Results for “kernels”

LuisaCompute API guide covering the DSL (kernels, callables, structs, buffers, atomics, control flow, dispatch), the runtime API (Context, Device, Stream, buffers, images, ray tracing, rasterization), AST usage markers and AST-to-XIR lowering, the core library (traits, types, logging, math, fiber, allocators, STL containers), and vstd containers (HashMap, queues, pools, variant, smart pointers, utilities).

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

Manual AST construction with FunctionBuilder for kernels and callables without DSL sugar.

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

Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.

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

Expert guide for Backend.AI distributed computing platform. Automatically activates when users ask about: - Backend.AI architecture, components (Manager, Agent, Storage Proxy, Webserver, App Proxy) - Features (session scheduling, Sokovan orchestrator, multi-tenancy, resource allocation) - APIs (REST, GraphQL), authentication, RBAC authorization - Container runtime (kernels, jail sandbox, hook library, virtual folders) - Accelerator support (CUDA, ROCm, TPU, NPU, Graphcore IPU) - Client SDKs (Python, Java, JavaScript, PHP) - Setup, requirements (Python 3.13+, Docker, PostgreSQL, Redis, etcd) - How WebUI connects to/interacts with Backend.AI backend - Plugin interfaces, development setup, infrastructure Use when user mentions "Backend.AI", "backend.ai", "Sokovan", component names, or asks about the backend platform this WebUI connects to. allowed-tools: WebFetch, Read

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

Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to operators, kernels, or when user mentions enabling unsigned types, barebones unsigned types, or uint support.

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

Operate Flash Linear Attention package workflows: setup, kernels, layers/models, KDA/context parallel, and benchmarking.

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

Use for writing, reviewing, debugging, testing, or optimizing Python code that uses PyArrow arrays, schemas, Tables, RecordBatches, RecordBatchReaders, compute kernels, Dataset scans, Parquet, or Arrow IPC and C data interfaces. Trigger on schema drift, null/type semantics, chunking, batch streaming, predicate projection pushdown, conversion, and memory ownership. Do not use for pandas-only transformations, Polars-only expressions, generic SQL, Arrow Flight service design, C++/Rust Arrow, or Parquet work that does not use PyArrow.

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

Review and upgrade MetaX model support against a target vLLM revision and installed MACA components, recursively including model-dependent attention and kernels. Establish quantization and cache differences, preserve upstream structure, and distinguish shared upstream bugs from adaptation defects. Use for model support work, not standalone monkey-patch or registry audits.

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

Expert guide for modern web shaders: Three Shading Language (TSL), WebGPU WGSL, Compute Shaders, and WebGL GLSL in Three.js and Babylon.js. Covers procedural generation, compute kernels, and post-processing / Panduan ahli shader web modern: TSL, WGSL, Compute Shader, dan GLSL.

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

Use when working on imp's quantization formats, loaders, or dequant paths - GGUF Q4_0...Q8_0/Q*_K/IQ4/MXFP4, NVFP4 two-level scaling (Modelopt vs compressed-tensors layouts), FP8 E4M3, StorageTier, decode cache, KV-cache dtypes (auto/fp16/fp8/int8/int4/nvfp4/mxfp4), "which quant should I use", "which KV dtype", scale-factor layout, dequant kernel wiring, imp-quantize / AWQ, judging quant quality (PPL). Do NOT use for writing/optimizing GEMM/GEMV kernels (sm120-cuda-expert) or measuring quant perf (benchmark-cuda).

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

Use when imp outputs differ between two paths that should agree - chunk sizes, prefill vs decode, prefix-cache resend, two engines (imp vs llama.cpp vs HF), two builds or kernels - "output changes with chunk size", "first token differs", "logprobs moved", "which layer diverges", "who is right, imp or llama.cpp", "near-tie". Do NOT use for repetition loops or garbage output (check-degeneration), a new arch that loads wrong (add-model-arch), or perf (benchmark-cuda).

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

Use when asking whether something in imp is actually finished - hunting stubs, placeholders, request fields that are parsed and then ignored, code paths or kernels that never run, gate-based features that silently no-op, tests that assert nothing. Triggers on "is this implemented", "unfinished", "stub", "placeholder", "dead path", "does this flag do anything", "accepted but ignored", "kernel never launched", "test asserts nothing", "neutral A/B on a gated feature". Do NOT use for structural debt / god files / duplication (codebase-audit) or for who-calls-what (code-graph).

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

Use when benchmarking, profiling, or A/B-testing CUDA kernels or end-to-end perf in the imp inference engine on RTX 5090 (sm_120), including refreshing tests/perf_baseline.json or publishing numbers to docs/BENCHMARKS.md and the README. Triggers on "benchmark kernel", "profile cuda", "ncu", "nsys", "kernel timing", "kernel sum", "occupancy", "bandwidth bound", "compute bound", "roofline", "perf baseline", "is this regression real", "decode dropped", "aggregate throughput", "two-image A/B", "prefill kernel A/B". Do NOT use for writing/optimizing kernel code (sm120-cuda-expert) or output-quality checks (check-degeneration).

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

Use when verifying that a model in the imp inference engine produces coherent output without repetition loops, token-stuck states, or state corruption across turns or streams. Triggers on "degenerates", "check degeneration", "repetition loop", "own own own", "stuck token", "empty content", "multi-turn regression", "does it still work", "NIAH", and after enabling CUDA graphs / changing forward pass / MoE routing / KV cache or KV dtype / GDN state or scan / sparse attention / PDL / speculation (MTP, n-gram) / ragged prefill / batched-decode kernels (smallm, producer quantize) / FA2 softmax.

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

Use when writing CUDA kernels, managing the thread, block, and grid hierarchy, tiling shared memory, using streams, setting nvcc flags, or using Thrust. Not for kernel debugging: use cuda-debugging.

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

Reference for MCP Jupyter tools (kernel management, cell execution, Papermill). Use when executing notebooks, managing kernels, or running code interactively via MCP.

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

Expert guide for Backend.AI distributed computing platform. Automatically activates when users ask about: - Backend.AI architecture, components (Manager, Agent, Storage Proxy, Webserver, App Proxy) - Features (session scheduling, Sokovan orchestrator, multi-tenancy, resource allocation) - APIs (REST, GraphQL), authentication, RBAC authorization - Container runtime (kernels, jail sandbox, hook library, virtual folders) - Accelerator support (CUDA, ROCm, TPU, NPU, Graphcore IPU) - Client SDKs (Python, Java, JavaScript, PHP) - Setup, requirements (Python 3.13+, Docker, PostgreSQL, Redis, etcd) - How WebUI connects to/interacts with Backend.AI backend - Plugin interfaces, development setup, infrastructure Use when user mentions "Backend.AI", "backend.ai", "Sokovan", component names, or asks about the backend platform this WebUI connects to. allowed-tools: WebFetch, Read

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

Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.

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

Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to operators, kernels, or when user mentions enabling unsigned types, barebones unsigned types, or uint support.

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

Quantização de pesos consciente da ativação para compressão de LLM de 4-bit com aceleração 3x e perda mínima de acurácia. Use ao implantar modelos grandes (7B-70B) em memória GPU limitada, quando precisa de inferência mais rápida que GPTQ com melhor preservação de acurácia, ou para modelos com instruções ajustadas e multimodais. Vencedor do Prêmio Melhor Artigo MLSys 2024.

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

Collapse meaningful ambiguity with contrastive examples. Use when a plan/design has several plausible meanings, requirements feel vague, examples would clarify intent faster than grilling, or the user asks to disambiguate, find ambiguity, use behavioral kernels, or ask contrastive questions.

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

F16 on BitNet-embedding-270M prefill (8 threads) Supports I2_S conversion with optimized kernels on x86 CPUs Lossless inference with 2 bits per weight

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

Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.

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