From d71a8639710454d9203ab91266039efbdc240d1d Mon Sep 17 00:00:00 2001 From: Gongfan Fang Date: Sun, 1 Dec 2024 15:49:11 +0800 Subject: [PATCH 1/3] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 7f27fd4..bf796eb 100644 --- a/README.md +++ b/README.md @@ -21,7 +21,7 @@ Torch-Pruning (TP) is a framework for structural pruning with the following features: * **General-purpose Pruning Toolkit:** TP enables structural pruning for a wide range of deep neural networks. Different from [torch.nn.utils.prune](https://pytorch.org/tutorials/intermediate/pruning_tutorial.html) that zeroizes parameters via masking, Torch-Pruning deploys an algorithm called ⚡ **[DepGraph](https://openaccess.thecvf.com/content/CVPR2023/html/Fang_DepGraph_Towards_Any_Structural_Pruning_CVPR_2023_paper.html)** to group and remove coupled parameters. -* **Examples**: Pruning off-the-shelf models from Huggingface, Timm, Torchvision, including [Large Language Models (LLMs)](https://github.com/VainF/Torch-Pruning/tree/master/examples/LLMs), [Segment Anything Model (SAM)](https://github.com/czg1225/SlimSAM), [Diffusion Models](https://github.com/VainF/Diff-Pruning), [Vision Transformers](https://github.com/VainF/Isomorphic-Pruning), [ConvNext](https://github.com/VainF/Isomorphic-Pruning), [Yolov7](examples/yolov7/), [yolov8](examples/yolov8/), [Swin Transformers](examples/transformers#swin-transformers-from-hf-transformers), [BERT](examples/transformers#bert-from-hf-transformers), FasterRCNN, SSD, ResNe(X)t, DenseNet, RegNet, DeepLab, etc. As detailed list can be found in 🎨 [Examples](examples). +* **Examples**: Pruning off-the-shelf models from Huggingface, Timm, Torchvision, including [Large Language Models (LLMs)](https://github.com/VainF/Torch-Pruning/tree/master/examples/LLMs), [Segment Anything Model (SAM)](https://github.com/czg1225/SlimSAM), [Diffusion Models](https://github.com/VainF/Diff-Pruning), [Vision Transformers](https://github.com/VainF/Isomorphic-Pruning), [ConvNext](https://github.com/VainF/Isomorphic-Pruning), [Yolov7](examples/yolov7/), [yolov8](examples/yolov8/), [Swin Transformers](examples/transformers#swin-transformers-from-hf-transformers), [BERT](examples/transformers#bert-from-hf-transformers), FasterRCNN, SSD, ResNe(X)t, DenseNet, RegNet, DeepLab, etc. A detailed list can be found in 🎨 [Examples](examples). For more technical details, please refer to our CVPR'23 paper. You can also find the [code](reproduce) to reproduce paper results. From daf2353778c2739710794fda9f7920002b3c3179 Mon Sep 17 00:00:00 2001 From: Gongfan Fang Date: Sun, 1 Dec 2024 15:51:46 +0800 Subject: [PATCH 2/3] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index bf796eb..bc47678 100644 --- a/README.md +++ b/README.md @@ -82,7 +82,7 @@ Here we provide a quick start for Torch-Pruning. More explained details can be f ### Why Torch-Pruning? -In structural pruning, the removal of a single parameter may affect multiple layers. For example, pruning an output dimension of a linear layer will require the removal of the corresponding input dimension in the following linear layer as shown in (a). This dependency between layers makes it challenging to prune complicated networks manually. Torch-Pruning addresses this issue by introducing a graph-based algorithm called ``DependencyGraph`` to automatically identify dependencies and collect groups for pruning. +In structural pruning, the removal of a single parameter may affect multiple layers. For example, pruning an output dimension of a linear layer will require the removal of the corresponding input dimension in the following linear layer as shown in (a). This dependency between layers makes it challenging to prune complicated networks manually. Torch-Pruning addresses this issue by introducing a graph-based algorithm called **[DepGraph](https://openaccess.thecvf.com/content/CVPR2023/html/Fang_DepGraph_Towards_Any_Structural_Pruning_CVPR_2023_paper.html)** to automatically identify dependencies and collect groups for pruning.
From 33075cfdb738e31d926b5b5a2b9661146bdaa564 Mon Sep 17 00:00:00 2001 From: Gongfan Fang Date: Sun, 1 Dec 2024 15:52:48 +0800 Subject: [PATCH 3/3] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index bc47678..23818a1 100644 --- a/README.md +++ b/README.md @@ -24,7 +24,7 @@ Torch-Pruning (TP) is a framework for structural pruning with the following feat * **Examples**: Pruning off-the-shelf models from Huggingface, Timm, Torchvision, including [Large Language Models (LLMs)](https://github.com/VainF/Torch-Pruning/tree/master/examples/LLMs), [Segment Anything Model (SAM)](https://github.com/czg1225/SlimSAM), [Diffusion Models](https://github.com/VainF/Diff-Pruning), [Vision Transformers](https://github.com/VainF/Isomorphic-Pruning), [ConvNext](https://github.com/VainF/Isomorphic-Pruning), [Yolov7](examples/yolov7/), [yolov8](examples/yolov8/), [Swin Transformers](examples/transformers#swin-transformers-from-hf-transformers), [BERT](examples/transformers#bert-from-hf-transformers), FasterRCNN, SSD, ResNe(X)t, DenseNet, RegNet, DeepLab, etc. A detailed list can be found in 🎨 [Examples](examples). -For more technical details, please refer to our CVPR'23 paper. You can also find the [code](reproduce) to reproduce paper results. +For more technical details, please refer to our CVPR'23 paper. > [**DepGraph: Towards Any Structural Pruning**](https://openaccess.thecvf.com/content/CVPR2023/html/Fang_DepGraph_Towards_Any_Structural_Pruning_CVPR_2023_paper.html) > *[Gongfan Fang](https://fangggf.github.io/), [Xinyin Ma](https://horseee.github.io/), [Mingli Song](https://person.zju.edu.cn/en/msong), [Michael Bi Mi](https://dblp.org/pid/317/0937.html), [Xinchao Wang](https://sites.google.com/site/sitexinchaowang/)* > *[Learning and Vision Lab](http://lv-nus.org/), National University of Singapore*