mindcv

mindcv copied to clipboard

## Motivation 1. add pretrain and finetune pipeline of mae 2. update vit: initialization method and pre-scale of attention ## Test Plan (How should this PR be tested? Do you...

sageyou

mindcv在下游任务作为backbone提取特征使用，需要新的支持

2

当前mindcv提供的backbone在分割、检测下游任务迁移，存在如下问题： 参考示例：https://github.com/mindspore-lab/models/blob/master/official/cv/OCRNet/README.md 1. 多尺度特征的获取： 针对没有skip-connection的模型，如vgg系列，backbone的多尺度特征无法获取，需要提供可获取模型特征的方法 2. 大模型缺乏特征unpatchfy机制。 transformer系列（vistformer、swin_transformer、vit等）其中，vistformer、swin_transformer仅能以固定的尺寸224、386作为特征提取器，并且不能进行多尺度特征提取；vit的提取器，除了固定尺寸、不能抽取多尺度特征的问题外，docoder部分不能剥离图片分类任务，直接将[B,C,H,W]的特征用于分割任务。 3. 所有的backbone特征，没有提供特征冻结接口，不能进行仅部分参数微调，需要全量微调，导致训练效率降低，所有下游模型模型参数都需要从头训练。

MiZhangWhuer

If this is your first time, please read our contributor guidelines: https://github.com/mindspore-lab/mindcv/blob/main/CONTRIBUTING.md **Describe the bug/ 问题描述 (Mandatory / 必填)** dpn92精度为0.1% - **Hardware Environment(`Ascend`/`GPU`/`CPU`) / 硬件环境**: > Please delete the backend...

787918582

Thank you for your contribution to the MindCV repo. Before submitting this PR, please make sure: - [x] You have read the [Contributing Guidelines on pull requests](https://github.com/mindspore-lab/mindcv/blob/main/CONTRIBUTING.md) - [x] Your...

Cui-yshoho

Thank you for your contribution to the MindCV repo. Before submitting this PR, please make sure: - [x] You have read the [Contributing Guidelines on pull requests](https://github.com/mindspore-lab/mindcv/blob/main/CONTRIBUTING.md) - [x] Your...

wcrzlh

The value Parameter (name=0.gamma, shape=(512,), dtype=Float32, requires_grad=True) , its name '0.gamma' already exists. Please set a unique name for the parameter.

1

If this is your first time, please read our contributor guidelines: https://github.com/mindspore-lab/mindcv/blob/main/CONTRIBUTING.md **Describe the bug/ 问题描述 (Mandatory / 必填)** A clear and concise description of what the bug is. When...

FEI120483

使用mindspore1.8.1报错

1

python train.py --model=swin_tiny --pretrained --opt=adamw --lr=0.001 --data_dir=/home/ma-user/work/vit_image_retrieval/data_set/dataset_split/ 请问mindcv需要的版本为多少

LeeLear

mindcv quick start 中报错 Could not load library libcudnn_cnn_infer.so.8. Error: /usr/local/cuda-10.1/targets/x86_64-linux/lib/libnvrtc.so: undefined symbol: nvrtcGetCUBIN

1

mindcv quick start 中报错 Could not load library libcudnn_cnn_infer.so.8. Error: /usr/local/cuda-10.1/targets/x86_64-linux/lib/libnvrtc.so: undefined symbol: nvrtcGetCUBIN

yyl-cloud

模型训练过程中训练模式突然崩溃

2

If this is your first time, please read our contributor guidelines: https://github.com/mindspore-lab/mindcv/blob/main/CONTRIBUTING.md **Describe the bug/ 问题描述 (Mandatory / 必填)** 模型在训练过程中会突然崩溃 - **Hardware Environment(`Ascend`) / 硬件环境**: - **Software Environment / 软件环境...

JingyangXiang

If this is your first time, please read our contributor guidelines: https://github.com/mindspore-lab/mindcv/blob/main/CONTRIBUTING.md **Describe the bug/ 问题描述 (Mandatory / 必填)** 精度劣化 - **Hardware Environment(`Ascend`/`GPU`/`CPU`) / 硬件环境**: > Please delete the backend...

787918582