AMD and PyTorch upstreamed FP8 training optimizations for AMD Instinct GPUs into TorchAO and TorchTitan, delivering a 13.4% throughput gain over BF16 on Llama3-8B dense models and recovering 89% of FP…
Meta introduced Muse Glimmer, an open-weight 30-billion-parameter model distilled from Muse Spark for on-device agentic workflows, with ExecuTorch adding end-to-end support for running it on NVIDIA GP…
PyTorch Conference North America 2026 will be held in San Jose, California, on October 20–21, 2026, with keynote speakers including Mark Collier, Executive Director of the PyTorch Foundation; Mazin Gi…
The inaugural Santa Cruz PyTorch Meetup, organized by Red Hat's Steve and UCSC OSPO's Stephanie Lieggi, drew 45 local engineers, students, and leaders for GPU/CUDA talks and lightning presentations on…
Meta's FBTriton infrastructure powers custom GPU compiler innovations like TLX and autoWS while staying synced with upstream Triton using agentic ingestion and a stratified L1/L2/L3 validation framewo…
The PyTorch Foundation announced a design contest for the 2026 flare pin for PyTorch Conference North America, with the winning entrant receiving one complimentary ticket to the conference in San Jose…
PyTorch serves as both a reference language and an implementation language for deep learning, with reference implementations in plain PyTorch used to verify the correctness of optimized production ker…
Helion, PyTorch's high-level DSL for writing performance-portable ML kernels, partnered with Google to build a TPU backend that compiles Helion kernels to Pallas, enabling PyTorch-friendly TPU kernel …
The PyTorch Foundation, which expanded into a multi-project foundation in April 2025, now hosts six projects including PyTorch, vLLM, DeepSpeed, Ray, Helion, and Safetensors. PyTorch released version …
The PyTorch Conference North America schedule is live, with the event taking place in San Jose on October 20–21, featuring sessions on training, inference, compiler innovations, responsible AI, and th…
The PyTorch-Triton 3.7 release introduces the Triton Plugin Extensions system, a framework for dynamically loading custom compiler passes, dialects, and DSL extensions into upstream Triton at runtime …
PyTorch 2.13 has been released with performance improvements including FlexAttention on Apple Silicon with up to 12x speedups, a CuTeDSL backend for Inductor, and fused nn.LinearCrossEntropyLoss reduc…
AMD has brought PyTorch Monarch to its Instinct GPUs with ROCm, enabling single-controller distributed training for large language models. The port addresses reliability challenges at scale by providi…
PyTorch's test infrastructure dynamically generates test names across devices and dtypes, causing CI failures to show names like TestLinalgCUDA.test_matmul_cuda_float32 that differ from source templat…
Over 100 participants across 20+ teams competed in the ExecuTorch Hackathon in San Francisco on June 27-28, 2026, building on-device AI applications for Snapdragon-powered Samsung Galaxy S25 Ultra dev…
Shopify has joined the PyTorch Foundation as a Platinum member, committing to invest in the open-source AI ecosystem and contribute engineering expertise to shape PyTorch for commerce and agentic AI. …
RadixArk released Miles, an open-source PyTorch-native framework for large-scale LLM reinforcement learning post-training. The framework integrates SGLang for rollout, NVIDIA Megatron-LM for training,…
PyTorch introduced the Cross-Repository CI Relay (CRCR), an automated pipeline that triggers and tracks CI in downstream repositories whenever changes are made to pytorch/pytorch. Results are displaye…
LightSeek Org open-sourced TokenSpeed-Kernel, a portable API and high-performance kernel subsystem for multi-silicon LLM inference, decoupling runtime from hardware-specific code to simplify backend c…
SGLang achieved a 5x throughput improvement for serving DeepSeek-V4 on NVIDIA GB300 hardware since its Day-0 launch, reaching ~11,200 tok/s/GPU at 50 tok/s/user through kernel and runtime optimization…