The output logit tensor is resized twice. #1529
Assignees
Labels
Comments
|
@ChenDirk Hi, it seems you're right. The exporting to ONNX follows original PyTorch pipeline. @MeowZheng Could you explain why there are two resizes here? Whether they can be merged into one? |
3 tasks
aravind-h-v
pushed a commit
to aravind-h-v/mmsegmentation
that referenced
this issue
Mar 27, 2023
The mask and instance image were being cropped in different ways without --center_crop, causing the model to learn to ignore the mask in some cases. This PR fixes that and generate more consistent results.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hello,
I find a small bug when I export onnx file. The output logit tensor is resized twice in the onnx file, before softmax opration. The problem is in mmsegmentation/mmseg/models/segmentors/encoder_decoder.py file, in the 'encode_decode' function, the output is resized at the first time(line 75), and the output is resized at the second time in the 'whole_inference' function(line 210). In my mind, they are repetitive and unnecessary.
The text was updated successfully, but these errors were encountered: