# microsoft/Olive

Olive: Simplify ML Model Finetuning, Conversion, Quantization, and Optimization for CPUs, GPUs and NPUs.

Repository: https://github.com/microsoft/Olive
Canonical: https://ross.abutalabs.com/products/microsoft-olive
Homepage: https://microsoft.github.io/Olive/
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-26T16:12:56+00:00

## Health v2 (maintenance only)
Score: 91/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 75, longevity 100
- inputs: {"age_days": 2578, "days_push": 7, "days_rel": 85, "gap_med": 49.5, "n_releases_24m": 13}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2382, forks 312 (observed 2026-08-28T04:06:42.391204+00:00)

## What it is
Olive is Microsoft's AI model optimization toolkit for the ONNX Runtime, automating finetuning, conversion, quantization, and compression of ML models for CPUs, GPUs, and NPUs. It composes optimization techniques into workflows that produce efficient ONNX models meeting accuracy and latency constraints for cloud or edge inference.

## Use cases
- quantize an LLM for ONNX Runtime inference
- convert a Hugging Face model to optimized ONNX format
- finetune a small language model then optimize it for deployment
- optimize models to run on NPU hardware
- compress a model while preserving accuracy within a latency budget
- prepare models for edge inference on Windows devices

## When to choose
- you deploy models with ONNX Runtime on CPU, GPU, or NPU
- you need automated quantization or conversion pipelines for ML models
- you want to finetune and optimize LLMs for Windows or edge devices

## When to avoid
- you deploy with PyTorch or TensorFlow runtimes rather than ONNX
- you need a general-purpose training framework rather than an optimization toolkit
- your target hardware has no ONNX Runtime support

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-training, llm-inference, cli, gpu-computing
- domain: machine-learning, deep-learning, large-language-models, developer-tools, gpu-computing
- platform: python, cli, cross-platform, windows
- tags: onnx-runtime, model-optimization, quantization, model-conversion, npu, finetuning, microsoft, linux, gpu

## Member repositories
- microsoft/Olive (main) score 91

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:42.391204+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 2026-08-30T02:34:43.274090+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/Olive (fetched 2026-08-28T04:06:42.391204+00:00, sha 6ed4740931f1)
  - homepage: https://microsoft.github.io/Olive/ (fetched 2026-08-29T10:15:36.946097+00:00, sha c389f5c6fa35)
- Data as of 2026-08-30T08:39:29.467469+00:00.