File size: 2,428 Bytes
5545f96
 
 
fd40e27
 
5545f96
4b5b300
5545f96
fd40e27
5545f96
 
 
 
 
 
 
 
fd40e27
5545f96
4b5b300
 
 
5545f96
fd40e27
5545f96
 
 
 
 
 
 
 
 
fd40e27
5545f96
 
 
 
 
 
fd40e27
 
 
 
 
 
 
 
 
5545f96
 
 
 
 
 
 
 
fd40e27
5545f96
 
 
 
 
 
 
 
 
 
 
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
62
63
64
65
66
67
68
69
70
"""Hugging Face Spaces entry point for the published CSA release."""

import hashlib
import io
import json
import os
import sys
from pathlib import Path
from zipfile import ZipFile

if os.environ.get("SPACE_ID") and os.environ.get("CSA_DEVICE", "cuda") == "cuda":
    import spaces
else:
    spaces = None

import torch
from huggingface_hub import hf_hub_download
from PIL import Image

# Spaces installs dependencies before copying the repository into the image.
sys.path.insert(0, str(Path(__file__).with_name("csa_inpainting-0.3.0-py3-none-any.whl")))

from csa_inpainting.inference import load_inference_generator, predict_stages
from csa_inpainting.web_demo import CSS, Source, build_demo

MODEL = "KumaPower/CSA"
REVISION = "bf82a33eb70a6a0182180ae9f880bb7d1a08c439"
FILENAME = "csa-stage2-step97397.pt"
SHA256 = "605b80c18dcb0856367e6276813df0b54ec42aabea782af1631052aedb7abdfe"

def download(filename):
    return Path(hf_hub_download(MODEL, filename, revision=REVISION, token=False))

checkpoint = Path(os.environ["CSA_CHECKPOINT"]) if "CSA_CHECKPOINT" in os.environ else download(FILENAME)
with checkpoint.open("rb") as stream:
    if hashlib.file_digest(stream, "sha256").hexdigest() != SHA256:
        raise ValueError("CSA release checkpoint checksum mismatch")

torch.set_num_threads(2)
model, config = load_inference_generator(checkpoint, torch.device("cuda" if spaces else "cpu"))

examples = {}
with ZipFile(Path(__file__).with_name("examples.zip")) as archive:
    for example in json.loads(archive.read("index.json")):
        label = example["label"]
        with Image.open(io.BytesIO(archive.read(example["image"]))) as image:
            source = Source(image.convert("RGB"), label)
        with Image.open(io.BytesIO(archive.read(example["mask"]))) as mask:
            examples[label] = source, mask.convert("L")

def run_stages(image, mask):
    rough, refined = predict_stages(model, image, mask)
    return rough.cpu(), refined.cpu()

if spaces is not None:
    run_stages = spaces.GPU(duration=10)(run_stages)

demo, theme = build_demo(model, config, examples, FILENAME, run_stages=run_stages)

if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0" if os.environ.get("SPACE_ID") else "127.0.0.1",
        server_port=int(os.environ.get("CSA_PORT", "7860")),
        share=False,
        theme=theme,
        css=CSS,
        max_file_size="16mb",
        show_error=False,
    )