iexamples

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,6 +1,6 @@
 ---
 title: Layerdiffusion Gradio Unofficial
-emoji: 🏆
+emoji: 🍰
 colorFrom: pink
 colorTo: blue
 sdk: gradio

app.py

@@ -12,6 +12,7 @@ from utils import (
     postprocess_image,
     preprocess_image,
     downloadModels,
+    examples,
 )
 
 sys.path.append(os.path.dirname("./ComfyUI/"))
@@ -42,7 +43,9 @@ downloadModels()
 
 with torch.inference_mode():
     ckpt_load_checkpoint = CheckpointLoaderSimple().load_checkpoint
-    ckpt = ckpt_load_checkpoint(ckpt_name="juggernautXL_v8Rundiffusion.safetensors")
+    ckpt = ckpt_load_checkpoint(
+        ckpt_name="juggernautXL_version6Rundiffusion.safetensors"
+    )
 
 cliptextencode = CLIPTextEncode().encode
 emptylatentimage_generate = EmptyLatentImage().generate
@@ -72,6 +75,7 @@ def predict(
     cfg: float,
     denoise: float,
 ):
+    seed = seed if seed != -1 else np.random.randint(0, 2**63 - 1)
     try:
         with torch.inference_mode():
             cliptextencode_prompt = cliptextencode(
@@ -139,7 +143,7 @@ def predict(
                 )
 
                 rgb_img = tensor_to_pil(vaedecode_sample[0])
-                return flatten([rgb_img])
+                return flatten([rgb_img]), seed
             else:
                 layereddiffusionapply_sample = ld_fg_apply_layered_diffusion(
                     config="SDXL, Conv Injection", weight=1, model=ckpt[0]
@@ -177,28 +181,37 @@ def predict(
                 mask = tensor_to_pil(mask[0])
                 rgb_img = tensor_to_pil(vaedecode_sample[0])
 
-                return flatten([rgba_img, mask])
+                return flatten([rgba_img, mask]), seed
                 # return flatten([rgba_img, mask, rgb_img, ld_image])
     except Exception as e:
         raise gr.Error(e)
 
 
-examples = [["An old men sit on a chair looking at the sky"]]
-
-
 def flatten(l: List[List[any]]) -> List[any]:
     return [item for sublist in l for item in sublist]
 
 
-def predict_examples(prompt, negative_prompt):
+def predict_examples(
+    prompt, negative_prompt, input_image=None, remove_bg=False, cond_mode=None
+):
     return predict(
-        prompt, negative_prompt, None, False, None, 0, "euler", "normal", 20, 8.0, 1.0
+        prompt,
+        negative_prompt,
+        input_image,
+        remove_bg,
+        cond_mode,
+        0,
+        "euler",
+        "normal",
+        20,
+        8.0,
+        1.0,
     )
 
 
 css = """
 .gradio-container{
-    max-width: 50rem;
+    max-width: 85rem !important;
 }
 """
 with gr.Blocks(css=css) as blocks:
@@ -223,31 +236,38 @@ with gr.Blocks(css=css) as blocks:
                         label="Remove Background",
                         value=False,
                     )
-                    input_image = gr.Image(label="Input Image", type="pil")
+                    input_image = gr.Image(
+                        label="Input Image",
+                        type="pil",
+                    )
             with gr.Accordion(open=False, label="Advanced Options"):
-                seed = gr.Slider(
-                    label="Seed",
-                    value=0,
-                    minimum=-1,
-                    maximum=0xFFFFFFFFFFFFFFFF,
-                    step=1,
-                    randomize=True,
-                )
+                with gr.Group():
+                    with gr.Row():
+                        seed = gr.Slider(
+                            label="Seed",
+                            value=-1,
+                            minimum=-1,
+                            maximum=0xFFFFFFFFFFFFFFFF,
+                            step=1,
+                        )
+                        curr_seed = gr.Number(
+                            value=-1, interactive=False, scale=0, label=" "
+                        )
                 sampler_name = gr.Dropdown(
                     choices=samplers.KSampler.SAMPLERS,
                     label="Sampler Name",
-                    value=samplers.KSampler.SAMPLERS[0],
+                    value="dpmpp_2m_sde",
                 )
                 scheduler = gr.Dropdown(
                     choices=samplers.KSampler.SCHEDULERS,
                     label="Scheduler",
-                    value=samplers.KSampler.SCHEDULERS[0],
+                    value="karras",
                 )
                 steps = gr.Slider(
-                    label="Steps", value=20, minimum=1, maximum=30, step=1
+                    label="Steps", value=20, minimum=1, maximum=50, step=1
                 )
                 cfg = gr.Number(
-                    label="CFG", value=8.0, minimum=0.0, maximum=100.0, step=0.1
+                    label="CFG", value=5.0, minimum=0.0, maximum=100.0, step=0.1
                 )
                 denoise = gr.Number(
                     label="Denoise", value=1.0, minimum=0.0, maximum=1.0, step=0.01
@@ -269,12 +289,12 @@ with gr.Blocks(css=css) as blocks:
         cfg,
         denoise,
     ]
-    outputs = [gallery]
+    outputs = [gallery, curr_seed]
 
     gr.Examples(
         fn=predict_examples,
         examples=examples,
-        inputs=[prompt, negative_prompt],
+        inputs=[prompt, negative_prompt, input_image, remove_bg, cond_mode],
         outputs=outputs,
         cache_examples=False,
     )

utils.py

@@ -20,25 +20,26 @@ def tensor_to_pil(images: torch.Tensor | List[torch.Tensor]) -> List[Image.Image
     return imgs
 
 
-def pad_image(input_image):
-    pad_w, pad_h = (
-        np.max(((2, 2), np.ceil(np.array(input_image.size) / 64).astype(int)), axis=0)
-        * 64
-        - input_image.size
-    )
-    im_padded = Image.fromarray(
-        np.pad(np.array(input_image), ((0, pad_h), (0, pad_w), (0, 0)), mode="edge")
-    )
-    w, h = im_padded.size
-    if w == h:
+def pad_image(input_image, background_color=(0, 0, 0)):
+    w, h = input_image.size
+    pad_w = (64 - w % 64) % 64
+    pad_h = (64 - h % 64) % 64
+
+    new_size = (w + pad_w, h + pad_h)
+    im_padded = Image.new(input_image.mode, new_size, background_color)
+    im_padded.paste(input_image, (pad_w // 2, pad_h // 2))
+
+    if im_padded.size[0] == im_padded.size[1]:
         return im_padded
-    elif w > h:
-        new_image = Image.new(im_padded.mode, (w, w), (0, 0, 0))
-        new_image.paste(im_padded, (0, (w - h) // 2))
+    elif im_padded.size[0] > im_padded.size[1]:
+        new_size = (im_padded.size[0], im_padded.size[0])
+        new_image = Image.new(im_padded.mode, new_size, background_color)
+        new_image.paste(im_padded, (0, (new_size[1] - im_padded.size[1]) // 2))
         return new_image
     else:
-        new_image = Image.new(im_padded.mode, (h, h), (0, 0, 0))
-        new_image.paste(im_padded, ((h - w) // 2, 0))
+        new_size = (im_padded.size[1], im_padded.size[1])
+        new_image = Image.new(im_padded.mode, new_size, background_color)
+        new_image.paste(im_padded, ((new_size[0] - im_padded.size[0]) // 2, 0))
         return new_image
 
 
@@ -95,10 +96,14 @@ def postprocess_image(result: torch.Tensor, im_size: list) -> np.ndarray:
 
 def downloadModels():
     MODEL_PATH = hf_hub_download(
-        repo_id="lllyasviel/fav_models",
-        subfolder="fav",
-        filename="juggernautXL_v8Rundiffusion.safetensors",
+        repo_id="RunDiffusion/Juggernaut-XL-v6",
+        filename="juggernautXL_version6Rundiffusion.safetensors",
     )
+    # MODEL_PATH = hf_hub_download(
+    #     repo_id="lllyasviel/fav_models",
+    #     subfolder="fav",
+    #     filename="juggernautXL_v8Rundiffusion.safetensors",
+    # )
     LAYERS_PATH = snapshot_download(
         repo_id="LayerDiffusion/layerdiffusion-v1", allow_patterns="*.safetensors"
     )
@@ -112,3 +117,35 @@ def downloadModels():
     )
     if not model_target_path.exists():
         os.symlink(MODEL_PATH, model_target_path)
+
+
+examples = [
+    [
+        "An old men sit on a chair looking at the sky",
+        "ugly distorted image, low quality, text, bad, not good ,watermark",
+        None,
+        False,
+        None,
+    ],
+    [
+        "A beautiful toucan bird flying in the sky",
+        "ugly distorted image, low quality, text, bad, not good ,watermark",
+        "./examples/bg.png",
+        False,
+        "SDXL, Background",
+    ],
+    [
+        "A men watching a concert",
+        "ugly distorted image, low quality, text, bad, not good ,watermark",
+        "./examples/lecun.png",
+        True,
+        "SDXL, Foreground",
+    ],
+    [
+        "A men watching a concert",
+        "ugly distorted image, low quality, text, bad, not good ,watermark",
+        "./examples/julien.png",
+        True,
+        "SDXL, Foreground",
+    ],
+]