Spaces:
Runtime error
Runtime error
Create server.py
Browse files
server.py
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import sys
|
| 3 |
+
import base64
|
| 4 |
+
from io import BytesIO
|
| 5 |
+
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
| 6 |
+
|
| 7 |
+
from fastapi import FastAPI
|
| 8 |
+
import numpy as np
|
| 9 |
+
from PIL import Image
|
| 10 |
+
|
| 11 |
+
import clip
|
| 12 |
+
from dalle.models import Dalle
|
| 13 |
+
from dalle.utils.utils import clip_score, download
|
| 14 |
+
|
| 15 |
+
print("Loading models...")
|
| 16 |
+
app = FastAPI()
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
url = "https://arena.kakaocdn.net/brainrepo/models/minDALL-E/57b008f02ceaa02b779c8b7463143315/1.3B.tar.gz"
|
| 20 |
+
root = os.path.expanduser("~/.cache/minDALLE")
|
| 21 |
+
filename = os.path.basename(url)
|
| 22 |
+
pathname = filename[: -len(".tar.gz")]
|
| 23 |
+
download_target = os.path.join(root, filename)
|
| 24 |
+
result_path = os.path.join(root, pathname)
|
| 25 |
+
if not os.path.exists(result_path):
|
| 26 |
+
result_path = download(url, root)
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
device = "cpu"
|
| 30 |
+
model = Dalle.from_pretrained(result_path) # This will automatically download the pretrained model.
|
| 31 |
+
model.to(device=device)
|
| 32 |
+
model_clip, preprocess_clip = clip.load("ViT-B/32", device=device)
|
| 33 |
+
model_clip.to(device=device)
|
| 34 |
+
|
| 35 |
+
print("Models loaded !")
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
@app.get("/")
|
| 39 |
+
def read_root():
|
| 40 |
+
return {"minDALL-E!"}
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
@app.get("/{generate}")
|
| 44 |
+
def generate(prompt):
|
| 45 |
+
images = sample(prompt)
|
| 46 |
+
images = [to_base64(image) for image in images]
|
| 47 |
+
return {"images": images}
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
def sample(prompt):
|
| 51 |
+
# Sampling
|
| 52 |
+
images = (
|
| 53 |
+
model.sampling(prompt=prompt, top_k=256, top_p=None, softmax_temperature=1.0, num_candidates=3, device=device)
|
| 54 |
+
.cpu()
|
| 55 |
+
.numpy()
|
| 56 |
+
)
|
| 57 |
+
images = np.transpose(images, (0, 2, 3, 1))
|
| 58 |
+
|
| 59 |
+
# CLIP Re-ranking
|
| 60 |
+
rank = clip_score(
|
| 61 |
+
prompt=prompt, images=images, model_clip=model_clip, preprocess_clip=preprocess_clip, device=device
|
| 62 |
+
)
|
| 63 |
+
images = images[rank]
|
| 64 |
+
|
| 65 |
+
pil_images = []
|
| 66 |
+
for i in range(len(images)):
|
| 67 |
+
im = Image.fromarray((images[i] * 255).astype(np.uint8))
|
| 68 |
+
pil_images.append(im)
|
| 69 |
+
|
| 70 |
+
return pil_images
|
| 71 |
+
|
| 72 |
+
|
| 73 |
+
def to_base64(pil_image):
|
| 74 |
+
buffered = BytesIO()
|
| 75 |
+
pil_image.save(buffered, format="JPEG")
|
| 76 |
+
return base64.b64encode(buffered.getvalue())
|