Fix prior preservation encoding in DreamBooth SD3(&its LoRA version) training scripts #12411
Conversation
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
…param instead of None; 2. tokenizers should be a list; 3. and max_sequence_length should be provided; 4. _encode_with_t5(...) now behave similar as train_dreambooth_lora_sd3.py does; when calling encode_prompt(...).
…instance_prompt should be passed to param prompt, when calling encode_prompt(...) here..
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
What does this PR do?
The current version of the examples folder contains a bug in the prior preservation calculation in train_dreambooth_sd3.py and train_dreambooth_lora_sd3.py.
Taking the LoRA version as an example: when using --with_prior_preservation, encode_prompt() incorrectly receives args.instance_prompt instead of prompts (i.e., instance_prompt + class_prompt). This prevents correct encoding of the class prior, which negatively affects subsequent training.
A similar issue exists in the non-LoRA version, where encode_prompt() is called incorrectly.
This PR aims to fix these bugs.
Before submitting
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
@sayakpaul