search

VITA-Group / FasterSeg

[ICLR 2020] "FasterSeg: Searching for Faster Real-time Semantic Segmentation" by Wuyang Chen, Xinyu Gong, Xianming Liu, Qian Zhang, Yuan Li, Zhangyang Wang
MIT License
524 stars 107 forks source link

I reported an error on my own data set while I was infering #48

Closed mx1mx2 closed 3 years ago

mx1mx2 commented 3 years ago

RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 1. Got 19 and 20 in dimension 2 at /pytorch/aten/src/THC/generic/THCTensorMath.cu:71

Best regards

chenwydj commented 3 years ago

Hi @mx1mx2,

Thank you for your interest in our work!

  1. Which script was this error coming from?
  2. What is the size of your image?
  3. What is the crop size you used for training?
mx1mx2 commented 3 years ago

Thank you very much,i have solved it .

when i go to the second workflow of search the architecture, the following error occur. [TensorRT] ERROR: Network must have at least one output

Best regards

chenwydj commented 3 years ago

Hi @mx1mx2,

Please take a look at this section.

Basically, you need to update the H and W here.

For the TensorRT error, I never met this before. Could you provide more error messages?