[BUG]: Use Teyvat but produce blur grid pictures

[FrankieDong]

🐛 Describe the bug

I am use the default teyvat.yaml to train a simple model, only modify the data->image_transform->size from 512 to 256 because of my gpu devices. and set nums_workers 0.

Environment

No response

[FrankieDong]

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

[feifeibear]

Hello, how long did you train the model. Epoches? Iterations? Considering that train a diffusion model from scratch is very slow to converge.

[FrankieDong]

Hello, how long did you train the model. Epoches? Iterations? Considering that train a diffusion model from scratch is very slow to converge.

I set epoch 10, around 10 hours to produce the last.ckpt

[yufengyao-lingoace]

This is because the code is stable diffusion 2.0, and it has no pre-trained model. So after a small epoch, it produces a picture with random noise.

[binmakeswell]

We have updated a lot. This issue was closed due to inactivity. Thanks.