Authoring

Authored the following liveProjects with Manning:

• Use Machine Learning to Detect Phishing Websites
• Summarize News Articles with NLP and TensorFlow (with Souradip Chakraborty)

Below are the blogs, articles, and tutorials I have written on Data Science, Machine Learning, and more. I am fortunate enough to collaborate with amazing folks from all around the globe. I am grateful to the GDE Program (ML Developer Programs team) that provides me with Google Cloud Platform credits which I use to run various experiments for my own curiosity and for blog posts.

Keras

Some favorites:

• Knowledge distillation recipes ¹
• Point cloud segmentation with PointNet (joint work with Soumik Rakshit) ²
• Handwriting recognition (joint work with Aakash Kumar Nain)
• Vector-Quantized Variational Autoencoders
• Distilling Vision Transformers ³
• Investigating Vision Transformer representations (joint work with Aritra Roy Gosthipaty) ⁴

NoteOthers

• MixUp augmentation for image classification
• RandAugment for Image Classification for Improved Robustness
• Self-supervised contrastive learning with SimSiam
• Consistency Training with Supervision
• Learning to Resize in Computer Vision
• Keypoint Detection with Transfer Learning
• Video Classification with a CNN-RNN Architecture
• Video Classification with Transformers
• Semi-supervision and domain adaptation with AdaMatch
• Compact Convolutional Transformers
• Conditional GAN
• Multimodal entailment
• Near-duplicate image search
• Large-scale multi-label text classification (joint work with Soumik Rakshit)
• FixRes: Fixing train-test resolution discrepancy
• Image classification with ConvMixer
• MobileViT: A mobile-friendly Transformer-based model for image classification
• Masked image modeling with Autoencoders (joint work with Aritra Roy Gosthipaty)
• Learning to tokenize in Vision Transformers (joint work with Aritra Roy Gosthipaty)
• GauGAN for conditional image generation (joint work with Soumik Rakshit)
• Class Attention Image Transformers with LayerScale
• Fine-tuning Stable Diffusion (joint work with Chansung Park)
• Semantic segmentation with SegFormer and Hugging Face Transformers
• DreamBooth (joint work with Chansung Park)
• Training a language model from scratch with 🤗 Transformers and TPUs (joint work with Matthew Carrigan)
• Segment Anything Model with 🤗 Transformers (joint work with Merve Noyan)

Datacamp

Some favorites:

• Demystifying Crucial Statistics in Python ⁵
• Turning Machine Learning Models into APIs in Python ⁶
• Essentials of Linear Regression in Python ⁷

NoteOthers

• Simplifying Sentiment Analysis in Python
• Introduction to Indexing in SQL
• Understanding Recursive Functions in Python
• Beginner’s Guide to Google’s Vision API in Python
• Beginner’s Guide to PostgreSQL
• Managing Databases in PostgreSQL
• Working with Spreadsheets in SQL
• Installing PostgreSQL on Windows and Mac OS X
• Using Order By Keyword in SQL
• Introduction to Alter Table Statement in SQL
• SQLite in Python
• Introduction to Where Clause in SQL
• Introduction to SQL Joins
• 10 command-line utilities in PostgreSQL
• CASE Statements in PostgreSQL
• Cleaning Data in SQL
• Materialized Views in PostgreSQL
• Argument Parsing in Python
• Ten Important Updates from TensorFlow 2.0
• Implementing Neural Style Transfer using TensorFlow 2.0
• TensorFlow 2.0 Case Study

FloydHub

Some favorites:

• An introduction to Q-Learning: Reinforcement Learning
• How to plan and execute your ML and DL projects

NoteOthers

• Introduction to Anomaly Detection in Python
• Introduction to K-Means Clustering in Python with scikit-learn
• Becoming One With the Data ⁸
• Training Neural Nets: a Hacker’s Perspective

Weights and Biases

Some favorites:

• Understanding the Effectivity of Ensembles in Deep Learning (joint work with Ayush Thakur)
• An Introduction to Adversarial Examples in Deep Learning ⁹
• Distilling Knowledge in Neural Networks
• Keras XLA Benchmarks (in collaboration with Soumik Rakshit and Ayush Thakur)

NoteOthers

• Running Hyperparameter Sweeps to Pick the Best Model
• arXiv Search: Generating Tags from Paper Titles
• How to Use GCP with Weights & Biases
• Mixed precision training with tf.keras
• Customizing Training Loops in TensorFlow 2.0
• Bayesian Hyperparameter Optimization - A Primer
• Visualize models in TensorBoard with Weights and Biases
• The effects of weight initialization on neural nets
• Introduction to image inpainting with deep learning (joint work with Ayush Thakur)
• Reproducible Models with W&B
• EvoNorm layers in TensorFlow 2
• A Tale of Model Quantization in TF Lite
• Towards self-supervised image understanding with SimCLR
• The Power of Random Features of a CNN
• Plotting top loss images while training models (joint work with Tulasi)
• Improving Image Classifiers with Supervised Contrastive Learning (joint work with Sweta Shaw)
• Model Pruning in Deep Learning
• Understanding the Effectivity of Ensembles in Deep Learning (joint work with Ayush Thakur)
• An Introduction to Adversarial Examples in Deep Learning ¹⁰
• Unsupervised Visual Representation Learning with SwAV (joint work with Ayush Thakur)

Google Cloud Platform

• Streamline your ML training workflow with Vertex AI (joint work with Karl Weinmeister of Google)
• Image search with natural language queries (joint work with Chansung Park)
• Dual deployments on Vertex AI (joint work with Chansung Park) ¹¹
• Model training as a CI/CD system - Part I, Part II (joint work with Chansung Park)

TensorFlow

Some favorites:

• How to Create a Cartoonizer with TensorFlow Lite (joint work with Margaret Maynard-Reid)
• Load-testing TensorFlow Serving’s REST Interface (joint work with Chansung Park)
• Serving With TF and GKE: Stable Diffusion (joint work with Chansung Park)

NoteOthers

• TensorFlow Addons Optimizers: CyclicalLearningRate
• Graph regularization for image classification using synthesized graphs ¹²
• Continuous Adaptation for Machine Learning System to Data Changes (joint work with Chansung Park) ¹³
• Automated Deployment of TensorFlow Models with TensorFlow Serving and GitHub Actions (joint work with Chansung Park)
• End-to-End Pipeline for Segmentation with TFX, Google Cloud, and Hugging Face (joint work with Chansung Park) ¹⁴

Hugging Face 🤗

NoteSeries on the deployment of TF vision models in 🤗

• Deploying TensorFlow Vision Models in Hugging Face with TF Serving
• Deploying 🤗 ViT on Kubernetes with TF Serving (with Chansung Park)
• Deploying 🤗 ViT on Vertex AI (with Chansung Park)

NoteSeries on Diffusers x quantization

• Memory-efficient Diffusion Transformers with Quanto and Diffusers
• In our previous post, Exploring Quantization Backends in Diffusers
• (LoRA) Fine-Tuning FLUX.1-dev on Consumer Hardware

You can check out the articles I have written or co-written on the hf.co/blog platform by visiting this link. Below are some of my favorite ones:

• The State of Computer Vision at Hugging Face 🤗
• A Dive into Vision-Language Models (with Alara Dirik)
• 🤗 PEFT: Parameter-Efficient Fine-Tuning of Billion-Scale Models on Low-Resource Hardware (with Sourab Mangrulkar)
• Instruction-tuning Stable Diffusion with InstructPix2Pix
• Personal Copilot: Train Your Own Coding Assistant (with Sourab Mangrulkar)
• 🤗 PEFT welcomes new merging methods (with Sourab Mangrulkar).
• State of open video generation models in Diffusers (with Aryan and Dhruv)
• Build awesome datasets for video generation (with hlky)
• Fast LoRA inference for Flux with Diffusers and PEFT (with Benjamin)
• Make your ZeroGPU Spaces go brrr with ahead-of-time compilation (with Charles, Linoy, and Poli)

Apart from these, I contribute to numerous documentation-related efforts, and they can be found by visiting these links: link 1, link 2.

NoteBlog posts outside of the Hugging Face blogging platform:

• Controlling Stable Diffusion with JAX, diffusers, and Cloud TPUs (with Merve Noyan)
• Accelerating Generative AI Part III: Diffusion, Fast (with Patrick von Platen). Tweet from PyTorch.
• Presenting Flux Fast: Making Flux go brrr on H100s. Tweet from PyTorch.
• torch.compile and Diffusers: A Hands-On Guide to Peak Performance. Tweet from PyTorch.
• Reducing AoT cold start compilation time with regional compilation (with Charles and Angela)

Carted

The first four posts were co-authored with Nilabhra Roy Chowdhury.

• Improving Dataflow Pipelines for Text Data Processing
• Variable-Length Sequences in TensorFlow
  • Part I
  • Part II
  • Part III
• Building an Efficient Machine Learning API
• Better Hardware Provisioning for ML Experiments

NoteMisc

• Lessons learned from a Deep Learning Hackathon
• “Reparameterization” trick in Variational Autoencoders
• Your First Machine Learning Project: Q and A with Sayak Paul, Google Developer Expert (GDE) in Machine Learning (Ep. 4)
• AMA with Sayak Paul - Hacktoberfest’19
• Predicting the publisher’s name from an article: A case study ¹⁵
• GDE Journey — Sayak Paul
• Multi-part tutorial series on Selfie2Anime with TFLite (joint work with Margaret Maynard-Reid) - Part I, Part II, Part III
• Multi-part tutorial series on Create Artistic Effect by Stylizing Image Background (joint work with Margaret Maynard-Reid and George Soloupis) - Part I, Part II, Part III ¹⁶
• Load-testing TensorFlow Serving and FastAPI on GKE (with Chansung Park) ¹⁷
• [ML Story] DreamBoothing Your Way into Greatness

Footnotes

1. #TFCommunitySpotlight awarded.

2. #TFCommunitySpotlight awarded.

3. #TFCommunitySpotlight awarded.

4. #TFCommunitySpotlight awarded.

5. This article got featured in “Python Top 10 Articles for the Past Month (v.Oct 2018)” and secured a rank of 4.

6. This article got featured in “Machine Learning Top 10 Articles for the Past Month (v.Nov 2018)” and secured a rank of 9.

7. This article got featured in “Python Top 10 Articles for the Past Month (v.Dec 2018)” and secured a rank of 10.

8. Featured in Sebastian Ruder’s monthly newsletter.

9. This one ranked eighth at a blogging competition.

10. This one ranked eighth at a blogging competition.

11. #TFCommunitySpotlight awarded.

12. In collaboration with the Neural Structured Learning team at Google.

13. #TFCommunitySpotlight awarded.

14. #TFCommunitySpotlight awarded.

15. This one won the ML GDE Dev Challenge.

16. #TFCommunitySpotlight awarded.

17. #TFCommunitySpotlight awarded.