Open-Source Spotlight - Super Gradients - Harpreet Sahota
Super Gradients: Build, train, and fine-tune production-ready deep learning SOTA vision models
00:00 Intro
01:30 Intro to SuperGradients
03:06 Why Use SuperGradients
05:30 Model Zoo
09:50 5 Steps to Train Your Model
10:50 Hands on Transfer Learning with SuperGradients
11:45 The Dataset We're Using
13:00 Initialize the SuperGradients Trainer
14:15 A Word About Image Augmentations
15:00 Architecture Definition
16:30 Training Recipe
18:00 Training and Evaluation
20:00 Label Smoothing Cross Entropy
23:00 The SuperGradients Repo
25:20 Career/Life Advice from Harpreet
27:00 The Last Epoch Has Trained and Let's Examine Results
Links:
GitHub repo: https://github.com/Deci-AI/super-grad...
ResNet in Action with SuperGradients: https://colab.research.google.com/dri...
EfficientNet in Action with SuperGradients: https://colab.research.google.com/dri...
RegNet in Action with SuperGradients: https://colab.research.google.com/dri...
Using the EfficientNet Architecture to Classify Car Brands: / using-the-efficientnet-architecture-to-cla...
ResNet: How One Paper Changed Deep Learning Forever: https://heartbeat.comet.ml/resnet-how...
A Better Way to Train Your Neural Networks: https://heartbeat.comet.ml/a-better-w...
Kaggle profile, with lots of SuperGradients examples: https://www.kaggle.com/harpdeci/code
ML Zoomcamp: https://github.com/alexeygrigorev/mlb...
Join DataTalks.Club: https://datatalks.club/slack.html
Our events: https://datatalks.club/events.html