Refactor: Flexible model architecture for dit models (Flux & SD3) #490
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
|
Thank you for your contribution. |
stduhpf
added a commit
to stduhpf/stable-diffusion.cpp
that referenced
this pull request
Dec 8, 2024
…#490) * Refactor: wtype per tensor * Fix default args * refactor: fix flux * Refactor photmaker v2 support * unet: refactor the refactoring * Refactor: fix controlnet and tae * refactor: upscaler * Refactor: fix runtime type override * upscaler: use fp16 again * Refactor: Flexible sd3 arch * Refactor: Flexible Flux arch * format code --------- Co-authored-by: leejet <[email protected]>
This was referenced Mar 27, 2025
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Outdated suggestions cannot be applied.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.
(Built on top of #455)
Motivations:
I stumbled upon this: https://huggingface.co/TencentARC/flux-mini, and thought it would be nice to run it on sdcpp.
The number of variants Flux and MMDiT (SD3.x) models supported is starting to get a bit overwhelming (3 each so far), and if people keep making these kinds of distillations or self merges, it would be impossible to support them all individually.
With this PR, the number of layers for each kind of block and the presence of some optional features is inferred from the tensor names in the model file when initializing the model runner, making it a lot more flexible.
New models supported with these chages (examples):
Flux Mini 3.2B (sucks at text): https://huggingface.co/TencentARC/flux-mini,
Flux Heavy 17B: https://huggingface.co/city96/Flux.1-Heavy-17B
