H2O Hydrogen Torch Starter Course

Name: H2O Hydrogen Torch Starter Course
Rating: 4.6 (12 reviews)

Master Deep Learning with H2O Hydrogen Torch: No-Code AI Models in Computer Vision, NLP, and Audio

Created byH2O.ai University, Andreea Turcu

Last updated 7/2024

English

What you'll learn

Craft robust deep learning models in computer vision, NLP, and audio without coding.
Master the experiment flow in H2O Hydrogen Torch, from importing datasets to observing completed experiments.
Gain practical knowledge in model tuning with grid search and identify optimal hyperparameters.
Understand the principles of deep learning through hands-on assignments and real-world applications.

Course content

6 sections • 22 lectures • 1h 6m total length

Class Intro1:56
What is H2O Hydrogen Torch2:25
Discover H2O Hydrogen Torch, an advanced tool that streamlines deep learning training with predefined and tunable hyperparameters, offering interactive charts to visualize impact and deployment to MLOps or Python environments.
H2O Hydrogen Torch use cases4:42
H2O Hydrogen Torch empowers data scientists to tackle computer vision, natural language processing, and audio tasks; access 100+ use cases and datasets in the H2O catalog.
Access H2O Hydrogen Torch1:33
Connect to the HAIC platform via the H2O AI cloud, launch an H2O Hydrogen Torch instance, and train state-of-the-art deep neural networks on diverse problem types.

Experiment flow2:26
Practice the H2O hydrogen torch workflow from dataset import to model deployment. Analyze training with hyperparameters and grid search, and inspect models with interactive graphs.
Explore the Home Page4:19
Explore the H2O Hydrogen Torch home page to access documentation, manage datasets and experiments, customize settings, run grid search, and generate predictions.
Start your first experiment2:01

Step 1 - Import dataset3:04
Import a preprocessed coins image regression dataset in H2O Hydrogen Torch, including 6,028-image coins set, and configure dataset settings to train a model that sums coin values in Brazilian real.
Step 2 - Explore dataset1:41
Explore the coins_image_regression dataset, view sample train data with sum labels, and build, train, and tune a model with h2o hydrogen torch to predict total coin value.
Step 3 - Build model (experiment)3:01
Build your first image regression model in h2o hydrogen torch by creating an experiment from the coins_image_regression dataset, using default hyperparameters and mae scorer, aiming for a zero validation score.
Step 4 - Observe running experiment2:23
Observe the image regression model's performance using interactive charts, track validation MAE as it decreases, and explore validation prediction insights and sample losses to assess accuracy.
Step 5 - Observe completed experiment4:53
Observe the completed experiment to view the final validation metric and best validation samples. Tune hyperparameters with grid search to improve MAE for an ATM coin counter, noting dark images.
Summary on starting a first experiment2:58
Welcome to your first assignment in the Hydrogen Torch Starter Course!
In this assignment, you will build on the concepts we've covered so far by starting your own experiment using H2O Hydrogen Torch. Your task is to use a new dataset, specifically the "flower_image_classification.zip" file located in our AWS S3 source. You have approximately 15 to 20 minutes to complete this task.
By completing this assignment, you will:
Gain hands-on experience with H2O Hydrogen Torch by setting up an experiment from scratch.
Develop the ability to navigate and utilize datasets from AWS S3.
Apply the foundational concepts learned in the course to a practical, real-world example.

Model tuning with grid search1:43
Explore grid search in H2O Hydrogen Torch to tune an image metric learning model that assesses bicycle image similarity, improving hyperparameter optimization and model performance.
Step 1 - Import and explore dataset3:37
Import and explore the preprocessed bicycle image metric learning dataset in H2O Hydrogen Torch, review data configuration, labels, and image columns, and validate samples with the train visualization before proceeding.
Step 2 - Build model1:33
Build an image metric learning model using default hyperparameters by running an experiment on the bicycle_image_metric_learning dataset, then monitor the queued and completed results to optimize mean average precision.
Step 3 - Identify hyperparameters2:28
Identify and tune key hyperparameters to boost mAP in object detection, focusing on backbone and learning rate, and explore grid search to optimize model performance.
Step 4 - Grid Search2:36
Explore hyperparameter tuning with grid search in H2O Hydrogen Torch, automatically exploring combinations of learning rate, hidden layers, and regularization to identify optimal model configurations.
Step 5 - Rebuild model10:17
Rebuild the model by tuning backbone architectures, embedding sizes (256 and 1024), and learning rates (0.0005 and 0.01) through grid search to determine the best configuration.
Step 6 - Explore new best-rebuilt model3:21
Welcome to your second assignment in the Hydrogen Torch Starter Course!
In this assignment, your goal is to enhance your model's performance by tuning the hyperparameters of the best-performing experiment we created together. You will use the "bicycle_image_metric_learning.zip" dataset for this task.
By completing this assignment, you will:
Learn to improve model performance by tuning hyperparameters.
Gain experience with sorting and selecting experiments based on validation metrics.
Understand how different hyperparameter settings can impact model results.
Step 7 - Best experiment settings1:07
Welcome to the final assignment of the Hydrogen Torch Starter Course!
In this challenge, you'll aim to surpass the best experiment score from Assignment 1 using the "flower_image_classification.zip" dataset. Unlike Assignment 2, we’ll use the Custom grid search mode to manually identify and tune hyperparameters.

Your objective is to adjust the hyperparameters to fine-tune the model and achieve improved image classification results. After tuning, analyze and compare the performance of your new model to the initial one to measure the improvement.

By completing this final assignment, you will:
Enhance your skills in manually tuning hyperparameters.
Gain deeper insights into the impact of various hyperparameters on model performance.
Achieve a better understanding of how to optimize models for superior results.

Requirements

No programming experience needed. You will learn everything you need to know.

Description

Welcome to the H2O Hydrogen Torch Starter Course!

This thrilling course, part of the H2O University and Certification Program, is designed to make cutting-edge AI accessible to everyone, regardless of coding experience. Whether you're a beginner or a seasoned data scientist, this course offers valuable insights into creating robust models in computer vision, natural language processing (NLP), and audio.

Andreea Turcu, Head of Global Training at H2O ai, guides you through the Hydrogen Torch Starter Course, providing a hands-on learning experience that covers the entire experiment flow. You'll start by importing and exploring datasets, followed by building and tuning models using grid search to identify optimal hyperparameters. The course emphasizes practical knowledge, allowing you to observe running experiments, explore completed ones, and understand the underlying principles of deep learning.

One of the standout features of this course is its focus on real-world applications. Learn from the best practices of Kaggle competitions and apply these techniques to your projects. By the end of the course, you'll have the skills to craft sophisticated deep learning models without the need for coding.

Upon completion, you will earn a Certificate of Completion from H2O University, which you can proudly showcase on LinkedIn. This certification not only validates your proficiency in deep learning but also positions you as a frontrunner in the dynamic field of AI and data science. Join us in this exciting journey and unlock the potential of deep learning with H2O Hydrogen Torch.

Who this course is for:

Data scientists and machine learning enthusiasts looking to expand their knowledge in computer vision, NLP, and audio.
Professionals seeking to leverage Kaggle best practices in their projects.
Beginners in deep learning and AI who are eager to learn without the need for coding.

H2O Hydrogen Torch Starter Course

What you'll learn

Explore related topics

Course content

Introduction to Hydrogen Torch4 lectures • 11min

Getting Started with H2O Hydrogen Torch3 lectures • 9min

Conducting Experiments6 lectures • 18min

Advanced Model Tuning8 lectures • 27min

Course Wrap Up1 lecture • 2min

Course Completion Quiz0

Requirements

Description

Who this course is for: