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.
A model compression and acceleration toolbox based on pytorch.
| Date | Stars |
|---|---|
| 2026-07-24 | 331 |
| 2026-07-25 | 331 |
| 2026-07-28 | 331 |
| 2026-07-30 | 331 |
| 2026-08-06 | 331 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
## [中文版](https://github.com/megvii-research/Sparsebit/blob/main/README_zh-CN.md) ## News - 2023.04.27: :fire: Pipeline parallelism is supported for [alpaca-qlora](https://github.com/megvii-research/Sparsebit/blob/main/large_language_models/alpaca-qlora) which enables fine-tuning llama-65b with 8*2080ti within 13 hours. - 2023.04.15: :fire: We release [alpaca-qlora](https://github.com/megvii-research/Sparsebit/blob/main/large_language_models/alpaca-qlora) which reduce a half model size gpu-memory than alpaca-lora. With alpaca-qlora support, you can use a single 2080ti to instruct fine-tuning llama-7b/13b. - 2023.03.20: :fire: We implemented a GPTQ cuda kernel with groupsize feature and add `--single_device_mode` to support all quant LLaMAs run in a single GPU(i.e. 2080ti). [GPTQ for LLaMA](https://github.com/megvii-research/Sparsebit/blob/main/large_language_models/llama/quantization). - 2023.03.08: Release a mix-precision quantization method based on [GPTQ for LLaMA](https://github.com/megvii-research/Sparsebit/blob/main/large_language_models/llama/quantization). - 2023.02.23: Release a PTQ example of [GPT2 on wikiText2](https://github.com/megvii-research/Sparsebit/blob/main/examples/post_training_quantization/wikitext) - 2022.11.24: Release a QAT example of [BEVDet](https://github.com/megvii-research/Sparsebit/blob/main/examples/quantization_aware_training/nuscenes/bevdet) - 2022.12.13: Release some examples of [BERT](https://github.com/megvii-research/Sparsebit/tree/main/examples/post_training_quantization/GLUE/CoLA). - 2022.12.14: Release a QAT example of [BEVDepth](https://github.com/megvii-research/Sparsebit/blob/main/examples/quantization_aware_training/nuscenes/bevdepth) - 2022.12.26: Release a QAT example of [BEVDet4D](https://github.com/megvii-research/Sparsebit/blob/main/examples/quantization_aware_training/nuscenes/bevdet4d) ## Introduction Sparsebit is a toolkit with pruning and quantization capabilities. It is designed to help researchers compress and accelerate neural network models by modifying only a few codes in existing pytorch project. ## Quantization Quantization turns full-precision params into low-bit precision params, which can compress and accelerate the model without changing its structure. This toolkit supports two common quantization paradigms, Post-Training-Quantization and Quantization-Aware-Training, with following features: - Benefiting from the support of torch.fx, Sparsebit operates on a QuantModel, and each operation becomes a QuantModule. - Sparsebit can easily be extended by users to accommodate their own researches. Users can register to extend important objects such as QuantModule, Quantizer and Observer by themselves. - Exporting [QDQ-ONNX](https://onnxruntime.ai/docs/tutorials/mobile/helpers/#qdq-format-model-helpers) is supported, which can be loaded and deployed by backends such as TensorRT and OnnxRuntime. ### Results - PTQ results on ImageNet-1k: [link](https://github.com/megvii-research/Sparsebit/blob/main/examples/post_training_quantization/imagenet1k/basecase/README.md) - PTQ results of Vision Transformer on ImageNet-1k: [link](https://github.com/megvii-research/Sparsebit/blob/main/examples/post_training_quantization/imagenet1k/deit/README.md) - PTQ results of YOLO related works on COCO: [link](https://github.com/megvii-research/Sparsebit/blob/main/examples/post_training_quantization/coco2017/yolo_series/README.md) - QAT results on ImageNet-1k: [link](https://github.com/megvii-research/Sparsebit/blob/main/examples/quantization_aware_training/imagenet1k/README.md) ## Sparse Sparse is often used in deep learning to refer to operations such as reducing network parameters or network computation. At present, Sparse supported by the toolbox has the following characteristics: - Supports two types of pruning: structured/unstructured; - Supports a variety of operation objects including: weights, activations, model-blocks, model-layers, etc.; - Supports multiple pruning algori
Excerpt of 5,768 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:61617d10127bee95, topic:quantization, topic:pruning, readme:quantization