textual_inversion

textual_inversion copied to clipboard

Reconstruction and Editability benchmark for experiments

4

Hello there! Thanks for providing a low level evaluation for benchmarking the quality of the latent space embeddings . 👍 💯 I have some questions regarding **Reconstruction** and Editability concepts:...

SoyGema

How to use with stable diffusion repo?

1

I have stable diffusion locally in a separate directory, working fine with the v1.4 ckpt. I generated embeddings.pt using textual inversion. How can I add the embedding to stable diffusion...

Nughu

The issue this resolves was closed without resolution, but it was bugging me enough and wasting my GPU time too much that I felt a need to fix it. Adds...

TaleirOfDeynai

#### What this does: This PR works to resolve all deprecation warnings and hard incompatibilities with current versions of PyTorch-Lightning. It also sets `1.7.7` as the requested version in `environment.yaml`....

TaleirOfDeynai

Artifacts arising from 256x256 data

5

Due to GPU constraints (RTX 3070, 8gb VRAM), I lowered my image training dimensions to half the 512x512 standard. Working with faces, a natural problem arose when priming the training...

GucciFlipFlops1917

Any way to force training into ignoring background and focus just on a subject ?

13

I can only get likeness with heavy overfitting but this also heavily learns background , any way to avoid it? white background? black? or a piece of code ?

1blackbar

NameError: name 'trainer' is not defined

4

``` (ldm) C:\Users\ekzer\stable_diffusion\textual_inversion-main>python main.py --base configs/stable-diffusion/v1-finetune.yaml -t --actual_resume models/ldm/text2img-large/modelSD.ckpt -n art --gpus 0, --data_root img/datasets/cat_statue --init_word art --placeholder_tokens "*art cat" --no-test Global seed set to 23 Running on GPUs 0,...

EKZERSIS

Is all the patching in ldm/modules/encoders/modules.py actually necessary?

1

This is more of a discussion than a bug/feature request. I've been using this repository and the AUTOMATIC1111 repository as references to implement textual inversion into my local version. In...

zwishenzug

Reproducing the results in the paper

25

Hi, I'm trying to reproduce the paper results, especially the 'cat_statue' concept below in Figure 1, but it seems like extra information is needed to do that.  ### Inversion...

shadow2496

How about training on more images of a domin? For example, 100~200 images?

1

In the paper, the training images is about 3-5 number of one object or 3-5 number of one style images. How about training on more images of a domin? For...

universewill