metadata
base_model: spamsoms/LCM-kotosmix_diffusers
tags:
- openvino
- openvino-export
This model was converted to OpenVINO from spamsoms/LCM-kotosmix_diffusers using optimum-intel
via the export space.
First make sure you have optimum-intel installed:
pip install optimum[openvino]
To load your model you can do as follows:
import huggingface_hub as hf_hub
from optimum.intel import OVStableDiffusionPipeline
from diffusers import LCMScheduler
import torch
model_id = "hsuwill000/LCM-kotosmix_diffusers-openvino"
HIGH = 1024
WIDTH = 1024
batch_size = -1 # Or set it to a specific positive integer if needed
prompt="agirl, anime,"
negative_prompt="(deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy,\
extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, \
mutation, mutated, ugly, disgusting, blurry, amputation"
pipe = OVStableDiffusionPipeline.from_pretrained(
model_id,
compile=False,
ov_config={"CACHE_DIR": ""},
torch_dtype=torch.bfloat16, # More standard dtype for speed
safety_checker=None,
use_safetensors=False,
)
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
print(pipe.scheduler.compatibles)
pipe.reshape(batch_size=batch_size, height=HIGH, width=WIDTH, num_images_per_prompt=1)
pipe.compile()
image = pipe(
prompt=prompt,
negative_prompt=negative_prompt,
width=WIDTH,
height=HIGH,
guidance_scale=2,
num_inference_steps=4,
num_images_per_prompt=1,
).images[0]
image.save("test.png")