CreamyLong comments

Results 37 comments of


                                            CreamyLong

predicted_indices not found in losses

> You can try to delete predicted_indices on autoencoder.py. I did on this https://github.com/SerdarHelli/latent-diffusion/blob/main/ldm/models/autoencoder.py . Then, probably you will get an error about version which is undefined. I just deleted....

Details about training inpainting model

The code and steps of training inpainting model is [here ](https://github.com/CreamyLong/stable-diffusion/tree/master?tab=readme-ov-file#Inpainting)

[Bug]: NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs:

Having the same issue, win10，RTX 3060，cuda11.1.

Where can I get the one good with 256x256

you could use this model _https://ommer-lab.com/files/latent-diffusion/text2img.zip_

Inference code for super-resolution?

There is a [link ](https://github.com/CreamyLong/stable-diffusion/tree/master?tab=readme-ov-file#super-resolution)for super resoulution

How to train and generate class conditional images based on custom data sets

There are some [config file](https://github.com/CreamyLong/stable-diffusion/tree/master/configs/latent-diffusion/conditional) may help you understand how to train

correct configuration for sd-v1-4.ckpt

_size mismatch for text_model.final_layer_norm.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([512])._ Fixed by install transformers==4.19.2.

How to train on latent size 128?

Sure, you need to train a new VQVAE model, and the image size is 512 if the f=4.

How to finetune inpainting on custom dataset？

This may help you https://github.com/CreamyLong/stable-diffusion/blob/master/scripts/train/train_inpaint.sh

How to retrain a image-condition model on my own dataset?

This [config](https://github.com/CreamyLong/stable-diffusion/tree/master/configs/latent-diffusion/conditional) is for conditional generation