AArthur Kaza·Sep 18Feeding the Beast: why your TPU is bored, and how to fix itA practical guide to data pipelines and profiling for JAX on TPU. Companion notebook: feeding_the_beast_jax_tpu.ipynb (Apache 2.0).4 min read·arthurkaza.com
SSamu Physicist·Sep 16Implementing Kolmogorov-Arnold Networks as Dense GEMMs in JAX: A Neural QMC ExperimentReplacing B-splines with orthogonal polynomials and jnp.einsum in a Slater–Jastrow–backflow wavefunction for H2.10 min read·samu-physicist.hashnode.dev
SSanskriti Harmukh·Sep 16Installing JAX with ROCm Acceleration on Ubuntu 24.04Install JAX with a ROCm-accelerated Docker container or Pip on Ubuntu 24.04 to run machine learning workloads on AMD GPUs.3 min read·vultr.hashnode.dev
WWesley Kambale·Aug 7Never lose your progress: Checkpointing with OrbaxPicture this. You have spent six hours training a model. The loss curve looks beautiful, accuracy is climbing, and you are one epoch away from a result worth writing home about. Then the power goes ou16 min read·kambale.dev1
WWesley Kambale·Jul 29Feeding the Beast: Data Pipelines with GrainOver the last seven weeks we have built models, written training loops, composed optimizers, bulletproofed our code with Chex, and learned to debug inside JIT-compiled functions. Every one of those ar13 min read·kambale.dev22N
WWesley Kambale·Jul 18When JIT hides your errors: Debugging JAXLast week, Chex gave us a firewall against shape mismatches and NaN gradients. That firewall catches the bugs that announce themselves; a tensor with the wrong shape, a gradient that explodes into inf13 min read·kambale.dev2
WWesley Kambale·Jul 12Catching bugs before they catch youHere's a scenario every ML engineer eventually lives through: you kick off a training run before leaving the office. Three hours later, you check back in. The loss curve looks like a flat line at NaN.14 min read·kambale.dev
WWesley Kambale·Jul 1Optax: Optimizers You Can Compose Like LEGOIn our last article on this blog in March, we built a complete training loop. We used optax.adamw() to create an optimizer, wrapped it in nnx.Optimizer, and watched our model learn. But we barely scra11 min read·kambale.dev2
AAshitesh·Jun 11 Scaling Quantum State-Vector Simulation to 36 Qubits on Google Cloud TPU Ashitesh Singh Independent Researcher, Supported by Google TPU Research Cloud (TRC) Program GitHub: https://github.com/AshiteshSingh/Tpu-Accelerated-Quantum-JAX4 min read·ashitesh.me11C
AAshitesh·Jun 2Simulating Quantum Computers at Scale: JAX, GPUs, and Cloud TPUsNote: The detailed research paper and full experimental results for this project have already been published. This post provides a high-level summary of the core system design, key benchmarks, and eng5 min read·ashitesh.me1