uP

README.md

@@ -1,14 +1,13 @@
 ---
-title: SD To Diffusers
+title: SD-XL To Diffusers
 emoji: 🎨➡️🧨
 colorFrom: indigo
-colorTo: blue
+colorTo: purple
 sdk: gradio
 sdk_version: 3.31.0
 app_file: app.py
 pinned: true
 license: mit
-duplicated_from: diffusers/sd-to-diffusers
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -7,8 +7,7 @@ The steps are the following:
 
 - Paste a read-access token from hf.co/settings/tokens. Read access is enough given that we will open a PR against the source repo.
 - Input a model id from the Hub
-- Input the filename from the root dir of the repo that you would like to convert, e.g. 'v2-1_768-ema-pruned.ckpt' or 'v1-5-pruned.safetensors'
-- Chose which Stable Diffusion version, image size, scheduler type the model has and whether you want the "ema", or "non-ema" weights.
+- Input the filename from the root dir of the repo that you would like to convert, e.g. 'xl-pruned.safetensors'
 - Click "Submit"
 - That's it! You'll get feedback if it works or not, and if it worked, you'll get the URL of the opened PR 🔥
 
@@ -16,7 +15,7 @@ The steps are the following:
 """
 
 demo = gr.Interface(
-    title="Convert any Stable Diffusion checkpoint to Diffusers and open a PR",
+    title="Convert any Stable Diffusion XL checkpoint to Diffusers and open a PR",
     description=DESCRIPTION,
     allow_flagging="never",
     article="Check out the [Diffusers repo on GitHub](https://github.com/huggingface/diffusers)",
@@ -24,10 +23,6 @@ demo = gr.Interface(
         gr.Text(max_lines=1, label="your_hf_token"),
         gr.Text(max_lines=1, label="model_id"),
         gr.Text(max_lines=1, label="filename"),
-        gr.Radio(label="Model type", choices=["v1", "v2", "ControlNet"]),
-        gr.Radio(label="Sample size (px)", choices=[512, 768]),
-        gr.Radio(label="Scheduler type", choices=["pndm", "heun", "euler", "dpm", "ddim"], value="dpm"),
-        gr.Radio(label="Extract EMA or non-EMA?", choices=["ema", "non-ema"], value="ema"),
     ],
     outputs=[gr.Markdown(label="output")],
     fn=convert,

convert.py

@@ -3,6 +3,7 @@ import requests
 import os
 import shutil
 from pathlib import Path
+from typing import Any
 from tempfile import TemporaryDirectory
 from typing import Optional
 
@@ -11,49 +12,22 @@ from io import BytesIO
 
 from huggingface_hub import CommitInfo, Discussion, HfApi, hf_hub_download
 from huggingface_hub.file_download import repo_folder_name
-from diffusers.pipelines.stable_diffusion.convert_from_ckpt import (
-    download_from_original_stable_diffusion_ckpt, download_controlnet_from_original_ckpt
-)
+from diffusers import StableDiffusionXLPipeline
 from transformers import CONFIG_MAPPING
 
 
-COMMIT_MESSAGE = " This PR adds fp32 and fp16 weights in PyTorch and safetensors format to {}"
+COMMIT_MESSAGE = " This PR adds fp32 and fp16 weights in safetensors format to {}"
 
 
-def convert_single(model_id: str, filename: str, model_type: str, sample_size: int, scheduler_type: str, extract_ema: bool, folder: str, progress):
-    from_safetensors = filename.endswith(".safetensors")
-
+def convert_single(model_id: str, filename: str, folder: str, progress: Any):
     progress(0, desc="Downloading model")
     local_file = os.path.join(model_id, filename)
     ckpt_file = local_file if os.path.isfile(local_file) else hf_hub_download(repo_id=model_id, filename=filename)
 
-    if model_type == "v1":
-        config_url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml"
-    elif model_type == "v2":
-        if sample_size == 512:
-            config_url = "https://raw.githubusercontent.com/Stability-AI/stablediffusion/main/configs/stable-diffusion/v2-inference.yaml"
-        else:
-            config_url = "https://raw.githubusercontent.com/Stability-AI/stablediffusion/main/configs/stable-diffusion/v2-inference-v.yaml"
-    elif model_type == "ControlNet":
-        config_url = (Path(model_id)/"resolve/main"/filename).with_suffix(".yaml")
-        config_url = "https://huggingface.co/" + str(config_url)
-
-    config_file = BytesIO(requests.get(config_url).content)
-
-    if model_type == "ControlNet":
-        progress(0.2, desc="Converting ControlNet Model")
-        pipeline = download_controlnet_from_original_ckpt(ckpt_file, config_file, image_size=sample_size, from_safetensors=from_safetensors, extract_ema=extract_ema)
-        to_args = {"dtype": torch.float16}
-    else:
-        progress(0.1, desc="Converting Model")
-        pipeline = download_from_original_stable_diffusion_ckpt(ckpt_file, config_file, image_size=sample_size, scheduler_type=scheduler_type, from_safetensors=from_safetensors, extract_ema=extract_ema)
-        to_args = {"torch_dtype": torch.float16}
-
-    pipeline.save_pretrained(folder)
-    pipeline.save_pretrained(folder, safe_serialization=True)
+    pipeline = StableDiffusionXLPipeline.from_single_file(ckpt_file)
 
-    pipeline = pipeline.to(**to_args)
-    pipeline.save_pretrained(folder, variant="fp16")
+    pipeline.save_pretrained(folder, safe_serialization=True)
+    pipeline = pipeline.to(torch_dtype=torch.float16)
     pipeline.save_pretrained(folder, safe_serialization=True, variant="fp16")
 
     return folder
@@ -71,7 +45,7 @@ def previous_pr(api: "HfApi", model_id: str, pr_title: str) -> Optional["Discuss
                 return discussion
 
 
-def convert(token: str, model_id: str, filename: str, model_type: str, sample_size: int = 512, scheduler_type: str = "pndm", extract_ema: bool = True, progress=gr.Progress()):
+def convert(token: str, model_id: str, filename: str, progress=gr.Progress()):
     api = HfApi()
 
     pr_title = "Adding `diffusers` weights of this model"
@@ -81,7 +55,7 @@ def convert(token: str, model_id: str, filename: str, model_type: str, sample_si
         os.makedirs(folder)
         new_pr = None
         try:
-            folder = convert_single(model_id, filename, model_type, sample_size, scheduler_type, extract_ema, folder, progress)
+            folder = convert_single(model_id, filename, folder, progress)
             progress(0.7, desc="Uploading to Hub")
             new_pr  = api.upload_folder(folder_path=folder, path_in_repo="./", repo_id=model_id, repo_type="model", token=token, commit_message=pr_title, commit_description=COMMIT_MESSAGE.format(model_id), create_pr=True)
             pr_number = new_pr.split("%2F")[-1].split("/")[0]