AnyLoRA - Checkpoint

3k
19k
61
Updated: May 04, 2023
clothinganimegirlsexyfemalemanga
Download (3.22 GB)
Verified:
SafeTensor
Type
Checkpoint Merge
Downloads
14,210
Uploaded
Mar 26, 2023
Base Model
SD 1.5
Hash
AutoV2
8A952CAFE9

AnyLoRA

Add a ❤️ to receive future updates.
Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee

Available on the following websites with GPU acceleration:

Remember to use the pruned version when training (less vram and no baked vae).

I made this model to ensure my future LoRA training is compatible with newer models, plus to get a model with a style neutral enough to get accurate styles with any style LoRA. Training on this model is much more effective conpared to NAI, so at the end you might want to adjust the weight or offset (I suspect that's because NAI is now much diluted in newer models). I usually find good results at 0.65 weigth that I later offset to 1.

This is good for inference (again, especially with styles) even if I made it mainly for training. It ended up being super good for generating pics and it's now my go-to anime model. It also eats very little vram.

The first version I'm uploading is a fp16-pruned with no baked vae, which is less than 2 GB, meaning you can get up to 6 epochs in the same batch on a colab.

Just make sure you use CLIP skip 2 and booru style tags when training.

Remember to use a good vae when generating, or images wil look desaturated. I suggest WD Vae or FT MSE. Or you can use the baked vae version.

Discussion

How would you do what you said here "I usually find good results at 0.65 weigth that I later offset to 1."? I can't seem to find any information about it in the discord or on the Kohya readme's.

Have tried experimenting in training an anime version of a real character?

For example if you would want to make John Wick/Keanu Reeves but in anime style - would that be possible? (meaning: would the end result resemble Keanu?)

Have you tried training with embedding?if so is the result good or is overfitting?

So would you recommend also using this model for more realistic training data sets like a real person, or is this more geared towards anime characters and such? Great model btw! :)

Can you upload at .safetensors?

There isn't any excuse anymore for .ckpt only uploads.

Anyone here uses any Colab for their work, Automatic1111 has been stopped for free users does anyone know any alternatives 😥

So, do you think training LoRas with AnyLoRa is better than using the base SD 1.5?

Whats the blend?