RuntimeError: The size of tensor a (645) must match the size of tensor b (646) at non-singleton dimension 2

stefanopini / simple-HRNet

Multi-person Human Pose Estimation with HRNet in Pytorch

GNU General Public License v3.0

573 stars 106 forks source link

RuntimeError: The size of tensor a (645) must match the size of tensor b (646) at non-singleton dimension 2 #89

Closed luoolu closed 3 years ago

luoolu commented 3 years ago

File "/home/xkjs/NDisk/PycharmProjects/simple-HRNet/models/hrnet.py", line 66, in forward x_fused[i] = x_fused[i] + self.fuse_layers[i]j RuntimeError: The size of tensor a (645) must match the size of tensor b (646) at non-singleton dimension 2 how to resolve?

stefanopini commented 3 years ago

Hi @luolugithub , can you share the configuration of your run so that I can investigate the issue? For instance, did you run the live demo script or use the class in your code? Did you load any pre-trained weights? Are you using the multiple people option with yolov3?

luoolu commented 3 years ago

just git repo and install requirements, use coco dataset, did not use pretrained

stefanopini commented 3 years ago

Sorry for the late reply. If you still haven't solved the issue, can you answer to the following questions? I'm not able to reproduce your issue at the moment. How did you use the coco dataset? Did you run python scripts/train_coco.py? Did you install YOLOv3 using the command git submodule update --init --recursive or downloading it from github?