# OpenPPL/ppl.nn

A primitive library for neural network

Repository: https://github.com/OpenPPL/ppl.nn
Canonical: https://ross.abutalabs.com/products/pplnn
Language: C++
License: Apache-2.0
License Family: permissive
Topics: neural-network, deep-learning, onnx
Last push: 2024-11-24T16:09:17+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 1890, "days_push": 647, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1367, forks 219 (observed 2026-08-28T04:04:31.446538+00:00)

## What it is
PPLNN is a high-performance deep-learning inference engine written in C++ that runs ONNX models on x86 CPUs and NVIDIA GPUs, with a dedicated LLM engine supporting models like LLaMA, ChatGLM, and Qwen. It provides Python APIs and optimizations such as Flash Attention, continuous batching, and tensor parallelism.

## Use cases
- run onnx models with a fast inference engine
- serve llama or qwen on nvidia gpus
- deploy openmmlab models in production
- quantize llm kv cache to int8
- run deep learning inference on x86 cpu

## When to choose
- you need high-performance ONNX inference on x86 or CUDA
- you want to serve open-source LLMs with continuous batching and tensor parallelism
- you deploy OpenMMLab models and want optimized support

## When to avoid
- you need training or fine-tuning rather than inference
- you need broad hardware support beyond x86 and NVIDIA GPUs
- you prefer a mature ecosystem like TensorRT or ONNX Runtime with extensive community support

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, machine-learning, deep-learning, gpu-computing
- domain: deep-learning, large-language-models, machine-learning, artificial-intelligence
- platform: cpp, python, cross-platform
- tags: onnx, inference-engine, llm-serving, openmmlab, cuda, x86, tensor-parallelism, flash-attention, kv-cache-quantization, linux, gpu

## Member repositories
- OpenPPL/ppl.nn (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:31.446538+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:41:05.658070+00:00, confidence not recorded.
  - readme: https://github.com/OpenPPL/ppl.nn (fetched 2026-08-28T04:04:31.446538+00:00, sha b79aabce3a47)
- Data as of 2026-08-30T08:39:29.467469+00:00.