search

CompVis / latent-diffusion

High-Resolution Image Synthesis with Latent Diffusion Models
MIT License
11.82k stars 1.53k forks source link

How to finetune inpainting? #151

Open TalTaiber opened 2 years ago

TalTaiber commented 2 years ago

Thank you for the excellent work.

I am trying to finetune the inpainting model. I have a dataset with this structure:

.../inpaint_ds/
├── train
│   ├── 0.png
│   ├── 1.png
│   ├── ...
├── val
│   ├── 0.png
│   ├── 0_mask.png
│   ├── ...

Which config file should I use? This open issue suggests using this config, but it does not have a 'data' section (also, it is unclear which loss should be defined for the model under 'lossconfig'). Could you please upload the config file used for training the inpainting model?

Cheers!

ImmortalSdm commented 1 year ago

I'm facing with the same problem, any update? Thx a lot! @TalTaiber @asanakoy

shensongli commented 1 year ago

hello,how did you train the inpainting model with your own dataset?Could you give me some advice about dataset setting.Thank you very much.

dreamlychina commented 1 year ago

hello,how did you train the inpainting model with your own dataset?Could you give me some advice about dataset setting.Thank you very much.

你可以训练了？？求救

Xijieupenn commented 1 year ago

Should be help to read #211

nickyisadog commented 1 year ago

hello,how did you train the inpainting model with your own dataset?Could you give me some advice about dataset setting.Thank you very much.

你可以训练了？？求救

Hello, I have simplified the inpaint fine tuning and made some inference example in my repo. Feel free to check it

https://github.com/nickyisadog/latent-diffusion-inpainting/tree/main