|
|
|
from diffusers import DiffusionPipeline, DDIMScheduler |
|
import argparse |
|
import torch |
|
from datasets import load_dataset |
|
import PIL |
|
|
|
IMAGE_OUTPUT_SIZE = (256, 256) |
|
NUM_INFERENCE_STEPS = 100 |
|
|
|
def resize(image: PIL.Image): |
|
return image.resize(IMAGE_OUTPUT_SIZE, resample=PIL.Image.Resampling.LANCZOS) |
|
|
|
def get_sd_eval(ckpt, guidance_scale=7.5): |
|
pipe = DiffusionPipeline.from_pretrained(ckpt, torch_dtype=torch.float16) |
|
pipe.to("cuda") |
|
pipe.scheduler = DDIMScheduler.from_config(pipe.config) |
|
|
|
def sd_eval(prompt): |
|
images = pipe(prompt, num_inference_steps=100, guidance_scale=guidance_scale).images |
|
images = [resize(image) for image in images] |
|
return images |
|
|
|
return sd_eval |
|
|
|
def get_karlo_eval(ckpt): |
|
pipe = DiffusionPipeline.from_pretrained(ckpt, torch_dtype=torch.float16) |
|
pipe.to("cuda") |
|
|
|
def karlo_eval(prompt): |
|
images = pipe(prompt, prior_num_inference_steps=50, decoder_num_inference_steps=100).images |
|
return images |
|
|
|
return karlo_eval |
|
|
|
def get_if_eval(ckpt): |
|
pipe_low = DiffusionPipeline.from_pretrained(ckpt, torch_dtype=torch.float16) |
|
pipe_low.enable_model_cpu_offload() |
|
|
|
pipe_up = DiffusionPipeline.from_pretrained("DeepFloyd/IF-II-L-v1.0", text_encoder=pipe_low.text_encoder, torch_dtype=torch.float16) |
|
pipe_up.enable_model_cpu_offload() |
|
|
|
def sd_eval(prompt): |
|
images = pipe_low(prompt, num_inference_steps=100, output_type="pt").images |
|
images = pipe_up(promtp=prompt, images=images, num_inference_steps=100).images |
|
return images |
|
|
|
return sd_eval |
|
|
|
MODELS = { |
|
"runwayml/stable-diffusion-v1-5": get_sd_eval, |
|
"stabilityai/stable-diffusion-v2-1": get_sd_eval, |
|
"kakaobrain/karlo-alpha": get_karlo_eval, |
|
"DeepFloyd/IF-I-XL-v1.0": get_if_eval, |
|
} |
|
|
|
|
|
|
|
|
|
if __name__ == "__main__": |
|
parser = argparse.ArgumentParser(description='Run Parti Prompt Evaluation') |
|
parser.add_argument('model_repo_or_id', type=str, help='ID or URL of the model repository.', required=True) |
|
parser.add_argument('--dataset_repo_or_id', type=str, default='diffusers/prompt_generations', help='ID or URL of the dataset repository (default: "diffusers/prompt_generations")') |
|
parser.add_argument('--batch_size', type=int, default=8, help="Batch size for the eval function") |
|
parser.add_argument('--upload_to_hub', action='store_true', help='whether to upload the dataset to the Hugging Face dataset hub') |
|
|
|
args = parser.parse_args() |
|
|
|
eval_fn = MODELS[args.model_repo_or_id](args.model_repo_or_id) |
|
|
|
dataset = load_dataset("nateraw/parti-prompts") |
|
|
|
def map_fn(batch): |
|
batch["images"] = eval_fn(batch["prompt"]) |
|
return batch |
|
|
|
dataset_images = dataset.map(map_fn, batched=True, batch_size=8) |
|
|
|
if args.upload_to_hub: |
|
dataset.push_to_hub(args.dataset_repo_or_id) |
|
else: |
|
dataset.save_to_disk(args.dataset_repo_or_id.split("/")[-1]) |
|
|
