0


timm使用swin-transformer

1.安装

pip install timm

2.timm中有多少个预训练模型

#timm中有多少个预训练模型
model_pretrain_list = timm.list_models(pretrained=True)print(len(model_pretrain_list), model_pretrain_list[:3])

在这里插入图片描述

3加载swin模型一般准会出错

model_ft = timm.create_model('swin_base_patch4_window7_224', pretrained=True, drop_path_rate =0.2)

在这里插入图片描述
报错的内容如下

Downloading:"https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window7_224_22kto1k.pth" to /root/.cache/torch/hub/checkpoints/swin_base_patch4_window7_224_22kto1k.pth

解决办法 去swin官网下载对应的

swin_base_patch4_window7_224.pth

(所有模型我都存自己百度网盘了)文件 然后根据提示 重命名为

swin_base_patch4_window7_224_22kto1k.pth

再将该文件移动到

/root/.cache/torch/hub/checkpoints/

该目录下
这样timm就可以爽歪歪的用了

4下载预训练模型的官网

在这里插入图片描述

注意convnext

  • connext输出与分类层的输入 一定要一样

在这里插入图片描述

timm中可用的swin模型

```python
#可用的swin模型
swin_transformer =['swin_base_patch4_window7_224','swin_base_patch4_window7_224_in22k','swin_base_patch4_window12_384','swin_base_patch4_window12_384_in22k','swin_large_patch4_window7_224','swin_large_patch4_window7_224_in22k','swin_large_patch4_window12_384','swin_large_patch4_window12_384_in22k','swin_s3_base_224','swin_s3_small_224','swin_s3_tiny_224','swin_small_patch4_window7_224','swin_tiny_patch4_window7_224','swinv2_base_window8_256','swinv2_base_window12_192_22k','swinv2_base_window12to16_192to256_22kft1k','swinv2_base_window12to24_192to384_22kft1k','swinv2_base_window16_256','swinv2_cr_small_224','swinv2_cr_small_ns_224','swinv2_cr_tiny_ns_224','swinv2_large_window12_192_22k','swinv2_large_window12to16_192to256_22kft1k','swinv2_large_window12to24_192to384_22kft1k','swinv2_small_window8_256','swinv2_small_window16_256','swinv2_tiny_window8_256','swinv2_tiny_window16_256',]
#可用的VIT模型
vision_tranformer =['visformer_small','vit_base_patch8_224','vit_base_patch8_224_dino','vit_base_patch8_224_in21k','vit_base_patch16_224','vit_base_patch16_224_dino','vit_base_patch16_224_in21k','vit_base_patch16_224_miil','vit_base_patch16_224_miil_in21k','vit_base_patch16_224_sam','vit_base_patch16_384','vit_base_patch16_rpn_224','vit_base_patch32_224','vit_base_patch32_224_clip_laion2b','vit_base_patch32_224_in21k','vit_base_patch32_224_sam','vit_base_patch32_384','vit_base_r50_s16_224_in21k','vit_base_r50_s16_384','vit_giant_patch14_224_clip_laion2b','vit_huge_patch14_224_clip_laion2b','vit_huge_patch14_224_in21k','vit_large_patch14_224_clip_laion2b','vit_large_patch16_224','vit_large_patch16_224_in21k','vit_large_patch16_384','vit_large_patch32_224_in21k','vit_large_patch32_384','vit_large_r50_s32_224','vit_large_r50_s32_224_in21k','vit_large_r50_s32_384','vit_relpos_base_patch16_224','vit_relpos_base_patch16_clsgap_224','vit_relpos_base_patch32_plus_rpn_256','vit_relpos_medium_patch16_224','vit_relpos_medium_patch16_cls_224','vit_relpos_medium_patch16_rpn_224','vit_relpos_small_patch16_224','vit_small_patch8_224_dino','vit_small_patch16_224','vit_small_patch16_224_dino','vit_small_patch16_224_in21k','vit_small_patch16_384','vit_small_patch32_224','vit_small_patch32_224_in21k','vit_small_patch32_384','vit_small_r26_s32_224','vit_small_r26_s32_224_in21k','vit_small_r26_s32_384','vit_srelpos_medium_patch16_224','vit_srelpos_small_patch16_224','vit_tiny_patch16_224','vit_tiny_patch16_224_in21k','vit_tiny_patch16_384','vit_tiny_r_s16_p8_224','vit_tiny_r_s16_p8_224_in21k','vit_tiny_r_s16_p8_384',]`

参考文章
[vison transformer](https://zhuanlan.zhihu.com/p/350837279)[swin](https://zhuanlan.zhihu.com/p/485716110#:~:text=Swin%20Transformer%20%E6%98%AF%E5%9C%A8%20Vision%20Transformer%20%E7%9A%84%E5%9F%BA%E7%A1%80%E4%B8%8A%E4%BD%BF%E7%94%A8%E6%BB%91%E5%8A%A8%E7%AA%97%E5%8F%A3%EF%BC%88shifted,windows,%20SW%EF%BC%89%E8%BF%9B%E8%A1%8C%E6%94%B9%E9%80%A0%E8%80%8C%E6%9D%A5%E3%80%82%20%E5%AE%83%E5%B0%86%20Vision%20Transformer%20%E4%B8%AD%E5%9B%BA%E5%AE%9A%E5%A4%A7%E5%B0%8F%E7%9A%84%E9%87%87%E6%A0%B7%E5%BF%AB%E6%8C%89%E7%85%A7%E5%B1%82%E6%AC%A1%E5%88%86%E6%88%90%E4%B8%8D%E5%90%8C%E5%A4%A7%E5%B0%8F%E7%9A%84%E5%9D%97%EF%BC%88Windows%EF%BC%89%EF%BC%8C%E6%AF%8F%E4%B8%80%E4%B8%AA%E5%9D%97%E4%B9%8B%E9%97%B4%E7%9A%84%E4%BF%A1%E6%81%AF%E5%B9%B6%E4%B8%8D%E5%85%B1%E9%80%9A%E3%80%81%E7%8B%AC%E7%AB%8B%E8%BF%90%E7%AE%97%E4%BB%8E%E8%80%8C%E5%A4%A7%E5%A4%A7%E6%8F%90%E9%AB%98%E4%BA%86%E8%AE%A1%E7%AE%97%E6%95%88%E7%8E%87%E3%80%82)

本文转载自: https://blog.csdn.net/weixin_41504611/article/details/129172232
版权归原作者 算法黑哥 所有, 如有侵权,请联系我们删除。

“timm使用swin-transformer”的评论:

还没有评论