CiaraStrawberry
commented
1 month ago Very likely a diffusers issue, I reccomend running something like this
import torch from diffusers import StableDiffusion3Pipeline
pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3-medium-diffusers", torch_dtype=torch.float16) pipe = pipe.to("cuda")
image = pipe( "A cat holding a sign that says hello world", negative_prompt="", num_inference_steps=28, guidance_scale=7.0, ).images[0] image
and see if you get the same issue.
Hi, I tried testing demo_sd3_instruct.py but for some reason it's not working. It tries to do something but then it stops the execution prematurely without errors. I think there is an issue caused by the following part:
pipe = StableDiffusion3PipelineExtraCFG.from_pretrained( base_model_path, feature_extractor=None, safety_checker=None, torch_dtype=torch.float16,
)
The output is:
2.4.0+cu124 2024-08-12 08:45:43.807334: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable
TF_ENABLE_ONEDNN_OPTS=0.2024-08-12 08:45:45.176026: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable
TF_ENABLE_ONEDNN_OPTS=0.Keyword arguments {'feature_extractor': None, 'safety_checker': None} are not expected by StableDiffusion3PipelineExtraCFG and will be ignored. Loading pipeline components...: 22%|██████████████████ | 2/9 [00:00<00:03, 2.16it/s]You set
add_prefix_space. The tokenizer needs to be converted from the slow tokenizers Loading pipeline components...: 44%|████████████████████████████████████ | 4/9 [00:01<00:02, 2.28it/s]Maybe there is something I am missing or there is an issue in the libraries I am using?
Kind Regards