Théo MartinBatch Normalization Showdown: Faster Convergence and Lower Losses! (Benchmark)TL;DR: Models converge faster and to a lower loss when using batch normalization.Jul 12, 20231Jul 12, 20231
Théo MartinSmart Batching (or How To Speed Up Your Training and Inferences For Free)TL;DR: Create batches with sequences of similar (tokenized) length. It gives ~2x faster inferences with t5-small on a T4 GPU.Jul 11, 2023Jul 11, 2023
Théo MartinDemystifying The Python Setup: Get Started in 3 Minutes (Including Download Time)After this 3 minutes setup, you will never worry again about virtual environments.Jul 10, 20231Jul 10, 20231
Théo MartinThe Ultimate Python Regex Cheat Sheet (With Some Real Life Examples)Regular expressions (regex) are a powerful tool for manipulating and analyzing text. In Python, we use the re module to work with regex.Jul 10, 20232Jul 10, 20232
Théo MartinA (Very Short) Visual Introduction to Learning Rate Schedulers (With Code)Learning rate is one of the most important hyperparameters in the training of neural networks, impacting the speed and effectiveness of the…Jul 9, 20231Jul 9, 20231
Théo MartinA (Very Short) Visual Introduction to Gradient Descent (With Code)Gradient descent is one of the most commonly used optimization algorithms in machine learning and deep learning. It’s a method to find the…Jul 9, 2023Jul 9, 2023