LECO icon indicating copy to clipboard operation
LECO copied to clipboard

Low-rank adaptation for Erasing COncepts from diffusion models.

Results 22 LECO issues
Sort by recently updated
recently updated
newest added

为什么要先根据随机初始化的噪音latens,用随机去噪步数对它进行基于lora的去噪操作得到denoised_latents呢?后续的positive_latents, neutral_latents, unconditional_latents都是基于denoised_latens在原始sd上继续去噪的

Same error for "euler_a" and "lms" noise schedulers. ```python Traceback (most recent call last): File "/content/LECO/./train_lora.py", line 334, in main(args) File "/content/LECO/./train_lora.py", line 321, in main train(config, prompts) File "/content/LECO/./train_lora.py",...

As I mentioned in https://github.com/p1atdev/LECO/issues/39, the jax library has [deprecated](https://jax.readthedocs.io/en/latest/changelog.html#jax-0-4-16-sept-18-2023) `jax.random.keyArray`. It was removed entirely in 0.4.24, so LECO has a broken dependency. ```bash /usr/local/lib/python3.10/dist-packages/transformers/utils/generic.py:311: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please...

Hi! ![1fbd895b06b2f8f720be7175a48c10be](https://github.com/p1atdev/LECO/assets/118778759/e0ba3f30-fc1d-4cc0-b7a8-6cc02cba3341) I'm having this problem trainng LECO... any solution to this? Thanks in advance

im on a 4090 and use sdp attention with cuda 12.1, xformers is limited to 11...I set xformers to false, as i assumed it would not not be used. however,...

- ⚠️ Classic training method option (experimental) - ⚠️ SDXL support on `sdxl` branch (experimental) - 🚧 Custom CLIP skip - 🚧 Noise offset? - 🚧 Pyramid noise? - 🛑...

Tried and some are getting a different error than I am, but the train_lora_XL.py script is broken. My error is that the optimizer args is empty.

First, I love this project. I have seen what others are making with it and it seems really powerful for fine tuning exactly the prompt you want. The only way...

I want head wings to be attached to every characters head but no wings on the back. When I prompt head wings I will sometimes get head wings AND back...