跳转到主要内容
此内容尚未提供您的语言版本,正在以英文显示。

VLLM High Performance LLM Serving

技能 已验证 活跃

Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.

目的

To enable users to deploy LLM APIs with high throughput and low latency using vLLM's advanced features for production environments.

功能

  • High-throughput LLM serving
  • Optimized inference latency
  • Efficient memory usage with PagedAttention
  • OpenAI-compatible API endpoint
  • Support for quantization (AWQ, GPTQ, FP8)
  • Tensor parallelism for distributed serving

使用场景

  • Deploying production LLM APIs
  • Optimizing inference latency and throughput
  • Serving large models with limited GPU memory
  • Building multi-user applications like chatbots

非目标

  • CPU-based inference
  • Research or prototyping with basic transformer implementations
  • NVIDIA-only, maximum-performance inference (TensorRT-LLM is an alternative)
  • Fine-tuning or training models

实践

  • Production deployment
  • Performance optimization
  • Quantization
  • Distributed serving

先决条件

  • NVIDIA GPU with CUDA installed
  • Python environment
  • vLLM library installed

Execution

  • info:Pinned dependenciesThe SKILL.md lists `dependencies: [vllm, torch, transformers]` but does not explicitly declare pinned interpreter versions or side-effect headers for any bundled scripts, although installation instructions point to `pip install vllm`.

安装

请先添加 Marketplace

/plugin marketplace add Orchestra-Research/AI-Research-SKILLs
/plugin install AI-Research-SKILLs@ai-research-skills

质量评分

已验证
97 /100
1 day ago 分析

信任信号

最近提交17 days ago
星标8.3k
许可证MIT
状态
查看源代码

类似扩展

Tensorrt Llm

98

Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.

技能
Orchestra-Research

Hqq Quantization

98

Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.

技能
Orchestra-Research

Hqq Quantization

96

Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.

技能
davila7

Llama Cpp

95

Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.

技能
Orchestra-Research

VLLM Inference Serving

93

Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.

技能
davila7

Llama Cpp

85

Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.

技能
davila7