search

liudakai2 / UnsupDIS-pytorch

A pytorch implementation of UnsupervisedDeepImageStitching
GNU General Public License v3.0
112 stars 20 forks source link

您好，在运行过程中出现weights和bias的形状无法相乘的错误 #31

Closed chenwinkk closed 1 year ago

chenwinkk commented 1 year ago

Traceback (most recent call last): File "D:\study\HMH\UnsupDIS-pytorch-main\UnsupDIS-pytorch-main\train.py", line 393, in train(opt) File "D:\study\HMH\UnsupDIS-pytorch-main\UnsupDIS-pytorch-main\train.py", line 254, in train pred = model(imgs) # forward File "D:\anaconda\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(*input, kwargs) File "D:\study\HMH\UnsupDIS-pytorch-main\UnsupDIS-pytorch-main\models\yolo.py", line 329, in forward return self.model[-1](feature1, feature2, x2, m2) File "D:\anaconda\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(*input, *kwargs) File "D:\study\HMH\UnsupDIS-pytorch-main\UnsupDIS-pytorch-main\models\yolo.py", line 141, in forward off = self.mi.unsqueeze(-1) # [bs, 8, 1], for matrix multiplication File "D:\anaconda\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(input, kwargs) File "D:\anaconda\lib\site-packages\torch\nn\modules\container.py", line 141, in forward input = module(input) File "D:\anaconda\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(*input, **kwargs) File "D:\anaconda\lib\site-packages\torch\nn\modules\linear.py", line 103, in forward return F.linear(input, self.weight, self.bias) RuntimeError: mat1 and mat2 shapes cannot be multiplied (2x51200 and 2048x512)

完全按照您的流程未作修改，却发现出现这个错误，weights使用的是yolov5x

liudakai2 commented 1 year ago

不太确定您是否对代码进行了更改，在models/yolo.py line 141理应是self.m[i]而非self.mi image

可以再git clone最新版的试试

chenwinkk commented 1 year ago

谢谢您！！ 还有我想请教一下就是为什么大佬您代码跑出来的psnr和ssim指标比原论文高这么多呀

liudakai2 commented 1 year ago

谢谢您！！ 还有我想请教一下就是为什么大佬您代码跑出来的psnr和ssim指标比原论文高这么多呀

可能是很多细节堆起来的吧，其实我们也没针对性做太多精度上的优化