

GSL - Zero-shot Synthesis with Group-Supervised Learning Figure: Zero-shot synthesis performance of our method with different dataset (iLab-20M, RaFD,

Andy_Ge 62 Dec 21, 2022
Official Pytorch Implementation of: "Semantic Diversity Learning for Zero-Shot Multi-label Classification"(2021) paper

Semantic Diversity Learning for Zero-Shot Multi-label Classification Paper Official PyTorch Implementation Avi Ben-Cohen, Nadav Zamir, Emanuel Ben Bar

null 28 Aug 29, 2022
Shared Attention for Multi-label Zero-shot Learning

Shared Attention for Multi-label Zero-shot Learning Overview This repository contains the implementation of Shared Attention for Multi-label Zero-shot

dathuynh 26 Dec 14, 2022
[ICCV 2021] Official Pytorch implementation for Discriminative Region-based Multi-Label Zero-Shot Learning SOTA results on NUS-WIDE and OpenImages

Discriminative Region-based Multi-Label Zero-Shot Learning (ICCV 2021) [arXiv][Project page >> coming soon] Sanath Narayan*, Akshita Gupta*, Salman Kh

Akshita Gupta 54 Nov 21, 2022
[ICCV 2021] Official Pytorch implementation for Discriminative Region-based Multi-Label Zero-Shot Learning SOTA results on NUS-WIDE and OpenImages

Discriminative Region-based Multi-Label Zero-Shot Learning (ICCV 2021) [arXiv][Project page >> coming soon] Sanath Narayan*, Akshita Gupta*, Salman Kh

Akshita Gupta 54 Nov 21, 2022
EMNLP 2021 Adapting Language Models for Zero-shot Learning by Meta-tuning on Dataset and Prompt Collections

Adapting Language Models for Zero-shot Learning by Meta-tuning on Dataset and Prompt Collections Ruiqi Zhong, Kristy Lee*, Zheng Zhang*, Dan Klein EMN

Ruiqi Zhong 42 Nov 3, 2022