Get a glimpse of how similar/different these libraries are: Pytorch vs Tensorflow on MNIST dataset
In each notebook, we shall train using free Google Colab resources and eventually deploy them as gradio/streamlit app (depending on projects).
- Tensorflow Fundamentals TF Tensors Basics
- Constants and Variables
- Compatibility with Numpy
- Random Generators
- Basic Operations
- Pytorch Fundamentals PT Tensors Basics
- Tensor Basic
- Interoperability with Numpy
- Basic Operations
- Regression Model Training with Custom Data on GPU
- Regression - Custom TF Model on Medical Insurance Dataset
- Minimal EDA
- k-Fold Cross Validation
- L1 Regularizers
- Gradio App
-
Image Classification - Custom TF Model on Cifar10 Dataset
- Image Augmentation
- LR Finder
- One-Cycle LR Scheduler
- GradCAM visualisation
- Gradio App
-
Multi-Label Image Classification - TF Transfer Learning on Custom Dataset
- Custom Dataset
- TF Record with Image Augmentation
- Custom Loss Function
- Transfer Learning
- Performance Profiling
- Gradio App
-
Image Generation - TF VAE Image Generation on Celeb Faces
- Custom Architecture using Probabilistic Layers
- Reduce LR on Plateau Scheduler
- New Generated Faces
- Reconstructing Faces
- Feature Manipulation
- Face Morphing
- Visualize clusters on UMAP-reduced 1D latent vector
-
Metric Learning - TF Similarity Models on Dog-Cat Breed Dataset
- Tensorflow Similarity
- Transfer Learning with an embedding layer and Multisimilarity loss
- ANN Search: Indexing, Calibration, Querying
- Precision-Recall Curve
- UMAP-reduced clustering with interactive visualization
-
Image Translation - TF Pix2Pix on Edges-to-Handbags Dataset
- Understanding Pix2Pix Architecture
- Training it from scratch with additional loss fucntion
- Focal Frequency Loss
- Using Tensorboard during model training
- Image Generation
-
Image Segmentation - HF SegFormer on Road-Sidewalk Dataset
- Understanding Semantic Segmentation using Transformers
- Fine-Tuning it using Huggingface Modules
- Mean IOU metric
- Publishing as HF Model
- Live Inference Model
-
Object Detection - PT YOLOS on Matterport Balloons Dataset
- Understanding Object Detection using Transformers
- Fine-Tuning YOLOS using Pytorch Lightning
- Detecting object on a video
- Viewing Attention Layers
- Publishing as HF Model
- Live Inference Model
- Pre-Neural NLP - Heuristics-based & Statistical Methods in NLP
- Basics of Sentiment Analysis
- Valence Aware Dictionary and Sentiment Reasoner (VADER)
- Support Vector Machines (SVM)
- Grid Search for Hyperparameters
- ROC Curve
- Understanding Vanilla Transformers - Vanilla Transformers
- Understanding Seq2Seq Models
- Understanding Attention Mechanism
- Understanding Transformer Architecture
- Vanilla Transformer Comment to Code - PT Train Vanilla Transformer (Sequence to Sequence)
- Dataset Augmentation
- Custom Tokenizer
- Build Complete Transformer Architecture
- Custom Loss
- Display Attention
- Gradio App
- Stable Diffusion - HF Stable Diffusion Text to Image
- Understanding Diffusion Models (Stable diffusion in particular)
- Exploring Diffusers Library
- Writing an inference pipeline
- Understanding the complete generative process during inference
-
JAX Basics - JAX Basics
- Why JAX?
- How randomness is handled
- Speed Comparison
- Asynchronous Dispatch
- JIT Compilation
- Auto-differentiation with grad
- Auto-vectorization with Vmap
- SPMD Programming with Pmap on TPU
- Device Memory Profiler
-
PySyft - Secure and Privacy AI
- Data Owner
- Data Scientist
- Differential Privacy
- Remote Data Science
- Covid-19 trends prediction
-
TenSeal - Homomorphic Encryption on Tensors
- Homomorphic Encryption Basic and Encrypted Inference
- Homomorphic Encryption NN Training
- Tenseal Context
- Basic Mathematical Operations on Encrypted Tensors
- Encrypted Evaluation on Encrypted Test Data
- Training Encrypted NN on Encrypted Data