App

app.py

@@ -0,0 +1,218 @@
+from contextlib import nullcontext
+import torch
+from torch import autocast
+from diffusers import StableDiffusionPipeline
+import gradio as gr
+
+
+CHECKPOINTS = [
+    "epoch-000025",
+    #"epoch-000081"
+]
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+context = autocast if device == "cuda" else nullcontext
+dtype = torch.float16 if device == "cuda" else torch.float32
+
+def load_pipe(checkpoint):
+
+    pipe = StableDiffusionPipeline.from_pretrained("Gazoche/sd-gundam-diffusers", revision=checkpoint, torch_dtype=dtype)
+    pipe = pipe.to(device)
+
+    # Disabling the NSFW filter as it's getting confused by the generated images
+    def null_safety(images, **kwargs):
+        return images, False
+    pipe.safety_checker = null_safety
+
+    return pipe
+
+pipes = {
+    checkpoint: load_pipe(checkpoint)
+    for checkpoint in CHECKPOINTS
+}
+
+def infer(prompt, n_samples, steps, scale, model):
+
+    checkpoint = "epoch-000025" if model == "normal" else "epoch-000081"
+
+    in_prompt = ""
+    guidance_scale = 0.0
+    if prompt is not None:
+        in_prompt = prompt
+        guidance_scale = scale
+
+    with context("cuda"):
+        images = pipes[checkpoint](
+            n_samples * [in_prompt],
+            guidance_scale=guidance_scale,
+            num_inference_steps=steps
+        ).images
+
+    return images
+
+
+def infer_random(n_samples, steps, scale, model):
+    return infer(None, n_samples, steps, scale, model)
+
+css = """
+        a {
+            color: inherit;
+            text-decoration: underline;
+        }
+        .gradio-container {
+            font-family: 'IBM Plex Sans', sans-serif;
+        }
+        .gr-button {
+            color: white;
+            border-color: #9d66e5;
+            background: #9d66e5;
+        }
+        input[type='range'] {
+            accent-color: #9d66e5;
+        }
+        .dark input[type='range'] {
+            accent-color: #dfdfdf;
+        }
+        .container {
+            max-width: 730px;
+            margin: auto;
+            padding-top: 1.5rem;
+        }
+        #gallery {
+            min-height: 22rem;
+            margin-bottom: 15px;
+            margin-left: auto;
+            margin-right: auto;
+            border-bottom-right-radius: .5rem !important;
+            border-bottom-left-radius: .5rem !important;
+        }
+        #gallery>div>.h-full {
+            min-height: 20rem;
+        }
+        .details:hover {
+            text-decoration: underline;
+        }
+        .gr-button {
+            white-space: nowrap;
+        }
+        .gr-button:focus {
+            border-color: rgb(147 197 253 / var(--tw-border-opacity));
+            outline: none;
+            box-shadow: var(--tw-ring-offset-shadow), var(--tw-ring-shadow), var(--tw-shadow, 0 0 #0000);
+            --tw-border-opacity: 1;
+            --tw-ring-offset-shadow: var(--tw-ring-inset) 0 0 0 var(--tw-ring-offset-width) var(--tw-ring-offset-color);
+            --tw-ring-shadow: var(--tw-ring-inset) 0 0 0 calc(3px var(--tw-ring-offset-width)) var(--tw-ring-color);
+            --tw-ring-color: rgb(191 219 254 / var(--tw-ring-opacity));
+            --tw-ring-opacity: .5;
+        }
+        #advanced-options {
+            margin-bottom: 20px;
+        }
+        .footer {
+            margin-bottom: 45px;
+            margin-top: 35px;
+            text-align: center;
+            border-bottom: 1px solid #e5e5e5;
+        }
+        .footer>p {
+            font-size: .8rem;
+            display: inline-block;
+            padding: 0 10px;
+            transform: translateY(10px);
+            background: white;
+        }
+        .dark .logo{ filter: invert(1); }
+        .dark .footer {
+            border-color: #303030;
+        }
+        .dark .footer>p {
+            background: #0b0f19;
+        }
+        .acknowledgments h4{
+            margin: 1.25em 0 .25em 0;
+            font-weight: bold;
+            font-size: 115%;
+        }
+"""
+
+block = gr.Blocks(css=css)
+
+with block:
+    gr.HTML(
+        """
+            <div style="text-align: center; max-width: 650px; margin: 0 auto;">
+              <div>
+                <h1 style="font-weight: 900; font-size: 3rem;">
+                  Gundam text to image
+                </h1>
+              </div>
+              <p style="margin-bottom: 10px; font-size: 94%">
+              From a text description, generate a mecha from the anime franchise Mobile Suit Gundam
+              </p>
+              <p style="margin-bottom: 10px; font-size: 94%">
+              Github: <a href="https://github.com/Askannz/gundam-stable-diffusion">https://github.com/Askannz/gundam-stable-diffusion</a>
+              </p>
+               <ul>
+                <li>More steps generally means less visual noise but fewer details</li>
+                <li>Text guidance controls how much the prompt influences the generation</li>
+                <li>The overfitted model gives cleaner but less original results</li>
+              </ul>
+            </div>
+        """
+    )
+    with gr.Group():
+
+        with gr.Box():
+            with gr.Row().style(mobile_collapse=False, equal_height=True):
+                text = gr.Textbox(
+                    label="Enter your prompt",
+                    show_label=False,
+                    max_lines=1,
+                    placeholder="Enter your prompt",
+                ).style(
+                    border=(True, False, True, True),
+                    rounded=(True, False, False, True),
+                    container=False,
+                )
+                btn = gr.Button("Generate from prompt").style(
+                    margin=False,
+                    rounded=(False, True, True, False),
+                )
+
+        with gr.Box():
+            with gr.Row().style(mobile_collapse=False, equal_height=True):
+                btn_rand = gr.Button("Random").style(
+                    margin=False,
+                    rounded=(False, True, True, False),
+                )
+
+        gallery = gr.Gallery(
+            label="Generated images", show_label=False, elem_id="gallery"
+        ).style(grid=[2], height="auto")
+
+
+        with gr.Row(elem_id="advanced-options"):
+            samples = gr.Slider(label="Images", minimum=1, maximum=4, value=1, step=1)
+            steps = gr.Slider(label="Steps", minimum=5, maximum=50, value=25, step=5)
+            scale = gr.Slider(
+                label="Text Guidance Scale", minimum=0, maximum=50, value=7.5, step=0.1
+            )
+
+        with gr.Row(elem_id="checkpoint"):
+            #model = gr.Radio(label="Model", choices=["normal", "overfitted"], value="normal")
+            model = gr.Radio(label="Model", choices=["normal"], value="normal")
+
+
+        text.submit(infer, inputs=[text, samples, steps, scale, model], outputs=gallery)
+        btn.click(infer, inputs=[text, samples, steps, scale, model], outputs=gallery)
+        btn_rand.click(infer_random, inputs=[samples, steps, scale, model], outputs=gallery)
+        gr.HTML(
+            """
+                <div class="footer">
+                    <p> Gradio Demo by 🤗 Hugging Face and Gazoche
+                    </p>
+                </div>
+           """
+        )
+
+block.launch()

requirements.txt

@@ -0,0 +1,9 @@
+diffusers==0.3.0
+transformers
+scipy
+ftfy
+gradio
+datasets
+fastapi
+uvicorn
+requests