Fix image input type and tensor shape mismatch in inference pipeline#2
Open
seongbae15 wants to merge 10 commits into
Open
Fix image input type and tensor shape mismatch in inference pipeline#2seongbae15 wants to merge 10 commits into
seongbae15 wants to merge 10 commits into
Conversation
…dimension mismatch - Fixed AttributeError caused by attempting to access '.shape' on a PIL.Image object. - Converted PIL image to tensor and added batch dimension using: 'image = transforms.ToTensor()(image).unsqueeze(0)' - Fixed RuntimeError from mismatched tensor sizes during torch.cat by ensuring input tensors have consistent dimensions.
Add time logger
Fix: AttributeError: module 'time' has no attribute 'tim'
Update from .debug to .info
Change logging.info -> print
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
First of all, thank you for your amazing research. I'm very interested in your work, and while running the inference script, I encountered a couple of errors related to the input image's data type and dimensions. I investigated the issues and made some fixes, which I’m sharing through this pull request.
This pull request fixes two runtime errors encountered during inference in the TEMU-VTOFF project.
Error 1: AttributeError — 'Image' object has no attribute 'shape'
Cause: vton_image was a PIL.Image object, which does not have a .shape attribute.
Error 2: RuntimeError — Sizes of tensors must match
Cause: Mismatch in tensor dimensions when concatenating vton_model_input, mask, and masked_vton_latents using torch.cat.
Fix
Changes Made
Added image preprocessing line to convert PIL.Image to 4D tensor
Test
Verified inference runs without errors on Colab.
Output image was generated successfully with no runtime exceptions.