MAC-AutoML/MindPipe
A unified compression and evaluation framework for LLMs and vision-language models supporting quantization and pruning across GPU and NPU hardware.
Not currently ranked — collecting fresh signals.
star history
MindPipe provides a single CLI pipeline for compressing large language and vision-language models through quantization and pruning techniques. It implements 11 quantization methods and 7 pruning methods, with integrated evaluation suites including PPL, lm-eval-harness, and VLMEvalKit benchmarks. The framework supports both NVIDIA CUDA GPUs and Huawei Ascend NPUs through a shared device abstraction layer.
Frequently asked
- What is MAC-AutoML/MindPipe?
- A unified compression and evaluation framework for LLMs and vision-language models supporting quantization and pruning across GPU and NPU hardware.
- Is MindPipe open source?
- Yes — MAC-AutoML/MindPipe is an open-source project tracked on heatdrop.
- What language is MindPipe written in?
- MAC-AutoML/MindPipe is primarily written in Python.
- How popular is MindPipe?
- MAC-AutoML/MindPipe has 1k stars on GitHub.
- Where can I find MindPipe?
- MAC-AutoML/MindPipe is on GitHub at https://github.com/MAC-AutoML/MindPipe.