Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Machine learning compiler based on MLIR for Sophgo TPU.
| Date | Stars |
|---|---|
| 2026-08-04 | 958 |
| 2026-08-06 | 958 |
| 2026-08-07 | 959 |
| 2026-08-08 | 960 |
| 2026-08-10 | 961 |
| 2026-08-12 | 963 |
| 2026-08-13 | 964 |
| 2026-08-14 | 965 |
| 2026-08-18 | 965 |
| 2026-08-20 | 966 |
| 2026-08-21 | 967 |
| 2026-08-22 | 966 |
| 2026-08-23 | 968 |
| 2026-08-26 | 970 |
| 2026-08-29 | 969 |
| 2026-08-31 | 968 |
| 2026-09-01 | 971 |
| 2026-09-02 | 973 |
| 2026-09-03 | 974 |
| 2026-09-04 | 975 |
| 2026-09-05 | 975 |
| 2026-09-09 | 978 |
| 2026-09-10 | 979 |
| 2026-09-15 | 980 |
| 2026-09-18 | 980 |
| 2026-09-19 | 982 |
| 2026-09-20 | 983 |
| 2026-09-21 | 984 |
| 2026-09-22 | 988 |
| 2026-09-24 | 989 |
| 2026-09-27 | 990 |
| 2026-09-28 | 992 |
| 2026-09-29 | 993 |
| 2026-09-30 | 992 |
| 2026-10-01 | 993 |
| 2026-10-03 | 994 |
| 2026-10-05 | 995 |
| 2026-10-08 | 996 |
| 2026-10-10 | 997 |
| 2026-10-12 | 997 |
| 2026-10-13 | 998 |
| 2026-10-16 | 999 |
| 2026-10-17 | 1000 |
| 2026-10-18 | 1001 |
| 2026-10-19 | 1003 |
| 2026-10-20 | 1006 |
| 2026-10-21 | 1008 |
| 2026-10-22 | 1009 |
| 2026-10-26 | 1008 |
| 2026-10-27 | 1009 |
| 2026-10-28 | 1010 |
| 2026-10-30 | 1013 |
| 2026-11-01 | 1014 |
| 2026-11-02 | 1015 |
| 2026-11-03 | 1016 |
| 2026-11-04 | 1017 |
Today
+1 stars today
This week
+7 stars this week
This month
+22 stars this month
Momentum
26.0
growth rate 0.69%/day
<p align="center"> <img src="./docs/assets/sophgo_cover.png" alt="TPU-MLIR" /> </p> <h1 align="center">TPU-MLIR</h1> <p align="center"> <em>An open-source, MLIR-based machine-learning compiler for TPUs.</em> </p> <p align="center"> <a href="https://pypi.org/project/tpu_mlir/"><img src="https://img.shields.io/pypi/v/tpu_mlir.svg" alt="PyPI"></a> <a href="https://pypi.org/project/tpu_mlir/"><img src="https://img.shields.io/pypi/pyversions/tpu_mlir.svg" alt="Python"></a> <a href="https://github.com/sophgo/tpu-mlir/stargazers"><img src="https://img.shields.io/github/stars/sophgo/tpu-mlir?style=social" alt="Stars"></a> <a href="https://github.com/sophgo/tpu-mlir/issues"><img src="https://img.shields.io/github/issues/sophgo/tpu-mlir.svg" alt="Issues"></a> <a href="https://arxiv.org/abs/2210.15016"><img src="https://img.shields.io/badge/arXiv-2210.15016-b31b1b.svg" alt="arXiv"></a> </p> <p align="center"> <a href="./README.md">English</a> · <a href="./README_cn.md">简体中文</a> · <a href="https://tpumlir.org/quick_start_en/index.html">Quick Start</a> · <a href="https://tpumlir.org/developer_manual_en/index.html">Docs</a> · <a href="https://github.com/sophgo/tpu-mlir/issues">Issues</a> </p> --- ## ✨ Overview **TPU-MLIR** converts pre-trained neural networks from mainstream frameworks into `bmodel` files that run efficiently on TPUs. Built on top of [MLIR](https://mlir.llvm.org/), it provides a unified IR, a clean lowering pipeline, and a rich set of tools for quantization, calibration, and deployment. <p align="center"> <img src="./docs/assets/tpumlir_workflow.png" alt="TPU-MLIR compilation workflow" width="800" /> </p> ## 🚀 Highlights - **Multi-framework front-ends** — PyTorch, ONNX, TFLite, Caffe (other frameworks via ONNX). - **LLM-ready** — one-shot conversion of HuggingFace LLMs (Qwen, MiniCPM-V, …) via `llm_convert.py`. - **Full quantization toolchain** — F32 / BF16 / F16 / INT8 (symmetric & asymmetric), AWQ / GPTQ / AutoRound passthrough, calibration, QAT. - **MLIR-based pipeline** — clean dialects (Top / Tpu), pattern rewrites, layer-group memory planning. - **Production tooling** — `model_runner`, `model_tool`, accuracy validation, simulator, visualizer. - **Bilingual docs & active community** — English / 中文 manuals, papers, and video tutorials. ## 📚 Table of Contents - [Installation](#-installation) - [Quick Start (LLM — Qwen)](#-quick-start--llm--qwen) - [Quick Start (Vision — YOLOv5)](#-quick-start--vision--yolov5) - [Auxiliary Tools](#-auxiliary-tools) - [Resources](#-resources) - [Citation](#-citation) - [Contributing](#-contributing) - [License](#-license) --- ## 🔧 Installation TPU-MLIR runs inside a prebuilt Docker image. After the container is running you can either install the Python wheel or build from source. ### 1. Pull the Docker image ```shell docker pull sophgo/tpuc_dev:latest ``` If the pull fails, download the tarball and load it manually: ```shell wget https://sophon-assets.sophon.cn/sophon-prod-s3/drive/25/04/15/16/tpuc_dev_v3.4.tar.gz docker load -i tpuc_dev_v3.4.tar.gz ``` Create and enter the container: ```shell docker run --privileged --name tpu-mlir -v $PWD:/workspace -it sophgo/tpuc_dev:latest ``` ### 2a. Install the prebuilt wheel > Requires Python ≥ 3.10 on Ubuntu 22.04 (already satisfied inside the Docker image). ```shell pip install tpu_mlir ``` ### 2b. Build from source (recommended) ```shell # git clone with --depth 1 cd /workspace/tpu-mlir pip install -r requirements.txt source ./envsetup.sh ./build.sh ``` --- ## 🤖 Quick Start — LLM (Qwen) Convert and run a HuggingFace LLM (here: Qwen) on a TPU. <details> <summary><b>Click to expand the full LLM walkthrough</b></summary> ### 1. Download the model A pre-quantized AWQ / GPTQ / AutoRound build is recommended. ```shell git lfs install git clone https://huggingface.co/Intel/Qwen3.5-2B-int4-AutoRound ``` ### 2. Compile to bmodel LLM compilation falls into two scenarios, controlled by `--use_
Excerpt of 19,540 characters
Read on GitHub
HarmonyHu · Sophgo
859
429
263
237
190
147
132
120
92
84
84
82
72
67
53
48
43
41
38
35
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:0ecacc6e1403f7da, topic:llm