README.md

metadata

dataset_info:
  features:
    - name: Prompt
      dtype: string
    - name: Category
      dtype: string
    - name: Challenge
      dtype: string
    - name: Note
      dtype: string
    - name: images
      dtype: image
    - name: model_name
      dtype: string
    - name: seed
      dtype: int64
  splits:
    - name: train
      num_bytes: 166170790
      num_examples: 1632
  download_size: 166034308
  dataset_size: 166170790

Images of Parti Prompts for "if-v-1.0"

Code that was used to get the results:

from diffusers import DiffusionPipeline
import torch

pipe_low = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-XL-v1.0", safety_checker=None, watermarker=None, torch_dtype=torch.float16, variant="fp16")
pipe_low.enable_model_cpu_offload()

pipe_up = DiffusionPipeline.from_pretrained("DeepFloyd/IF-II-L-v1.0", safety_checker=None, watermarker=None, text_encoder=pipe_low.text_encoder, torch_dtype=torch.float16, variant="fp16")
pipe_up.enable_model_cpu_offload()

prompt = "" # a parti prompt
generator = torch.Generator("cuda").manual_seed(0)

prompt_embeds, negative_prompt_embeds = pipe_low.encode_prompt(prompt)
images = pipe_low(prompt_embeds=prompt_embeds, negative_prompt_embeds=negative_prompt_embeds, num_inference_steps=100, generator=generator, output_type="pt").images
images = pipe_up(prompt_embeds=prompt_embeds, negative_prompt_embeds=negative_prompt_embeds, image=images, num_inference_steps=100, generator=generator).images[0]