Idea | Status | References/Papers |
---|---|---|
FIFO CogVideoX | In Progress | https://jjihwan.github.io/projects/FIFO-Diffusion |
Flux Image Inversion using RNRI | Blocked. Understanding of prior for flow matching too low | https://barakmam.github.io/rnri.github.io/ |
CogVideoX Attention Scaling | Paused. Need to recheck for higher res | https://arxiv.org/abs/2306.08645 |
RB Modulation for FLUX | Planned. SOC is straightforward. AFA needs to be seen in detail. Planning to replace CSD with other image opeerators for different manifold explorations. Soft histograms for relative color palette retention ? | https://rb-modulation.github.io/ |
CogVideoX distillation using FineVideo and PeRFlow | Planned. Needs compute grant. May be scrapped once BFLs video model is out. | https://arxiv.org/abs/2405.07510 |
Underwater Image Colorization as an Inverse Problem | Planned. Needs better underestanding of inverse problems | https://github.com/LituRout/PSLD |
Comparision of different cfg like methods | Planned. | Smoothed Energy Guidance, Guidance embedding, CFG and CFG++ |
Flux generation steering using SAE for CLIP | Planned. Need better understanding of SAEs & apply them to T5 as well | https://www.lesswrong.com/posts/Quqekpvx8BGMMcaem/interpreting-and-steering-features-in-images |
LoRA (move to MoRA ?) ControlNet layer | Planned. Compute ∆W for Flux dev & its controlnet layer. Decompose to LoRA and see decomposition error. If its low enough, LoRA should be enough | ChatGPT conversation |
MoRA finetuning Flux | I have a hypothesis: MoRA might give better samples than LoRA for Flux. I'll try it out sometime next week maybe. TLDR: 1. Full finetuning > LoRA for personalization. 2. Full finetuning > MoRA > DoRA > LoRA. 3. MoRA should converge fast like LoRA but give better quality/diversity like finetuning. There should be no free lunch though. Hmm | 1. MoRA: High-rank PEFT Approach 2. Full Finetuning of Flux 3. GitHub: MoRA |
Transformer layers as Painters for DiTs | Complete. Results published here | https://arxiv.org/abs/2407.09298 |
Pinned Loading
-
Transformer-layers-as-painters-DiT
Transformer-layers-as-painters-DiT PublicRepo for the article: Extending transformer layers as painters to DiT's
Python 1
-
Attention-Scaling-CogVideoX
Attention-Scaling-CogVideoX PublicAttention scaling for inference time entropy correction as suggested in https://arxiv.org/abs/2306.08645 and https://jfischoff.github.io/blog/motion_control_with_attention_scaling.html
-
-
Flux-Latent-Inversion
Flux-Latent-Inversion PublicFlux Image inversion tests based on naive and RNRI methods proposed here: https://barakmam.github.io/rnri.github.io/
Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.
If the problem persists, check the GitHub status page or contact support.