Results for “gpu”

Comprehensive guide for developing WebGPU-enabled Three.js applications using TSL (Three.js Shading Language). Covers WebGPU renderer setup, TSL syntax and node materials, compute shaders, post-processing effects, and WGSL integration. Use this skill when working with Three.js WebGPU, TSL shaders, node materials, or GPU compute in Three.js.

The English summary is being prepared. Showing the original description.

Web 线上展览开发 (虚拟展厅) (Web 线上展览 / 虚拟展厅开发 (从业者/工程视角) — 用 web 技术开发线上展览、虚拟展厅、云展会、数字展馆，覆盖 3D 空间、全景漫游、WebXR、文博数字化、营销云展厅、3D 商品展示等形态。覆盖: (a) 技术栈 — 3D web(Three.js / React Three Fiber / Babylon.js / Google model-viewer / WebGL / WebGPU)、全景(Pannellum / Marzipano / Photo Sphere Viewer / krpano / 720云)、WebXR(WebXR Device API / A-Frame)、资产管线(glTF·GLB / Draco / Meshopt / KTX2 纹理压缩 / gltf-transform / LOD); (b) 自研 vs SaaS 平台选型(国内: 众趣 / 酷雷曼 / 比目鱼 / 会鸽 / 3DVista; 海外: Matterport / Kuula / Artsteps / Spatial); (c) 性能优化(大场景加载 / 模型轻量化 / draw call / 移动端 / 首屏 / 显存)——web 3D 第一性难题; (d) 策展与导览(动线设计 / 热点 hotspot / 语音讲解 / 交互); (e) 文博数字化(IIIF / 数字孪生 / 高清文物) 与 会展·电商商业化(营销云展厅 / 数据埋点 / 线索转化); (f) 部署(CDN / 流式加载 / 跨端兼容 / WebGL 降级)。学派分歧: 自研 Three.js vs SaaS 平台、真 3D 重场景 vs 720 全景轻量、WebGL vs WebGPU、文博考据派 vs 营销转化派、沉浸优先 vs 加载优先。不含: 线下展陈 / 展台搭建、原生 VR App(非 web)、通用 3D 游戏开发、纯 3D 建模教程。) Master OS — automated mastery of Web 线上展览 / 虚拟展厅开发 (从业者/工程视角) — 用 web 技术开发线上展览、虚拟展厅、云展会、数字展馆，覆盖 3D 空间、全景漫游、WebXR、文博数字化、营销云展厅、3D 商品展示等形态。覆盖: (a) 技术栈 — 3D web(Three.js / React Three Fiber / Babylon.js / Google model-viewer / WebGL / WebGPU)、全景(Pannellum / Marzipano / Photo Sphere Viewer / krpano / 720云)、WebXR(WebXR Device API / A-Frame)、资产管线(glTF·GLB / Draco / Meshopt / KTX2 纹理压缩 / gltf-transform / LOD); (b) 自研 vs SaaS 平台选型(国内: 众趣 / 酷雷曼 / 比目鱼 / 会鸽 / 3DVista; 海外: Matterport / Kuula / Artsteps / Spatial); (c) 性能优化(大场景加载 / 模型轻量化 / draw call / 移动端 / 首屏 / 显存)——web 3D 第一性难题; (d) 策展与导览(动线设计 / 热点 hotspot / 语音讲解 / 交互); (e) 文博数字化(IIIF / 数字孪生 / 高清文物) 与 会展·电商商业化(营销云展厅 / 数据埋点 / 线索转化); (f) 部署(CDN / 流式加载 / 跨端兼容 / WebGL 降级)。学派分歧: 自研 Three.js vs SaaS 平台、真 3D 重场景 vs 720 全景轻量、WebGL vs WebGPU、文博考据派 vs 营销转化派、沉浸优先 vs 加载优先。不含: 线下展陈 / 展台搭建、原生 VR App(非 web)、通用 3D 游戏开发、纯 3D 建模教程。: top builders' mental models, tool stack, current workflows, jargon, and where to keep up. Trigger this skill when the user works on Web 线上展览 / 虚拟展厅开发 (从业者/工程视角) — 用 web 技术开发线上展览、虚拟展厅、云展会、数字展馆，覆盖 3D 空间、全景漫游、WebXR、文博数字化、营销云展厅、3D 商品展示等形态。覆盖: (a) 技术栈 — 3D web(Three.js / React Three Fiber / Babylon.js / Google model-viewer / WebGL / WebGPU)、全景(Pannellum / Marzipano / Photo Sphere Viewer / krpano / 720云)、WebXR(WebXR Device API / A-Frame)、资产管线(glTF·GLB / Draco / Meshopt / KTX2 纹理压缩 / gltf-transform / LOD); (b) 自研 vs SaaS 平台选型(国内: 众趣 / 酷雷曼 / 比目鱼 / 会鸽 / 3DVista; 海外: Matterport / Kuula / Artsteps / Spatial); (c) 性能优化(大场景加载 / 模型轻量化 / draw call / 移动端 / 首屏 / 显存)——web 3D 第一性难题; (d) 策展与导览(动线设计 / 热点 hotspot / 语音讲解 / 交互); (e) 文博数字化(IIIF / 数字孪生 / 高清文物) 与 会展·电商商业化(营销云展厅 / 数据埋点 / 线索转化); (f) 部署(CDN / 流式加载 / 跨端兼容 / WebGL 降级)。学派分歧: 自研 Three.js vs SaaS 平台、真 3D 重场景 vs 720 全景轻量、WebGL vs WebGPU、文博考据派 vs 营销转化派、沉浸优先 vs 加载优先。不含: 线下展陈 / 展台搭建、原生 VR App(非 web)、通用 3D 游戏开发、纯 3D 建模教程。 problems and wants industry-grade thinking, tool selection, or workflow guidance. 触发词：「线上展览」「虚拟展厅」「web 线上展览」「云展厅」「数字展馆」

The English summary is being prepared. Showing the original description.

Enable a Skia + Dawn GPU build of Pulp (MacGpuWindowHost, Skia Graphite). Covers the prebuilt skia-builder binaries, the headers-only fresh-worktree trap, reusing another checkout's cached libs via SKIA_DIR, FindSkia layouts, verifying PULP_HAS_SKIA / MacGpuWindowHost, and the raster-fallback + GPU-wedge gotchas. Use whenever GPU rendering "doesn't work" or a build silently came up CPU-only.

The English summary is being prepared. Showing the original description.

Estimate the cost per million tokens, units of model text, on hardware you own using power consumption, electricity rates, hardware costs, and measured generation speed. Compare utilization levels to find when local operation costs less than a hosted model.

• Compare local model costs with a hosted API
• Estimate costs at different utilization levels
• Include electricity and hardware depreciation in costs

Run protein structure prediction, molecular design, and docking tools on Tamarind Bio without a local GPU. Guides tool selection, input validation, job submission, batch processing, and result retrieval through its REST API or MCP server.

• Predicting protein structures in the cloud
• Running binder or antibody design tools
• Submitting batches of sequences for analysis

Inspect CPU, memory, disk, and accelerator information alongside container and scheduler limits. Produce a JSON snapshot with identifying details removed and conservative worker and memory estimates for planning local workloads, without running stress tests.

• Choosing worker counts before a large local job
• Checking CPU and memory limits inside a container
• Comparing resource snapshots before and after changes

Export an approved fine-tuned model into a format suited to its hardware and runtime. Choose weight packaging and quantization, which reduces numerical precision, then compare outputs before and after export to check for regressions.

• Exporting a checkpoint approved for deployment
• Choosing a model format for a target GPU or laptop
• Comparing outputs before and after export

Measure denoising latency in SGLang diffusion models and identify slow operations with profiler traces. Compare performance and output quality on the same GPUs, check existing optimizations, and gather evidence before starting low-level kernel work.

• Measuring diffusion denoising latency
• Ranking slow operations from profiler traces
• Checking whether existing optimizations address a bottleneck

Guides language model training with GRPO in TRL, comparing several responses and rewarding better results. Covers dataset preparation, scoring rules for correctness and format, training configuration, monitoring, and saving models for later use.

• Training models to follow JSON or XML formats
• Designing rewards for math and coding tasks
• Adjusting training to available GPU memory

Train language models with SimPO using pairs of preferred and rejected answers. Configure preference training without a separate reference model, adapt settings for base or instruction models, and address unstable training, memory limits, or lost capabilities.

• Training a base model with preferred and rejected answers
• Further tuning an instruction model with preference data
• Adjusting training settings for math and code tasks

Diagnose unresponsive GKE GPU or TPU nodes and pods stuck terminating through read-only checks of node status, logs, serial output, and repair history. Receive documented recovery guidance to apply yourself.

• Investigate nodes stuck in NodeStatusUnknown
• Find kernel panic or hardware failure evidence
• Check node auto-repair settings and failures

Finds the base image needed to create a custom GKE node image. It translates requirements such as GKE version, operating system, processor architecture, accelerators, and sandbox settings into a matching image name.

• Finding a base image for custom GKE nodes
• Looking up images for GPU or TPU configurations
• Matching images to COS, Ubuntu, ARM64, or X86_64

Plan ablation studies, experiments that remove or replace parts of a machine learning method to test their contribution. Define reviewer questions, expected outcomes, priorities, execution order, and compute estimates, then implement experiments and record findings.

• Testing the contribution of individual method components
• Planning additional experiments before paper submission
• Prioritizing experiments within a GPU budget

Guide language model training with SimPO using pairs of preferred and rejected answers. Covers configuration, learning rate tuning, and memory troubleshooting for preference alignment, which adjusts model responses toward human preferences without a separate reference model.

• Training on preferred and rejected answer pairs
• Exploring an alternative to DPO without a reference model
• Tuning preference training for math and code tasks

Convert language models to GGUF and reduce their size through quantization, which stores model weights at lower precision. Guides cover compression choices, local inference on CPUs and Apple Silicon, and integration with tools such as Ollama.

• Running a language model on a CPU-only computer
• Choosing model compression for available memory
• Configuring Metal acceleration on Apple Silicon

Adapt large language models to specific tasks through fine-tuning, or additional training. Prepares and validates datasets, configures LoRA or QLoRA to train a small subset of parameters, and supplies training settings, evaluation scripts, and deployment optimization guidance.

• Preparing and validating JSONL training data
• Choosing LoRA or QLoRA for available GPU memory
• Configuring training and monitoring overfitting

Build and publish a browser demo for a supplied LoRA, a model adaptation for image or video tasks. Tailor the Gradio interface to its inputs and follow the model’s recommended settings on Hugging Face Spaces.

• Share a trained LoRA through a browser demo
• Build an interface tailored to image editing or video generation
• Apply recommended trigger words and generation settings

Build JavaScript or TypeScript applications that run text, image, and audio models in browsers or server runtimes. Guides compatible model selection, GPU or CPU execution, caching, and memory cleanup without requiring a Python server.

• Add text classification or translation to a browser app
• Implement image classification and object detection
• Transcribe audio using JavaScript

Set up NVIDIA infrastructure for synthetic data generation on MicroK8s or Azure AKS. Coordinate Kubernetes, inference services, and OSMO workflow execution, with prerequisite checks, deployment validation, capacity planning, and failure investigation before workloads are retried.

• Set up a synthetic data generation environment on MicroK8s
• Deploy an Azure AKS environment with OSMO and inference services
• Check GPU quotas and capacity before scaling

Create game images, transparent PNGs, textured 3D models, animated humanoids, sound effects, and voice lines using local GPU tools. Paid API options extend production to assets such as music and animated sprites.

• Creating transparent character and item images
• Building textured 3D characters and props
• Adding character movement and lip-synced speech

Build particle trails that respond to cursor or touch movement with consistent spacing at different speeds. Reuse a fixed pool of GPU-rendered particles, add keyboard-triggered bursts, and handle reduced motion, resizing, and cleanup.

• Creating particle trails for cursor and touch movement
• Making pollen or dust rise from a surface
• Adding particle bursts triggered by a keyboard-accessible button

Plan task handoffs between agents on different machines using Git to share YAML task and result files. This implementation specification describes validation, execution limits, result reporting, and GitHub Issues as an additional route for urgent requests.

• Designing GPU task handoffs from a laptop to a DevBox
• Standardizing task and result files across machines
• Designing urgent requests through labeled GitHub Issues

Diagnose failed canary ferry runs, automated checks for failures or regressions, when invoked by CI with the required context. It gathers logs, classifies the cause, and files a GitHub issue with supporting evidence.

• Investigate GPU or TPU canary failures triggered by CI
• Distinguish infrastructure failures from training or metric problems
• File a GitHub issue with logs and run context

Set up NVIDIA infrastructure for synthetic data generation on MicroK8s or Azure AKS. Coordinate Kubernetes, inference services, and OSMO workflow execution, with prerequisite checks, deployment validation, capacity planning, and failure investigation before workloads are retried.

• Set up a synthetic data generation environment on MicroK8s
• Deploy an Azure AKS environment with OSMO and inference services
• Check GPU quotas and capacity before scaling