scalelsd 笔记 线段识别 本地部署 模型架构

ant-research/scalelsd | DeepWiki

https://arxiv.org/html/2506.09369?_immersive_translate_auto_translate=1

https://gitee.com/njsgcs/scalelsd

https://github.com/ant-research/scalelsd

https://huggingface.co/cherubicxn/scalelsd

模型链接： 

https://cdn-lfs-us-1.hf.co/repos/4b/b7/4bb773266cdc6bd3898cae785776f1dd52f835c2d4fcdf132ca4c487ecac7942/d0a7333c4de518a70af7d6ea4a283aa0ce9ffe1047525373b94070b1d1d4003f?response-content-disposition=attachment%3B+filename*%3DUTF-8%27%27scalelsd-vitbase-v2-train-sa1b.pt%3B+filename%3D%22scalelsd-vitbase-v2-train-sa1b.pt%22%3B&Expires=1752922960&Policy=eyJTdGF0ZW1lbnQiOlt7IkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTc1MjkyMjk2MH19LCJSZXNvdXJjZSI6Imh0dHBzOi8vY2RuLWxmcy11cy0xLmhmLmNvL3JlcG9zLzRiL2I3LzRiYjc3MzI2NmNkYzZiZDM4OThjYWU3ODU3NzZmMWRkNTJmODM1YzJkNGZjZGYxMzJjYTRjNDg3ZWNhYzc5NDIvZDBhNzMzM2M0ZGU1MThhNzBhZjdkNmVhNGEyODNhYTBjZTlmZmUxMDQ3NTI1MzczYjk0MDcwYjFkMWQ0MDAzZj9yZXNwb25zZS1jb250ZW50LWRpc3Bvc2l0aW9uPSoifV19&Signature=fCSnqgEbqmYV7X3VhNo-uEz-ZO6gRvRUqtS4TzrEO2iSJ6U0LU%7EbrVwWwhc69FveoaXgoq7ikmJRWv7MZxomAF5qxPxQIr3Nhor1KyDeEp267ikFN6ASBAA8IcJ67unIWzT4d3CTrbNVh8wT0pn1JsL4gGK-Xl2A0x7%7ErVb9exw-8h7AUGDejkvg5yFbzrRIw6U4QqyuwOFpicMlZXNaPBPnjfyKuQmfdz4UzbZ4C5WeKKoUAa6sxsMTu-GEF4wv%7ETBbjQS1ev5ZQ2%7EYaZxZnICvEwcwdNxSpJPiZuJrrZq95xSOcTaTFORO-1wDT7zXJY9pdFPofzIV2CBv1i6AAQ__&Key-Pair-Id=K24J24Z295AEI9

https://cdn-lfs-us-1.hf.co/repos/4b/b7/4bb773266cdc6bd3898cae785776f1dd52f835c2d4fcdf132ca4c487ecac7942/a7eaabbfc5e79f8fc407251f923f791b39c739042eb585ca52bb85abe64c4ee7?response-content-disposition=attachment%3B+filename*%3DUTF-8%27%27scalelsd-vitbase-v1-train-sa1b.pt%3B+filename%3D%22scalelsd-vitbase-v1-train-sa1b.pt%22%3B&Expires=1752923360&Policy=eyJTdGF0ZW1lbnQiOlt7IkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTc1MjkyMzM2MH19LCJSZXNvdXJjZSI6Imh0dHBzOi8vY2RuLWxmcy11cy0xLmhmLmNvL3JlcG9zLzRiL2I3LzRiYjc3MzI2NmNkYzZiZDM4OThjYWU3ODU3NzZmMWRkNTJmODM1YzJkNGZjZGYxMzJjYTRjNDg3ZWNhYzc5NDIvYTdlYWFiYmZjNWU3OWY4ZmM0MDcyNTFmOTIzZjc5MWIzOWM3MzkwNDJlYjU4NWNhNTJiYjg1YWJlNjRjNGVlNz9yZXNwb25zZS1jb250ZW50LWRpc3Bvc2l0aW9uPSoifV19&Signature=rtwQzCIgS2POke1hHELX2G7I-9MbRqzH51BWuHemwviaTAHyp1HaxOf3e%7EX1bGOld1k57ZR%7EHzeWh6D1PRLpaO-aG%7EgyhmCydrSIgkUzHd9VB2vGdGkoky8l9vP3HpUUkt7ERXCLInynopDd5191g3diZ0WyYy05d69Qel1ld-vhWlBOFrp8BzY%7EzLWnfITlCu0PGSlexlM86P9k0HaWor3vwqUYFJ0b0FVJ7DSrbRKIayP-S4M1fR4amT17tCXNDn3Jb-B2z4Eu1un-YpP67GeiItMuS0RtB4qkjHbJpVFKIQSvP0i08qb0HQuxiiN%7Ez%7EepWWWW1XxraAaWhbFcWA__&Key-Pair-Id=K24J24Z295AEI9

python predictor/predict.py -i Drawing_Annotation_Recognition8.v1i.yolov12/test/images/Snipaste_2025-07-12_19-54-54_png.rf.a562ea098219605eff9cb4ce58f09d0e.jpg

 

 

ptsd了看到这个界面就想到ai画图

 

比霍尔识别好，知道这里分两段，但是无中生有了线也缺线了，

 ​​​​​​​

 

有encoder

 

 

torchinfo看这个模型会报错好像是动态模型，模型里带if结构

 

 with open("model_structure.txt", "w") as f:f.write(str(model))
ScaleLSD((backbone): DPTFieldModel((pretrained): Module((model): VisionTransformer((patch_embed): HybridEmbed((backbone): ResNetV2((stem): Sequential((conv): StdConv2dSame(3, 64, kernel_size=(7, 7), stride=(2, 2), bias=False)(norm): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(pool): MaxPool2dSame(kernel_size=(3, 3), stride=(2, 2), padding=(0, 0), dilation=(1, 1), ceil_mode=False))(stages): Sequential((0): ResNetStage((blocks): Sequential((0): Bottleneck((downsample): DownsampleConv((conv): StdConv2dSame(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): Identity()))(conv1): StdConv2dSame(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(1): Bottleneck((conv1): StdConv2dSame(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(2): Bottleneck((conv1): StdConv2dSame(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 64, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))))(1): ResNetStage((blocks): Sequential((0): Bottleneck((downsample): DownsampleConv((conv): StdConv2dSame(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)(norm): GroupNormAct(32, 512, eps=1e-05, affine=True(drop): Identity()(act): Identity()))(conv1): StdConv2dSame(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(128, 128, kernel_size=(3, 3), stride=(2, 2), bias=False)(norm2): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 512, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(1): Bottleneck((conv1): StdConv2dSame(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 512, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(2): Bottleneck((conv1): StdConv2dSame(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 512, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(3): Bottleneck((conv1): StdConv2dSame(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 128, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 512, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))))(2): ResNetStage((blocks): Sequential((0): Bottleneck((downsample): DownsampleConv((conv): StdConv2dSame(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)(norm): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity()))(conv1): StdConv2dSame(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(2, 2), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(1): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(2): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(3): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(4): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(5): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(6): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(7): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True))(8): Bottleneck((conv1): StdConv2dSame(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm1): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv2): StdConv2dSame(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(norm2): GroupNormAct(32, 256, eps=1e-05, affine=True(drop): Identity()(act): ReLU(inplace=True))(conv3): StdConv2dSame(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)(norm3): GroupNormAct(32, 1024, eps=1e-05, affine=True(drop): Identity()(act): Identity())(drop_path): Identity()(act3): ReLU(inplace=True)))))(norm): Identity()(head): ClassifierHead((global_pool): SelectAdaptivePool2d(pool_type=, flatten=Identity())(drop): Dropout(p=0.0, inplace=False)(fc): Identity()(flatten): Identity()))(proj): Conv2d(1024, 768, kernel_size=(1, 1), stride=(1, 1)))(pos_drop): Dropout(p=0.0, inplace=False)(patch_drop): Identity()(norm_pre): Identity()(blocks): Sequential((0): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(1): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(2): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(3): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(4): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(5): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(6): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(7): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(8): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(9): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(10): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity())(11): Block((norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(attn): Attention((qkv): Linear(in_features=768, out_features=2304, bias=True)(q_norm): Identity()(k_norm): Identity()(attn_drop): Dropout(p=0.0, inplace=False)(norm): Identity()(proj): Linear(in_features=768, out_features=768, bias=True)(proj_drop): Dropout(p=0.0, inplace=False))(ls1): Identity()(drop_path1): Identity()(norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(mlp): Mlp((fc1): Linear(in_features=768, out_features=3072, bias=True)(act): GELU(approximate='none')(drop1): Dropout(p=0.0, inplace=False)(norm): Identity()(fc2): Linear(in_features=3072, out_features=768, bias=True)(drop2): Dropout(p=0.0, inplace=False))(ls2): Identity()(drop_path2): Identity()))(norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)(fc_norm): Identity()(head_drop): Dropout(p=0.0, inplace=False)(head): Linear(in_features=768, out_features=1000, bias=True))(act_postprocess1): Sequential((0): Identity()(1): Identity()(2): Identity())(act_postprocess2): Sequential((0): Identity()(1): Identity()(2): Identity())(act_postprocess3): Sequential((0): ProjectReadout((project): Sequential((0): Linear(in_features=1536, out_features=768, bias=True)(1): GELU(approximate='none')))(1): Transpose()(2): Unflatten(dim=2, unflattened_size=torch.Size([24, 24]))(3): Conv2d(768, 768, kernel_size=(1, 1), stride=(1, 1)))(act_postprocess4): Sequential((0): ProjectReadout((project): Sequential((0): Linear(in_features=1536, out_features=768, bias=True)(1): GELU(approximate='none')))(1): Transpose()(2): Unflatten(dim=2, unflattened_size=torch.Size([24, 24]))(3): Conv2d(768, 768, kernel_size=(1, 1), stride=(1, 1))(4): Conv2d(768, 768, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))))(scratch): Module((layer1_rn): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(layer2_rn): Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(layer3_rn): Conv2d(768, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(layer4_rn): Conv2d(768, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(refinenet1): FeatureFusionBlock_custom((out_conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))(resConfUnit1): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(resConfUnit2): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(skip_add): FloatFunctional((activation_post_process): Identity()))(refinenet2): FeatureFusionBlock_custom((out_conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))(resConfUnit1): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(resConfUnit2): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(skip_add): FloatFunctional((activation_post_process): Identity()))(refinenet3): FeatureFusionBlock_custom((out_conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))(resConfUnit1): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(resConfUnit2): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(skip_add): FloatFunctional((activation_post_process): Identity()))(refinenet4): FeatureFusionBlock_custom((out_conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))(resConfUnit1): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(resConfUnit2): ResidualConvUnit_custom((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(activation): ReLU()(skip_add): FloatFunctional((activation_post_process): Identity()))(skip_add): FloatFunctional((activation_post_process): Identity()))(output_conv): Sequential((0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(1): ReLU(inplace=True)(2): MultitaskHead((heads): ModuleList((0): Sequential((0): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(1): ReLU(inplace=True)(2): Conv2d(32, 3, kernel_size=(1, 1), stride=(1, 1)))(1-2): 2 x Sequential((0): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(1): ReLU(inplace=True)(2): Conv2d(32, 1, kernel_size=(1, 1), stride=(1, 1)))(3-4): 2 x Sequential((0): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(1): ReLU(inplace=True)(2): Conv2d(32, 2, kernel_size=(1, 1), stride=(1, 1))))))))(loss): CrossEntropyLoss()(bce_loss): BCEWithLogitsLoss()
)

本文来自互联网用户投稿，该文观点仅代表作者本人，不代表本站立场。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。
如若转载，请注明出处：http://www.pswp.cn/pingmian/89706.shtml
繁体地址，请注明出处：http://hk.pswp.cn/pingmian/89706.shtml

如若内容造成侵权/违法违规/事实不符，请联系多彩编程网进行投诉反馈email:809451989@qq.com，一经查实，立即删除！

相关文章

Python, C ++开发个体户/个人品牌打造APP

Python, C ++开发个体户/个人品牌打造APP

个体户/个人品牌打造APP开发方案（Python C）一、技术选型与分工1. Python- 核心场景：后端API开发、数据处理、内容管理、第三方服务集成（如社交媒体分享、支付接口）。- 优势：开发效率高，丰富的库…
阅读更多...
SQLAlchemy 常见问题笔记

SQLAlchemy 常见问题笔记

文章目录SQLAlchemy Session对象如何操作数据库SQLAlchemy非序列化对象如何返回1.问题分析2.解决方案方法1：使用 Pydantic 响应模型（推荐）方法2：手动转换为字典（简单快速）方法3：使用 SQLAlchemy…
阅读更多...
Shell脚本-uniq工具

Shell脚本-uniq工具

一、前言在 Linux/Unix 系统中，uniq 是一个非常实用的文本处理命令，用于对重复的行进行统计、去重和筛选。它通常与 sort 搭配使用，以实现高效的文本数据清洗与统计分析。无论是做日志分析、访问频率统计，还是编写自动化脚本&…
阅读更多...
氛围编码（Vice Coding）的工具选择方式

氛围编码（Vice Coding）的工具选择方式

一、前言 在写作过程中，我受益于若干优秀的博客分享，它们给予我宝贵的启发： 《5分钟选对AI编辑器，每天节省2小时开发时间让你早下班！》：https://mp.weixin.qq.com/s/f0Zm3uPTcNz30oxKwf1OQQ 二、AI编辑的…
阅读更多...
[硬件电路-57]：根据电子元器件的受控程度，可以把电子元器件分为：不受控、半受控、完全受控三种大类

[硬件电路-57]：根据电子元器件的受控程度，可以把电子元器件分为：不受控、半受控、完全受控三种大类

根据电子元器件的受控程度，可将其分为不受控、半受控、完全受控三大类。这种分类基于元器件的工作状态是否需要外部信号（如电压、电流、光、热等）的主动调控，以及调控的精确性和灵活性。以下是具体分类及实例说明：一、…
阅读更多...
基于Pytorch的人脸识别程序

基于Pytorch的人脸识别程序

人脸识别原理详解人脸识别是模式识别和计算机视觉领域的重要研究方向，其目标是从图像或视频中识别出特定个体的身份。现代人脸识别技术主要基于深度学习方法，特别是卷积神经网络 (CNN)，下面从多个维度详细解析其原理：1. 人脸识别的…
阅读更多...
ubuntu 开启ssh踩坑之旅

ubuntu 开启ssh踩坑之旅

文章目录确认当前用户为普通用户 or root命令使用ssh还是sshd服务名称的由来apt update和apt upgrade的关系apt upgrade报错：“E: 您在 /var/cache/apt/archives/ 上没有足够的可用空间”开启ssh步骤错误排查查看日志修改sshd_config文件允许防火墙通过22端口确认当…
阅读更多...
力扣：动态规划java

力扣：动态规划java

sub07 线性DP - O(1) 状态转移2_哔哩哔哩_bilibili 跳楼梯 class Solution {public int climbStairs(int n) {if (n < 1) {return 1; // 处理边界情况}int[] dp new int[n 1]; // 创建长度为n1的数组&#xff0c;比方说跳二级楼梯dp[0] 1; // 初始值设定dp[1] 1;for (…
阅读更多...
React Native打开相册选择图片或拍照 -- react-native-image-picker

React Native打开相册选择图片或拍照 -- react-native-image-picker

官方文档&#xff1a;https://www.npmjs.com/package/react-native-image-picker 场景&#xff1a;点击按钮打开相册选择图片或者点击按钮拍照 import { launchCamera, launchImageLibrary } from react-native-image-picker;// ... <TouchableOpacityactiveOpacity{0.7}o…
阅读更多...
USRP B210生成信号最大带宽测试之Frank

USRP B210生成信号最大带宽测试之Frank

书接上文&#xff1a; USRP B210生成LFM,SFM,BPSK,Frank信号的最大带宽测试&#xff08;一&#xff09; USRP B210生成信号最大带宽测试&#xff08;二&#xff09;SFM USRP B210生成信号最大带宽测试&#xff08;三&#xff09;LFM USRP B210生成信号最大带宽测试之BPSK …
阅读更多...
pages.json页面路由中，globalStyle的各个属性

pages.json页面路由中，globalStyle的各个属性

欢迎来到我的UniApp技术专栏&#xff01;&#x1f389; 在这里&#xff0c;我将与大家分享关于UniApp开发的实用技巧、最佳实践和项目经验。 专栏特色&#xff1a; &#x1f4f1; 跨平台开发一站式解决方案 &#x1f680; 从入门到精通的完整学习路径 &#x1f4a1; 实战项目经…
阅读更多...
[前端技术基础]CSS选择器冲突解决方法-由DeepSeek产生

[前端技术基础]CSS选择器冲突解决方法-由DeepSeek产生

在 CSS 中&#xff0c;当多个选择器对同一元素的相同属性&#xff08;如颜色&#xff09;定义发生冲突时&#xff0c;浏览器会通过层叠规则&#xff08;Cascading&#xff09;解决冲突。具体优先级从高到低如下&#xff1a;1. !important 规则&#xff08;最高优先级&#xff0…
阅读更多...
解决 IDEA 中 XML 文件的 “URI is not registered” 报错

解决 IDEA 中 XML 文件的 “URI is not registered” 报错

解决 IDEA 中 XML 文件的 “URI is not registered” 报错 在使用 IDEA 开发时&#xff0c;XML 文件&#xff08;尤其是带有 DTD 约束的配置文件&#xff0c;如 MyBatis、Spring 配置文件&#xff09;常出现 URI is not registered (Settings | Languages & Frameworks | S…
阅读更多...
FreeBSD Conda Python3.12下安装GPT4Free（g4f）0.5.7.3版本

FreeBSD Conda Python3.12下安装GPT4Free（g4f）0.5.7.3版本

FreeBSD下不能直接安装g4f&#xff0c;因为Curl_cffi这个库装不上。0.5.0.3这个版本不需要这个库&#xff0c;所以可以安装。 那么就没有办法安装新版本了吗&#xff1f; 有的&#xff0c;就是在linux仿真环境下。 Linux仿真环境安装g4f 最简单的方法是使用chroot进入linux仿…
阅读更多...
Node.js 中基于请求 ID 实现简单队列（即时阻止策略/排队等待策略）

Node.js 中基于请求 ID 实现简单队列（即时阻止策略/排队等待策略）

在Node.js 中基于请求 ID 实现简单队列 下面示例演示两种策略&#xff0c;以同一个请求 ID 为单位&#xff1a; 即时阻止策略&#xff1a;如果已有相同 ID 的请求在处理&#xff0c;直接报错并返回。 排队等待策略&#xff1a;后续相同 ID 的请求不报错&#xff0c;而是挂起&…
阅读更多...
详解如何解决Mysql主从复制延迟

详解如何解决Mysql主从复制延迟

解决 MySQL 主从复制延迟需要从架构设计、参数调优、硬件优化等多维度综合处理。一、根本原因分析主从延迟的本质是&#xff1a;从库的 SQL 线程重放速度 < 主库的写入速度 常见瓶颈点&#xff1a;单线程回放&#xff08;MySQL 5.6 前&#xff09;从库硬件配置低&…
阅读更多...
Spring之事务使用指南

Spring之事务使用指南

Spring之事务使用指南一、事务的基础概念1.1 什么是事务&#xff1f;1.2 事务的ACID特性1.3 Spring事务的核心优势二、Spring事务的核心配置三、事务传播行为&#xff08;Propagation&#xff09;3.1 常用传播行为详解3.1.1 REQUIRED&#xff08;默认值&#xff09;3.1.2 SUPPO…
阅读更多...
基于FPGA的多级流水线加法器verilog实现,包含testbench测试文件

基于FPGA的多级流水线加法器verilog实现,包含testbench测试文件

目录 1.课题概述 2.系统仿真结果 3.核心程序 4.系统原理简介 5.参考文献 6.完整工程文件 1.课题概述 流水线&#xff08;Pipeline&#xff09;技术源于工业生产中的装配线理念&#xff0c;在数字电路中&#xff0c;它将一个复杂运算任务分解为若干个子任务&#xff0c;每…
阅读更多...
5.1.4习题精讲

5.1.4习题精讲

一、单项选择题 01. 下列部件不属于控制器的是&#xff08; C &#xff09;。 题目原文 下列部件不属于控制器的是&#xff08; &#xff09;。 A. 指令寄存器 B. 程序计数器 C. 程序状态字寄存器 D. 时序电路 正确答案&#xff1a;C 题目解析 考点分析&#xff1a; 本题考察CP…
阅读更多...
华为云Flexus+DeepSeek征文｜低代码 × 强推理：华为云 Flexus 搭建可部署的 AI Agent 实践方案【搭建宠物养护小知识AI助手】

华为云Flexus+DeepSeek征文｜低代码 × 强推理：华为云 Flexus 搭建可部署的 AI Agent 实践方案【搭建宠物养护小知识AI助手】

文章目录华为云FlexusDeepSeek征文&#xff5c;低代码 强推理&#xff1a;华为云 Flexus 搭建可部署的 AI Agent 实践方案【搭建宠物养护小知识AI助手】&#x1f680; 引言一、核心技术概览1. 华为云 Flexus X2. DeepSeek-R1 模型3. Dify 平台二、总体架构设计三、环境准备与资…
阅读更多...
最新文章

Copyright @ 2022~2023