Burn
Burn 0.22.0: Faster Builds, Easier Extensions, and Smarter Autotuning
Faster builds, easier custom operations, adaptive memory management, and smarter autotuning…
Blog
Latest updates
Explore what's new, our releases, technical work, and practical guides from across the Tracel ecosystem.
Burn 0.22.0: Faster Builds, Easier Extensions, and Smarter Autotuning
Faster builds, easier custom operations, adaptive memory management, and smarter autotuning…
Burn ONNX 0.21.0 Release
A dedicated repository, broader model coverage, and a new ONNX compliance gate…
Burn 0.21.0 Release: Up to 8× Lower Framework Overhead, Differentiable Collectives and Improved Kernels
Distributed training, lower overhead, improved kernels, and early reinforcement-learning support…
5× Faster than Rust Standard Channel: Optimizing Burn Communication Layer
Specialized asynchronous channels and memory arenas accelerate Burn communication…
Burn 0.20.0 Release: Unifying CPU & GPU kernels with CubeCL
CubeCL unifies optimized CPU and GPU kernels, with Blackwell support…
Burn: End of the Year Review
A year of performance work and tighter integration across Burn and CubeCL…
Burn
Burn ONNX 0.21.0 Release
A dedicated repository, broader model coverage, and a new ONNX compliance gate…
Burn
Burn 0.21.0 Release: Up to 8× Lower Framework Overhead, Differentiable Collectives and Improved Kernels
Distributed training, lower overhead, improved kernels, and early reinforcement-learning support…
Burn
5× Faster than Rust Standard Channel: Optimizing Burn Communication Layer
Specialized asynchronous channels and memory arenas accelerate Burn communication…
Burn / CubeCL
Burn 0.20.0 Release: Unifying CPU & GPU kernels with CubeCL
CubeCL unifies optimized CPU and GPU kernels, with Blackwell support…
Burn / CubeCL
Burn: End of the Year Review
A year of performance work and tighter integration across Burn and CubeCL…
CubeCL
When Profilers Mislead You: Learning from a Pointless Optimization
How asynchronous GPU behavior can hide the real performance bottleneck…
Burn
Burn 0.19.0 Release: Quantization, Distributed Training, and LLVM Backend
Quantization, multi-GPU training, LLVM CPU execution, and a renewed training loop…
Tracel
Tracel AI Raised $3M in Funding
New funding supports Tracel’s mission to broaden access to AI compute…
Burn
Announcing Burn-LM (alpha): LLM Inference Engine
An alpha inference engine for LLMs, VLMs, training, and fine-tuning…
CubeCL
State-of-the-Art Multiplatform Matrix Multiplication Kernels
Portable matrix-multiplication kernels designed to rival vendor-specific libraries…
Burn
Burn 0.18.0 Release Notes
Performance, reliability, and optimization improvements across the framework…
Burn / CubeCL
Burn 0.17.0 Release Notes
A Metal backend, expanded tensor fusion, and CubeCL-powered runtimes…
Burn
Why Quantization Matters
Why memory bandwidth makes efficient model quantization essential for inference…
CubeCL
Improve Rust Compile Time by 108×
Compiler-focused techniques cut matrix-multiplication benchmark build time dramatically…
Burn / CubeCL
Burn 0.16.0 Release Notes
Faster tensor operations, experimental runtimes, and new quantization foundations…
Tracel
Going Big and Small for 2025
A roadmap for performance, portability, and flexibility across the compute stack…
Burn / CubeCL
Burn 0.15.0 Release Notes
Major tensor-operation improvements and experimental multi-backend capabilities…
Tracel
Becoming the Fastest: Introduction
A performance program for AI beyond hardware and software constraints…
Burn
Transitioning From PyTorch to Burn
Build a ResNet workflow in Burn and import ImageNet-trained weights…
Burn
Building Blocks #1: Dataset & Data Loading
An introduction to Burn datasets, batchers, and the data-loading workflow…
Burn / CubeCL
Burn 0.14.0 Release Notes
CubeCL joins Burn for cross-platform GPU programming, alongside broader improvements…
Burn
Burn 0.13.0 Release Notes
New tensor operations, autodiff advances, backend refactoring, and framework features…
Burn
Optimal Performance without Static Graphs by Fusing Tensor Operation Streams
Burn’s eager tensor streams enable fused kernels without static graphs…
Burn
Autotune for GPU Kernels: Ensuring Consistent Peak Performance
How Burn selects high-performing GPU kernels for each workload and device…
Tracel
Tracel AI: A New Player in the Field of High-Performance Computing in Deep Learning
A new company from Burn’s creator, built around high-performance and portable AI infrastructure…
Burn
Creating High Performance Asynchronous Backends With Burn-Compute
Burn-Compute brings asynchronous execution, memory management, and autotuning to backends…
Burn
Burn's New Cross-Platform GPU Backend
A Rust GPU backend built for portable deep-learning deployments…
Burn
Reduced Memory Usage: Burn's Rusty Approach to Tensor Handling
Rust ownership patterns make tensor allocation and reuse substantially more efficient…
Burn
A Case for Rust in Deep Learning
Why Rust’s safety and concurrency abstractions fit modern deep-learning systems…
No posts match these filters.