# PyTorch PyTorch is an optimized tensor library for deep learning using GPUs and CPUs. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Fault tolerant distributed training on Amazon EKS using NVRx DevFeed: [Fault tolerant distributed training on Amazon EKS using NVRx]() Original publisher: [Read original article]() Author: Aravind Neelakantan Published: 2026-09-16T18:59:25Z Content type: tutorial Language: en Sources: [Artificial Intelligence]() Topics: [NCCL](), [Amazon Elastic Kubernetes Service](), [PyTorch](), [distributed-training](), [Resilience](), [GPU]() Tags: [advanced-300](), [amazon-eks](), [amazon-elastic-kubernetes-service](), [async](), [benchmarks](), [distributed-training](), [gpu](), [pytorch](), [resiliency](), [technical-how-to]() ### AI overview This tutorial integrates NVIDIA Resiliency Extension (NVRx) with PyTorch FSDP training on Amazon EKS. It covers asynchronous checkpointing, in-process restart, and in-job restart, and reports H100 benchmarks at 2- to 8-node scale with 99%+ training efficiency and recovery measured in seconds. ### Source excerpt Integrate NVIDIA Resiliency Extension (NVRx) into PyTorch FSDP training on Amazon EKS to overlap checkpoint I/O with training and recover from GPU faults in seconds. This post covers async checkpointing, in-process restart, and ft_launcher in-job restart, with H100 benchmarks at 2 to 8 nodes showing 99%+ training efficiency and second-scale recovery. ## HeyGen x Google Cloud: Bringing Avatar IV to TPUs DevFeed: [HeyGen x Google Cloud: Bringing Avatar IV to TPUs]() Original publisher: [Read original article]() Published: 2026-09-12T11:04:33.891311Z Content type: article Language: en Sources: [Google Developers Blog]() Topics: [Google](), [Cloud](), [Artificial Intelligence](), [Optimization](), [AI Infrastructure](), [Transformer](), [PyTorch](), [Compiler](), [Streaming](), [real-time](), [Hardware](), [API]() Tags: [ai](), [ai-infrastructure](), [api](), [cloud](), [code](), [compiler](), [engineering](), [generation](), [google](), [hardware](), [model](), [optimization](), [performance](), [pytorch](), [real-time](), [streaming](), [time](), [tpu](), [transformers](), [video]() ### AI overview HeyGen and Google Cloud describe porting the 18B+ parameter Avatar IV talking-head video generation pipeline to an eight-chip Trillium TPU host. Using torchax, JAX, XLA, FSDP sharding, Ulysses sequence parallelism, and custom Pallas kernels, the team improved performance by 1.86x for real-time chunked streaming while preserving output quality through strict quality gates. ### Source excerpt HeyGen ported their 18B+ parameter Avatar IV video generation model to Google Cloud's Trillium (v6e) TPUs via torchax and XLA, utilizing FSDP and Ulysses sequence parallelism across an eight-chip mesh. To achieve a 1.86x speedup for real-time streaming, the engineering team pipelined exposed all-to-all collectives, aligned sparse attention block sizes to eliminate mask padding, and bypassed softmax serial dependencies using a precomputed Cauchy-Schwarz upper bound. These custom Pallas kernel and compiler optimizations were deployed only after passing rigorous two-tier quality gates to guarantee byte-identical or mathematically equivalent pixel outputs. ## 9 GPU Concepts Every AI Engineer Should Know DevFeed: [9 GPU Concepts Every AI Engineer Should Know]() Original publisher: [Read original article]() Author: Dr. Ashish Bamania Published: 2026-08-31T15:19:52Z Content type: tutorial Language: en Sources: [Dr. Ashish Bamania]() Topics: [GPU](), [PyTorch](), [Large Language Model](), [Artificial Intelligence]() Tags: [ai-engineer](), [artificial-intelligence](), [coding](), [compute](), [cpu](), [data-science](), [gpu](), [guide](), [internals](), [llms](), [machine-learning](), [nvidia](), [programming](), [pytorch](), [technology]() ### AI overview A practical guide to nine GPU concepts relevant to AI engineers. It introduces GPU computation hardware, memory, program execution, and multi-GPU connections, including streaming multiprocessors in NVIDIA GPUs and their role in parallel computation. ### Source excerpt A simple and practical guide to GPU internals that AI engineers actually need. Continue reading on Level Up Coding " ## How AI Coding Agents Can Unlock Materials Simulation with NVIDIA ALCHEMI Toolkit DevFeed: [How AI Coding Agents Can Unlock Materials Simulation with NVIDIA ALCHEMI Toolkit]() Original publisher: [Read original article]() Author: Elizabeth Goodman Published: 2026-08-18T18:00:00Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [ALCHEMI](), [AI-assisted coding](), [Agent Skills](), [PyTorch](), [CUDA](), [GPU](), [Python](), [AI Development]() Tags: [agent-skills](), [agentic-ai-generative-ai](), [ai-agent](), [ai-coding](), [alchemi](), [coding](), [computational-chemistry-materials-science](), [compute](), [cuda](), [developer-tools-techniques](), [featured](), [gpu](), [python](), [pytorch](), [simulation-modeling-design]() ### AI overview This article presents an end-to-end workflow for using AI coding agents with NVIDIA ALCHEMI Toolkit to build GPU-accelerated atomistic materials simulations. It explains how ALCHEMI agent skills and reference files provide API knowledge, describes the Python, PyTorch, CUDA and NVIDIA GPU environment, and reports validation across 45 generated pipelines. ### Source excerpt Atomistic simulation requires three things: knowledge of the science, compute-efficient implementation of simulations, and accessible interfaces to the... ## Run High-Performance Core Math at Scale with NVIDIA nvmath-python DevFeed: [Run High-Performance Core Math at Scale with NVIDIA nvmath-python]() Original publisher: [Read original article]() Author: Michelle Horton Published: 2026-07-30T22:43:04Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [Python](), [CUDA](), [math](), [GPU](), [NumPy](), [cpu](), [PyTorch](), [pip]() Tags: [agentic-ai-generative-ai](), [c-plus-plus](), [ci-cd](), [cpu](), [cuda](), [cuda-x](), [data-science](), [developer-tools-techniques](), [device-apis](), [featured](), [gpu](), [multi-gpu](), [nvidia](), [performance](), [python](), [pytorch](), [scale]() ### AI overview NVIDIA nvmath-python 1.0 provides a Pythonic interface to CUDA-X and NVPL math libraries, enabling optimized numerical operations on CPUs, CUDA GPUs, and distributed multi-GPU, multi-node systems. The article covers its sparse-tensor approach, flexible installation options, and interoperability with NumPy, CuPy, and PyTorch. ### Source excerpt NVIDIA nvmath-python is a library designed to bridge the gap between the Python scientific community and NVIDIA CUDA-X math libraries. It gives Python users... ## Optimizing a Neural Reconstruction Pipeline Using NVIDIA Nsight Developer Tools DevFeed: [Optimizing a Neural Reconstruction Pipeline Using NVIDIA Nsight Developer Tools]() Original publisher: [Read original article]() Author: Tanya Lenz Published: 2026-06-30T16:00:00Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [Omniverse](), [Optimization](), [Tooling](), [GPU](), [Nvidia](), [CUDA](), [PyTorch](), [Physical AI](), [Robotics](), [Synthetic Data Generation](), [real-time](), [Reinforcement learning]() Tags: [3d](), [ai](), [autonomous](), [autonomous-vehicles](), [cuda](), [developer-tools](), [developer-tools-techniques](), [driving](), [engineering](), [featured](), [gpu](), [lidar](), [machine-learning](), [nvidia](), [omniverse](), [optimization](), [performance](), [physical-ai](), [real-time](), [reinforcement-learning](), [simulation-modeling-design](), [synthetic-data-generation]() ### AI overview This article explains how NVIDIA Nsight Developer Tools can optimize the NVIDIA Omniverse NuRec neural reconstruction pipeline. It focuses on reducing GPU-intensive reconstruction and rendering costs to improve engineering iteration and move toward real-time performance. ### Source excerpt NVIDIA Omniverse NuRec is a neural reconstruction pipeline for building high-fidelity 3D representations of real-world environments from multisensor data such... ## Accelerating Transformers Fine-Tuning with NVIDIA NeMo AutoModel DevFeed: [Accelerating Transformers Fine-Tuning with NVIDIA NeMo AutoModel]() Original publisher: [Read original article]() Author: Adil Asif; Alexandros Koumparoulis; Wenwen Gao; Sylendran Arunagiri; David Messina; Bernard Nguyen Published: 2026-06-24T16:00:13Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [NeMo](), [Fine-tuning](), [AI Models](), [Mixture of Experts (MoE)](), [distributed-training](), [Generative AI](), [Artificial Intelligence](), [GPU](), [PyTorch](), [vllm](), [sglang]() Tags: [ai](), [ai-models](), [building](), [compute](), [distributed-training](), [fine-tuning](), [framework](), [generative](), [generative-ai](), [gpu](), [infrastructure](), [kernels](), [memory](), [model](), [moe](), [nemo](), [nemotron](), [nvidia](), [open](), [transformers]() ### AI overview This article explains how NVIDIA NeMo AutoModel accelerates fine-tuning of mixture-of-experts models by extending Transformers v5 with Expert Parallelism, DeepEP fused all-to-all dispatch, and TransformerEngine kernels. It describes API compatibility, distributed execution, dynamic weight loading, and reported gains of 3.4-3.7x higher training throughput and 29-32% lower GPU memory use. ### Source excerpt NVIDIA NeMo AutoModel is an open library part of the NVIDIA NeMo framework for building custom generative AI models at scale. NeMo AutoModel builds cleanly on top of v5, adding Expert Parallelism, DeepEP fused all-to-all dispatch, and TransformerEngine kernels, and it leans on v5's dynamic weight loading to bring those optimizations to a broad and growing set of model families. ## The next chapter in flood resilience: Open sourcing Google's hydrology framework DevFeed: [The next chapter in flood resilience: Open sourcing Google's hydrology framework]() Original publisher: [Read original article]() Published: 2026-06-03T18:37:00Z Content type: article Language: en Sources: [The latest research from Google]() Topics: [Google](), [Artificial Intelligence](), [AI Models](), [Resilience](), [Open Source](), [GitHub](), [Earth AI](), [Python](), [PyTorch](), [Google AI]() Tags: [ai](), [ai-models](), [architecture](), [climate-sustainability](), [data](), [earth-ai](), [generative-ai](), [github](), [google](), [open-source](), [open-source-models-datasets](), [python](), [pytorch](), [research](), [resilience](), [sustainability](), [training]() ### AI overview Google Research is open-sourcing a Python and PyTorch hydrology framework for AI-based riverine flood forecasting, enabling meteorological and hydrological agencies to use local data, train models, and improve flood warnings. ### Source excerpt Climate & Sustainability ## Executable UDFs are now in public beta on ClickHouse Cloud DevFeed: [Executable UDFs are now in public beta on ClickHouse Cloud]() Original publisher: [Read original article]() Author: San Tran; Jia Xu; Zach Naimon; Ilya Andreev; Hanzi Jiang; Kevin Zhang Published: 2026-05-30T00:00:00Z Content type: release Language: en Sources: [ClickHouse Blog]() Topics: [clickhouse](), [Python](), [SQL](), [Cloud](), [Inference](), [real-time](), [PyTorch](), [data](), [model-deployment](), [Databases](), [Embeddings](), [Next.js]() Tags: [clickhouse](), [cloud](), [data](), [deployment](), [embeddings](), [inference](), [model](), [next-js](), [python](), [pytorch](), [real-time](), [sql]() ### AI overview ClickHouse Cloud has released executable user-defined functions in public beta. Developers can upload Python functions and invoke them from SQL, including in materialized views, while ClickHouse manages sandboxed processes near the data. The article demonstrates real-time anomaly detection on equity trade ticks with a PyTorch autoencoder and a Next.js front end. ### Source excerpt Today we're excited to announce that executable UDFs are now available in public beta on ClickHouse Cloud. You can write a function in Python, upload it as a zip to your cluster, and call it from SQL like any built-in. ClickHouse manages a pool of long-li ## Running the Latest vLLM on the NVIDIA DGX Spark DevFeed: [Running the Latest vLLM on the NVIDIA DGX Spark]() Original publisher: [Read original article]() Author: Techno Tim Published: 2026-05-21T13:00:00Z Content type: tutorial Language: en Sources: [Techno Tim]() Topics: [vllm](), [DGX Spark](), [Docker](), [Dockerfile](), [GitHub Actions](), [NCCL](), [CUDA](), [PyTorch](), [GitHub](), [NGC](), [GPU](), [Ubuntu]() Tags: [ai](), [cuda](), [dgx-spark](), [docker](), [docker-image](), [github](), [github-actions](), [gpu](), [homelab](), [nccl](), [ngc](), [nvidia-dgx](), [pytorch](), [self-hosted](), [ubuntu](), [vllm]() ### AI overview This tutorial presents a reproducible Docker image pipeline for running current vLLM releases on the NVIDIA DGX Spark's GB10 ARM64 platform. It explains why NVIDIA's NGC images can lag upstream and describes a GitHub Actions build on a self-hosted Spark runner with pinned inputs and generated lockfiles. ### Source excerpt When I built my local AI cluster on a pair of ASUS Ascent GX10s, the hard part was not serving a model. The hard part was getting a working vLLM image with current components. NVIDIA's official image was already over a month behind by the time I needed it, and waiting on their release schedule was not an option. If you saw that post, you know the GX10 is an ARM64 machine built around NVIDIA's ... ## PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend DevFeed: [PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend]() Original publisher: [Read original article]() Author: AlexZhang; Cuicheng; Jun Zhang; Manhui Lin; Yue Zhang Published: 2026-05-18T15:12:46Z Content type: release Language: en Sources: [Hugging Face - Blog]() Topics: [document ai](), [Inference](), [hugging face](), [PyTorch](), [Retrieval Augmented Generation (RAG)]() Tags: [document-ai](), [hugging-face](), [inference](), [ocr](), [python](), [pytorch](), [rag](), [rocm](), [transformers]() ### AI overview PaddleOCR 3.5 adds Transformers as a supported inference backend for OCR and document parsing models, while PaddleOCR continues to manage the underlying pipelines. The release simplifies integration with Hugging Face-centered environments and downstream document workflows such as RAG, search, analytics, and automation. ### Source excerpt PaddleOCR continues to provide OCR model series such as PP-OCRv5 and document parsing model series such as PaddleOCR-VL 1.5, while Transformers becomes one of the supported backends for running them. Try the live demo on Hugging Face Spaces: PaddleOCR 3.5 introduces a more flexible inference-engine interface. Developers can select the backend through the parameter and pass backend-specific options through . ## Building Blocks for Foundation Model Training and Inference on AWS DevFeed: [Building Blocks for Foundation Model Training and Inference on AWS]() Original publisher: [Read original article]() Author: Keita Watanabe; Pavel Belevich; Aman Shanbhag Published: 2026-05-11T23:18:26Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [foundation-models](), [Amazon Web Services](), [Training AI Models](), [Inference](), [distributed-training](), [cloud-infrastructure](), [Orchestration](), [PyTorch](), [Kubernetes](), [Grafana](), [Prometheus](), [Monitoring]() Tags: [ai](), [aws](), [building](), [distributed-training](), [frameworks](), [grafana](), [hardware](), [inference](), [infrastructure](), [kubernetes](), [latency](), [low-latency](), [model](), [model-development](), [monitoring](), [networking](), [observability](), [open-source](), [prometheus](), [training]() ### AI overview An introductory technical article about the infrastructure and open-source software building blocks required for foundation-model pre-training, post-training, and inference on AWS. It discusses accelerator compute, low-latency networking, distributed storage, orchestration, ML frameworks, and observability tools. ### Source excerpt Figure: Adapted from "AI's Three Scaling Laws, Explained" (NVIDIA Blog). Taken together, these scaling regimes push the foundation-model lifecycle--pre-training, post-training, and inference--toward convergent infrastructure requirements: tightly coupled accelerator compute, a high-bandwidth low-latency network, and a distributed storage backend. ## Kubeflow SDK User Survey 2026 - Feedback, Insights and Roadmap DevFeed: [Kubeflow SDK User Survey 2026 - Feedback, Insights and Roadmap]() Original publisher: [Read original article]() Author: Kubeflow SDK Team Published: 2026-04-28T05:00:00Z Content type: article Language: en Sources: [Kubeflow]() Topics: [SDK](), [Developer experience](), [Orchestration](), [Model Development](), [PyTorch](), [Kubernetes](), [GPU]() Tags: [2026](), [community](), [developer](), [developer-experience](), [gpu](), [insights](), [jupyter-notebook](), [kubernetes](), [orchestration](), [pytorch](), [sdk](), [survey](), [tooling]() ### AI overview The article reports findings from a Kubeflow SDK user survey involving practitioners from the Kubeflow ecosystem. It describes how Kubeflow is used in machine-learning workflows, highlights common tooling and components, and summarizes challenges involving infrastructure complexity, resource management, and debugging. ### Source excerpt To better understand the needs of our community, the Kubeflow SDK working group recently conducted a user survey focused on the SDK and developer workflows. The goal was to gather feedback from practitioners across the ecosystem about their current tooling, common challenges, and the features they would most like to see improved. ## Safetensors is Joining the PyTorch Foundation DevFeed: [Safetensors is Joining the PyTorch Foundation]() Original publisher: [Read original article]() Author: Luc Georges; Lysandre Published: 2026-04-08T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [PyTorch](), [Open Source](), [Open Source Models & Datasets](), [hugging face](), [Maintainers](), [Malware](), [JSON]() Tags: [community](), [contributors](), [data](), [hugging-face](), [json](), [maintainers](), [models](), [open-source](), [open-source-collab](), [partnerships](), [pytorch](), [safety]() ### AI overview Safetensors is joining the PyTorch Foundation, bringing its vendor-neutral governance under the Linux Foundation while preserving the existing format, APIs, Hub integration, and model compatibility. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Reverse-engineering a neural network puzzle with mechanistic interpretability DevFeed: [Reverse-engineering a neural network puzzle with mechanistic interpretability]() Original publisher: [Read original article]() Author: Ricson Cheng Published: 2026-02-24T00:00:00Z Content type: tutorial Language: en Sources: [Jane Street]() Topics: [Neural Network](), [Machine learning](), [PyTorch]() Tags: [capture](), [ml](), [models](), [neural](), [puzzle](), [pytorch](), [research]() ### AI overview This article explains a Jane Street machine-learning puzzle in which solvers receive a neural network specification, including its weights, and must determine what the network does. It describes why ordinary brute-force approaches fail and presents a solver's reverse-engineering process using mechanistic interpretability. ### Source excerpt A lot of "capture-the-flag" style ML puzzles give you a black box neural net, and your job is to figure out what it does. When we were thinking of creating our own ML puzzle early last year, we wanted to do something a little different. We thought it'd be neat to give users a complete specification of the neural net, weights and all. They would then be forced to use the tools of mechanistic interpretability to reverse engineer the network--which is a situation we sometimes find ourselves facing in our own research, when trying to interpret features of complex models. ## Custom Kernels for All from Codex and Claude DevFeed: [Custom Kernels for All from Codex and Claude]() Original publisher: [Read original article]() Author: ben burtenshaw; Sayak Paul; Aritra Roy Gosthipaty; shaun smith Published: 2026-02-13T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [CUDA](), [Agent Skills](), [AI-assisted coding](), [PyTorch](), [diffusers](), [Large Language Model](), [GPU](), [kernel-builder]() Tags: [agent-skills](), [agents](), [benchmarks](), [claude-code](), [community](), [cuda](), [cursor](), [diffusers](), [gpu](), [kernels](), [memory](), [open-source](), [pytorch](), [transformers]() ### AI overview The article describes an agent skill that teaches coding agents to write production CUDA kernels. Claude and Codex were used to create working kernels for a diffusers pipeline and a transformers model, including PyTorch bindings and benchmarks. The skill packages GPU architecture, memory optimization, project structure, and integration expertise for on-demand use. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## The maturity gap in ML pipeline infrastructure DevFeed: [The maturity gap in ML pipeline infrastructure]() Original publisher: [Read original article]() Published: 2026-01-26T00:00:00Z Content type: opinion Language: en Sources: [Chainguard: Unchained]() Topics: [Security](), [Architecture & Design](), [Tooling](), [Machine learning](), [Software Engineering](), [PyTorch]() Tags: [ai-coding](), [attacks](), [chainguard-pytorch-image](), [infrastructure](), [machine-learning](), [machine-learning-pipelines](), [ml](), [ml-ops](), [pipeline](), [python](), [pytorch](), [secure-by-default](), [security](), [security-best-practices](), [serialization](), [serialization-format](), [tooling]() ### AI overview This article argues that ML pipeline infrastructure in 2026 has a security maturity gap: common tooling does not yet provide the secure-by-default protections expected in software engineering. It examines risks including data poisoning, model laundering, and insecure model serialization, and discusses short-term mitigations and longer-term industry improvements. ### Source excerpt ML pipelines in 2026 still lack secure-by-default tooling. Learn the key security gaps in ML Ops and how teams can reduce risk today. ## This Shit is Hard: Building hardened PyTorch wheels with upstream parity DevFeed: [This Shit is Hard: Building hardened PyTorch wheels with upstream parity]() Original publisher: [Read original article]() Published: 2026-01-22T00:00:00Z Content type: article Language: en Sources: [Chainguard: Unchained]() Topics: [PyTorch](), [Dependency management](), [toolchain](), [Python](), [Open Source]() Tags: [chainguard](), [chainguard-libraries](), [chainguard-libraries-for-python](), [compatibility](), [dependency](), [python](), [pytorch](), [pytorch-libraries](), [pytorch-packages](), [toolchain]() ### AI overview Chainguard describes the engineering challenges of building hardened PyTorch wheels with upstream feature parity. The article covers dependency resolution, toolchain constraints, manylinux and binary compatibility, test infrastructure, and SLSA Level 2 provenance. ### Source excerpt Learn how Chainguard tackled dependency resolution, toolchain challenges, and manylinux compatibility to deliver hardened PyTorch wheels. ## Technical Deep Dive: How DigitalOcean and AMD Delivered a 2x Production Inference Performance Increase for Character.ai DevFeed: [Technical Deep Dive: How DigitalOcean and AMD Delivered a 2x Production Inference Performance Increase for Character.ai]() Original publisher: [Read original article]() Author: Karnik Modi Published: 2026-01-13T12:30:00Z Content type: article Language: en Sources: [DigitalOcean]() Topics: [Artificial Intelligence](), [Digital Ocean](), [GPU](), [Inference](), [Inference Performance](), [GPU optimization](), [moe](), [Kubernetes](), [model-deployment](), [PyTorch](), [qwen]() Tags: [ai](), [digitalocean](), [engineering](), [gpu](), [inference](), [inference-performance](), [infrastructure](), [kubernetes](), [mixture-of-experts](), [news](), [performance](), [platforms](), [qwen](), [technical]() ### AI overview This technical deep dive describes how Character.ai, AMD, and DigitalOcean optimized inference for the Qwen3-235B Instruct FP8 model on AMD Instinct MI300X and MI325X GPUs. The reported optimizations, including parallelization, FP8 execution paths, optimized kernels, topology-aware allocation, and Kubernetes orchestration, produced up to a 2x improvement in production request throughput under stated latency and concurrency constraints. ### Source excerpt Background: How Character.ai worked with DigitalOcean and AMD to optimize performance Character.ai, a leading AI entertainment platform with about 20 million worldwide users, wanted to optimize GPU performance and achieve lower inference costs for its application, which requires low-latency performance at large scale. They approached DigitalOcean and AMD in order to achieve this goal. Working closely together, the Character.ai, AMD, and DigitalOcean teams optimized AMD Instinct™ MI300X and MI325X GPU platforms, resulting in a 2x production inference throughput. In optimized configurations, DigitalOcean delivered high request density per node while maintaining exceptional p90 responsiveness for initial token and sustained token generation throughput, outperforming prior deployments on generic, non-optimized GPU infrastructure. These gains were achieved through platform-level optimizations, including clever parallelization strategies for large Mixture-of-Experts models, efficient FP8 execution paths, optimized kernels with AITER, topology-aware GPU allocation, and production-ready Kubernetes orchestration through DigitalOcean Kubernetes (DOKS). Together, these capabilities allowed Character.ai to scale inference predictably without increasing operational burden. In this post, we will explore the specific orchestration and tuning strategies that made these gains possible. Technical deep dive overview Character.ai leverages multiple models like Qwen, Mistral and more to power their applications. This document is focused on how we optimized the Qwen3-235B Instruct FP8 model on a cluster of DigitalOcean featuring AMD Instinct GPUs. This workload was migrated from a generic, non-optimized setup on other providers to AMD Instinct™ MI325X platform on DigitalOcean, and following the outlined optimizations we were able to achieve up to a 2x improvement in request throughput (QPS) under strict latency and concurrency constraints. The Character.ai team has a demanding workload, ## Easily Build and Share ROCm Kernels with Hugging Face DevFeed: [Easily Build and Share ROCm Kernels with Hugging Face]() Original publisher: [Read original article]() Author: Abdennacer Badaoui; Daniel Huang; colorswind; Zesen Liu Published: 2025-11-17T00:00:00Z Content type: tutorial Language: en Sources: [Hugging Face - Blog]() Topics: [kernels](), [rocm](), [hugging face](), [PyTorch](), [Deep learning](), [Open Source](), [benchmarking](), [CUDA]() Tags: [amd](), [architectures](), [best-practices](), [build](), [building](), [compiler](), [compute](), [core](), [cuda](), [deep-learning](), [deployment](), [developer](), [gpu](), [hardware](), [hugging-face](), [innovation](), [kernels](), [open-source](), [pytorch](), [rocm]() ### AI overview This tutorial explains how to build, test, package, deploy, and share ROCm-compatible GPU kernels with Hugging Face's kernels library and kernel-builder. It focuses on integrating kernels with PyTorch and optimizing them for AMD GPUs, using a high-performance FP8 GEMM kernel for the AMD Instinct MI300X as an example. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Arm at the PyTorch Conference: AI Application Development, Workshops, and Developer Sessions DevFeed: [Arm at the PyTorch Conference: AI Application Development, Workshops, and Developer Sessions]() Original publisher: [Read original article]() Author: EricSondhi Published: 2025-10-10T17:35:04Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [Arm](), [PyTorch](), [AI Development](), [Responsibility & Safety]() Tags: [ai](), [arm](), [conference](), [developers](), [pytorch](), [responsible-ai](), [workshops]() ### AI overview Arm announces its participation in the PyTorch Conference on October 22-23, featuring AI technology discussions, networking opportunities, one-on-one product usability and responsible AI workshops, and Voice of the Developer sessions. ### Source excerpt Co-Authored by Michelle Yung @ Arm Join us on site October 22-23 to see how Arm empowers developers to build and deploy AI applications with ease using PyTorch and ExecuTorch. Learn about the latest AI technologies from Arm and our ecosystem while expanding your professional network alongside like-minded AI engineers. Fuel up ahead of the conference with an evening of food, drinks, and good conversation . ## SOTA OCR with Core ML and dots.ocr DevFeed: [SOTA OCR with Core ML and dots.ocr]() Original publisher: [Read original article]() Author: Christopher Fleetwood; Pedro Cuenca Published: 2025-10-02T00:00:00Z Content type: tutorial Language: en Sources: [Hugging Face - Blog]() Topics: [MLX](), [PyTorch]() Tags: [apple](), [architecture](), [battery](), [coreml](), [cpu](), [developers](), [framework](), [gpu](), [high-performance](), [images](), [mlx](), [model](), [ocr](), [on-device](), [parameter](), [precision](), [pytorch](), [repo](), [tools](), [vision]() ### AI overview This tutorial explains how to convert dots.ocr from PyTorch to Core ML for on-device execution on Apple hardware. It discusses the roles of the Neural Engine, GPU, MLX, and Core ML, then outlines a staged conversion process beginning with GPU execution, FLOAT32 precision, and static shapes. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## exploits.club Weekly Newsletter 87 - NVIDIA Merlin Bugs, GrapheneOS's Allocator, Intel CPU Bugs, And More DevFeed: [exploits.club Weekly Newsletter 87 - NVIDIA Merlin Bugs, GrapheneOS's Allocator, Intel CPU Bugs, And More]() Original publisher: [Read original article]() Author: exploits.club Published: 2025-09-26T15:00:49Z Content type: news Language: en Sources: [exploits.club]() Topics: [Nvidia](), [GrapheneOS](), [Exploit](), [intel](), [PyTorch](), [recommendation systems](), [GPU]() Tags: [exploit](), [intel](), [nvidia](), [pytorch](), [rca](), [recommendation-systems](), [updates]() ### AI overview The 87th exploits.club weekly newsletter reviews developer and security news, including a remote code execution vulnerability in NVIDIA Merlin Transformers4Rec, GrapheneOS's hardened memory allocator, and an Intel GPU driver crash during power transitions. It also mentions Google Cloud VRP updates and other security resources. ### Source excerpt Happy Friday. Almost that time of the week again: Annnnnyway 👇 In Case You Missed It... Google Cloud VRP: Enhancing Transparency and Impact in Our Rewards Program - Some VRP updates for Google Cloud with more transparency and consistency, less ambiguity New chITchat by pamoutaf Episode - A chat with @PinkDraconian that ## \`LeRobotDataset:v3.0\`: Bringing large-scale datasets to \`lerobot\` DevFeed: [\`LeRobotDataset:v3.0\`: Bringing large-scale datasets to \`lerobot\`]() Original publisher: [Read original article]() Author: Francesco Capuano; Michel Aractingi; Quentin Lhoest; Caroline Pascal; Pepijn Kooijmans; Jade Choghari; Remi Cadene; Simon Alibert; Adil Zouitine; Martino Russi Published: 2025-09-16T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [lerobot](), [datasets](), [hugging face](), [PyTorch](), [Streaming]() Tags: [datasets](), [hugging-face](), [lerobot](), [open-source](), [pytorch](), [robotics](), [streaming]() ### AI overview LeRobotDataset v3.0 introduces a standardized, extensible format for multimodal robotics data, including sensorimotor readings, camera feeds, teleoperation status, and collection metadata. It integrates with the Hugging Face and PyTorch ecosystems and adds streaming support so large datasets can be processed without downloading them in full. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. [Next page]()