PhotoMaker
PhotoMaker copied to clipboard
TencentARC
PhotoMaker
Fetching 18 files: 17%|██████████▌ | 3/18 [00:01
It seems that both parties avoid mentioning each other in the paper and code repository https://github.com/modelscope/facechain
   如此5张图 差距比较大啊
I use demo "https://github.com/TencentARC/PhotoMaker/blob/main/photomaker_style_demo.ipynb " in google colab, and use v100 16G vram, but it throw exception :  how can i fix it, only increase more vram ? colab...
```python python gradio_demo/app.py Traceback (most recent call last): File "E:\2023_Code_WEB\PhotoMaker-main\gradio_demo\app.py", line 10, in import spaces File "E:\anaconda3\envs\photomaker\lib\site-packages\spaces\__init__.py", line 10, in from .zero.decorator import GPU File "E:\anaconda3\envs\photomaker\lib\site-packages\spaces\zero\decorator.py", line 18, in from...
ValueError: PhotoMaker currently does not support multiple trigger words in a single prompt. Trigger word: img, Prompt: a beautiful girl on Italy street, best quality.
The modification replaces this method with a token-based approach using a tokenizer. This tokenizes the entire prompt, identifies the token corresponding to the trigger word, removes it, and then decodes...
We have seen very good results on SDXL. How does this method perform on SD1.5? Are there any plans to offer a version for SD1.5? Thank you very much.
``` base_model_path = 'SG161222/RealVisXL_V3.0' photomaker_path = 'checkpoints/photomaker-v1.bin' device = "cuda" save_path = "./outputs" pipe = PhotoMakerStableDiffusionXLPipeline.from_pretrained( base_model_path, torch_dtype=torch.bfloat16, use_safetensors=True, variant="fp16", # local_files_only=True, ).to(device) # Load PhotoMaker checkpoint pipe.load_photomaker_adapter( os.path.dirname(photomaker_path), subfolder="",...
