How to train stable-diffusion-xl-base-1.0 without lora?

[KimmiShi]

Hi, I want to train stable-diffusion-xl-base-1.0 without lora, how to do this?

I can run train_text_to_image_lora_sdxl.py .
But train_text_to_image.py with MODEL_NAME="stabilityai/stable-diffusion-xl-base-1.0" with raise an error:

diffusers/models/unet_2d_condition.py:836 in forward                                                        │
│   833 │   │   │   aug_emb = self.add_embedding(text_embs, image_embs)        │
│   834 │   │   elif self.config.addition_embed_type == "text_time":           │
│   835 │   │   │   # SDXL - style                                             │
│ ❱ 836 │   │   │   if "text_embeds" not in added_cond_kwargs:                 │
│   837 │   │   │   │   raise ValueError(                                      │
│   838 │   │   │   │   │   f"{self.__class__} has the config param `addition_ │
│   839 │   │   │   │   )                                                      │
╰──────────────────────────────────────────────────────────────────────────────╯
TypeError: argument of type 'NoneType' is not iterable

the added_cond_kwargs is none in this case.

[sayakpaul]

It won't work. I am working on #4505 to fix this.

[KimmiShi]

It won't work. I am working on #4505 to fix this.

Thanks, I found your PR just a short while ago and I am testing this code.

[muyuWang]

Hi, I have the same problem as you mentioned. Have you fixed it already？ And could you tell me how you solved this? Thanks

[sayakpaul]

Refer to #4505.