Bad Reconstruction Results

[Fire-friend]

Thank you for your amazing work. I attempt to reconstruct the original image using the provided code and image, but the outcome is unsatisfactory. Can you offer any advice?

from src.eunms import Model_Type, Scheduler_Type
from src.utils.enums_utils import get_pipes
from src.config import RunConfig
from PIL import Image
from main import run as invert

model_type = Model_Type.SDXL_Turbo
scheduler_type = Scheduler_Type.DDIM
pipe_inversion, pipe_inference = get_pipes(model_type, scheduler_type, device='cuda')

input_image = Image.open("example_images/lion.jpeg").convert("RGB").resize((512, 512))
prompt = ""

config = RunConfig(model_type = model_type,
scheduler_type = scheduler_type)

edit_img, inv_latent, noise, all_latents = invert(input_image,
prompt,
config,
pipe_inversion=pipe_inversion,
pipe_inference=pipe_inference,
do_reconstruction=True,
# edit_prompt="a lion in the field"
)

edit_img.save("result.png")

[danielgaribi]

Do you use the same environment as mentioned in the requirements file?