Youngbin Lee·2d agoUnder the Hood of JAX on TPU: From Asynchronous Dispatch to Custom Kernels with PallasHow JAX hides host latency with futures, why naive benchmarks fail, and how to explicitly program TPU VMEM and DMA using Pallas
Youngbin Lee·2d agoBenchmarking JAX on Cloud TPU v5e vs. PyTorch on NVIDIA A100: Why Peak TFLOPs/s LiesA Roofline analysis and real-world hardware experiment showing how batch size, XLA fusion, and host transfers dominate inference latency
Youngbin Lee·2d agoHow to Think About TPUs vs. GPUs: TensorCore, MXU, VMEM, and Compiler-Driven ExecutionWhy NVIDIA’s “Tensor Core” and Google’s “TensorCore” are completely different architectural units — and how VMEM scratchpads change the…
Youngbin Lee·2d agoTracing Y = ReLU(XW + b) from Python to Silicon: JAX (OpenXLA) vs. PyTorch 2How a single line of linear algebra transforms through Jaxpr, StableHLO, and XLA HLO fusions compared to TorchDynamo, AOTAutograd, and…
Ozy·Oct 28JAX Basics — Week 2: vmap, Gradient Checks and How Tracing WorksNot just using JAX transformations, but understanding what they do to your code.
Ali Nawaz·Oct 22Inside jax.jit: Tracing, Jaxpr, Compilation, and ExecutionInside jax.jit: Tracing, Jaxpr, Compilation, and Execution
Ozy·Oct 18JAX Basics Week 1: Arrays, Immutability and Your First GradientsWhy JAX, and what Week 1 covers
Luca Massaron·Oct 18From TPU Silicon to the Browser: Shipping a JAX Model to the Edge with jax2tf and TFLiteIn this tutorial, you will train on a Cloud TPU, convert the resulting model with jax2tf, quantize it with TensorFlow Lite, and run the…
InGoogle Cloud - Communitybyxbill·Oct 10Gemma 4 E2B in Pure JAX on a Colab TPU: Google's 4-Bit Export Against an Exact RepackThis article provides a step by step guide to a Colab notebook that serves Gemma 4 E2B on a single TPU v5e chip with a pure-JAX engine and…
Ruqiya Bin Safi·Oct 7Look Inside Gemma 3 on a Free TPULanguage models feel like black boxes. You send text in, text comes out, and everything in between is “the model”. But the in-between is…
