textual_inversion VS diffusers

Checking out https://github.com/rinongal/textual_inversion, which has some possibly informative examples and scripts.

I think currently you will need about 20GB VRam..., options are: 1. https://github.com/rinongal/textual_inversion - localy

Rinongal and nicolai256 versions, the latter of which is also the one explained in Nerdy Rodent's youtube video https://www.youtube.com/watch?v=WsDykBTjo20, work but they also have an issue of lacking editability in comparison to one made by huggingface's collab which is followed up in a very long issue on Rinongal's Github. You can add accumulate_grad_batches: 4 to the end of the finetune files like shown in Nerdy Rodent's video at this time stamp to try to alleviate this issue, but the quality isn't as good as one made in the online collab.

That may mostly solve that but it’s too early right now: https://github.com/rinongal/textual_inversion

For fun I tried to make an entire animated music video but it took over one week of processing and basically fell apart coherently by 30 seconds so just did one third:

https://youtu.be/f3GfUKJBUYA

The huggingface models don't work with the local stable diffusion, only the models trained locally with this repo https://github.com/rinongal/textual_inversion can be installed, at least for now.

I'm missing something about the random factor, from the sample code from https://github.com/huggingface/diffusers/blob/main/README.md

Hey people, I noticed that combining these two methods in a single forward pass increases the controllability of the generation quite a bit. I was kind of puzzled that sometimes ControlNet yielded better results than T2IAdapter for some cases, and sometimes it was the other way around, so I decided to test both at the same time, and results were quite nice. Some visuals and more motivation here: https://github.com/huggingface/diffusers/issues/5847 And it was already merged here: https://github.com/huggingface/diffusers/pull/5869

Open Parti Prompts: The better way to evaluate diffusion models (repo)

You have disabled the safety checker for by passing `safety_checker=None`. Ensure that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered results in services or applications open to the public. Both the diffusers team and Hugging Face strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling it only for use-cases that involve analyzing network behavior or auditing its results. For more information, please have a look at https://github.com/huggingface/diffusers/pull/254 .

For developers, TensorRT support also exists for the diffusers library via community pipelines. [1] It's limited, but if you're only supporting a subset of features, it can help.

In general, these insane speed boosts comes at the cost of bleeding edge features.

[1] https://github.com/huggingface/diffusers/blob/28e8d1f6ec82a6...

I was training sdxl UNET base model, with the diffusers library, which was going great until around step 210k when the weights suddenly turned back to their original values and stayed that way. I also tried with the ema version, which didn't change at all. I also looked at the tensor's weight values directly which confirmed my suspicions.

Merging LoRAs is essentially taking a weighted average of the LoRA adapter weights. It's more common in other UIs.

diffusers is working on a PR for it: https://github.com/huggingface/diffusers/pull/4473