File size: 2,030 Bytes
5a0ae56 286875c 5a0ae56 5d7f667 5a0ae56 286875c 5d7f667 5a0ae56 6e2f4ca f1f2c1f 5a0ae56 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
from typing import Dict, List, Any
import torch
from torch import autocast
from huggingface_hub import hf_hub_download
from diffusers import DiffusionPipeline
import base64
from io import BytesIO
from safetensors.torch import load_file
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print("device ~>", device)
class EndpointHandler:
def __init__(self, path=""):
print("path ~>", path)
self.pipe = DiffusionPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
torch_dtype=torch.float16 if device.type == "cuda" else None,
variant="fp16",
).to(device)
self.pipe.load_lora_weights("SvenN/sdxl-emoji", weight_name="lora.safetensors")
self.pipe.fuse_lora(lora_scale=0.6)
embedding_path = hf_hub_download(
repo_id="SvenN/sdxl-emoji", filename="embeddings.pti", repo_type="model"
)
state_dict = load_file(embedding_path)
self.pipe.load_textual_inversion(state_dict["text_encoders_0"], token=["<s0>", "<s1>"], text_encoder=self.pipe.text_encoder, tokenizer=self.pipe.tokenizer)
self.pipe.load_textual_inversion(state_dict["text_encoders_1"], token=["<s0>", "<s1>"], text_encoder=self.pipe.text_encoder_2, tokenizer=self.pipe.tokenizer_2)
def __call__(self, data: Any) -> List[List[Dict[str, float]]]:
"""
Args:
data (:obj:):
includes the input data and the parameters for the inference.
Return:
A :obj:`dict`:. base64 encoded image
"""
inputs = data.pop("inputs", data)
# Automatically add trigger tokens to the beginning of the prompt
images = self.pipe(
inputs,
**data['parameters']
).images
image = images[0]
return image
if __name__ == "__main__":
handler = EndpointHandler()
print(handler)
output = handler({"inputs": "emoji of a tiger face, white background"})
print(output)
|