File size: 3,763 Bytes
18ce093 e841823 18ce093 e8d5783 18ce093 e841823 18ce093 |
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 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 |
import gradio as gr
from convert_diffusion_to_gguf import SUPPORTED_ARCHS, qconfig_map, convert
from huggingface_hub import create_repo, upload_file
from argparse import Namespace
from pathlib import Path
def upload(args, outfile):
url = ""
if args.host_repo_id and args.hf_token:
repo_id = create_repo(args.host_repo_id, repo_type="model", exist_ok=True, token=args.hf_token).repo_id
info = upload_file(repo_id=repo_id, path_in_repo=str(outfile), path_or_fileobj=str(outfile), token=args.token)
url = info.commit_url
print(f"Uploaded to {url}")
return url
def go_gguf(model_repo_id, subfolder, arch, outtype, outfile_name, bigendian, verbose, host_repo_id, hf_token):
args = Namespace(
model=model_repo_id,
subfolder=subfolder,
arch=arch,
outtype=outtype,
outfile=Path(outfile_name),
bigendian=bigendian,
verbose=verbose,
host_repo_id=host_repo_id,
hf_token=hf_token,
)
try:
convert(args)
return upload(args)
except Exception as e:
return e
with gr.Blocks(theme=gr.themes.Soft()) as demo:
gr.Markdown("<h1><center>GGUF Converter for Diffusers format model checkpoints</center></h1>")
gr.Markdown(
"Convert `diffusers` format model checkpoints from the Hub to GGUF format and optionally upload them back. Based on [this repo](https://github.com/ngxson/diffusion-to-gguf)."
)
with gr.Row():
with gr.Column(scale=1):
gr.Markdown("### 📥 Input Model")
model_repo_id = gr.Textbox(label="Model Repo ID", placeholder="e.g., Qwen/Qwen-Image")
subfolder = gr.Textbox(label="Subfolder (Optional)", placeholder="e.g., transformer")
gr.Markdown("### ⚙️ Conversion Settings")
arch = gr.Dropdown(choices=SUPPORTED_ARCHS, label="Architecture")
outtype = gr.Dropdown(choices=list(qconfig_map.keys()), label="Quantization Type", value="F16")
outfile_name = gr.Textbox(label="Output Filename", value="{ftype}.gguf")
with gr.Accordion("Advanced Settings", open=False):
bigendian = gr.Checkbox(label="Use Big Endian")
verbose = gr.Checkbox(label="Verbose Logging", value=True)
gr.Markdown("### 📤 Upload to Hub (Optional)")
host_repo_id = gr.Textbox(label="Your Hub Repo ID", placeholder="e.g., YourUsername/My-GGUFs")
hf_token = gr.Textbox(label="Hugging Face Token", type="password", placeholder="hf_...")
convert_btn = gr.Button("Convert & Upload", variant="primary")
with gr.Column(scale=2):
gr.Markdown("### 🚀 Result")
url_output = gr.Markdown()
gr.Examples(
examples=[
[
"black-forest-labs/FLUX.1-schnell",
"transformer",
"flux",
"Q4_0",
"flux-schnell-q4.gguf",
False,
False,
"YourUsername/MyGGUFs",
"hf_...",
],
[
"Qwen/Qwen-Image",
"transformer",
"qwen",
"Q8_0",
"qwen-q4.gguf",
False,
False,
"YourUsername/MyGGUFs",
"hf_...",
],
],
inputs=[model_repo_id, subfolder, arch, outtype, outfile_name, bigendian, verbose, host_repo_id, hf_token],
)
convert_btn.click(
fn=go_gguf,
inputs=[model_repo_id, subfolder, arch, outtype, outfile_name, bigendian, verbose, host_repo_id, hf_token],
outputs=[url_output],
)
if __name__ == "__main__":
demo.launch()
|