Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,46 +1,65 @@
|
|
1 |
-
# app.py
|
2 |
-
|
3 |
-
import os
|
4 |
import gradio as gr
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
-
def
|
8 |
with gr.Blocks() as demo:
|
9 |
-
gr.
|
10 |
-
|
11 |
-
with gr.Tab("Fine-tune Model"):
|
12 |
-
with gr.Row():
|
13 |
-
instance_images = gr.File(label="Instance Images", file_count="multiple")
|
14 |
-
class_images = gr.File(label="Class Images", file_count="multiple")
|
15 |
-
with gr.Row():
|
16 |
-
instance_prompt = gr.Textbox(label="Instance Prompt")
|
17 |
-
class_prompt = gr.Textbox(label="Class Prompt")
|
18 |
with gr.Row():
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
21 |
|
22 |
with gr.Tab("Generate Images"):
|
23 |
with gr.Row():
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
fine_tune_button.click(fine_tune_model, inputs=[instance_images, class_images, instance_prompt, class_prompt, num_train_steps], outputs=huggingface_link)
|
41 |
-
|
42 |
-
generate_button.click(generate_images, inputs=[prompt, negative_prompt, num_samples, height, width, num_inference_steps, guidance_scale], outputs=output_images)
|
43 |
-
|
44 |
-
push_button.click(push_to_huggingface, inputs=[HfFolder.path, repo_name], outputs=huggingface_link)
|
45 |
-
|
46 |
-
demo.launch()
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
+
import os
|
3 |
+
import shutil
|
4 |
+
from pathlib import Path
|
5 |
+
from main import fine_tune_model
|
6 |
+
from diffusers import StableDiffusionPipeline, DDIMScheduler
|
7 |
+
import torch
|
8 |
+
|
9 |
+
MODEL_NAME = "runwayml/stable-diffusion-v1-5"
|
10 |
+
OUTPUT_DIR = "/content/stable_diffusion_weights/custom_model"
|
11 |
+
|
12 |
+
def fine_tune(instance_prompt, images):
|
13 |
+
instance_data_dir = "/content/instance_images"
|
14 |
+
if os.path.exists(instance_data_dir):
|
15 |
+
shutil.rmtree(instance_data_dir)
|
16 |
+
os.makedirs(instance_data_dir, exist_ok=True)
|
17 |
+
|
18 |
+
for i, img in enumerate(images):
|
19 |
+
img.save(os.path.join(instance_data_dir, f"instance_{i}.png"))
|
20 |
+
|
21 |
+
fine_tune_model(instance_data_dir, instance_prompt, MODEL_NAME, OUTPUT_DIR)
|
22 |
+
return "Model fine-tuning complete."
|
23 |
+
|
24 |
+
def generate_images(prompt, num_samples, height, width, num_inference_steps, guidance_scale):
|
25 |
+
pipe = StableDiffusionPipeline.from_pretrained(OUTPUT_DIR, safety_checker=None, torch_dtype=torch.float16).to("cuda")
|
26 |
+
pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config)
|
27 |
+
g_cuda = torch.Generator(device='cuda').manual_seed(1337)
|
28 |
+
|
29 |
+
with torch.autocast("cuda"), torch.inference_mode():
|
30 |
+
images = pipe(
|
31 |
+
prompt, height=height, width=width, num_images_per_prompt=num_samples,
|
32 |
+
num_inference_steps=num_inference_steps, guidance_scale=guidance_scale, generator=g_cuda
|
33 |
+
).images
|
34 |
+
|
35 |
+
return images
|
36 |
|
37 |
+
def gradio_app():
|
38 |
with gr.Blocks() as demo:
|
39 |
+
with gr.Tab("Fine-Tune Model"):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
with gr.Row():
|
41 |
+
with gr.Column():
|
42 |
+
instance_prompt = gr.Textbox(label="Instance Prompt")
|
43 |
+
image_input = gr.Image(label="Upload Images", source="upload", tool="editor", type="pil", multiple=True)
|
44 |
+
fine_tune_button = gr.Button("Fine-Tune Model")
|
45 |
+
output_text = gr.Textbox(label="Output")
|
46 |
+
fine_tune_button.click(fine_tune, inputs=[instance_prompt, image_input], outputs=output_text)
|
47 |
|
48 |
with gr.Tab("Generate Images"):
|
49 |
with gr.Row():
|
50 |
+
with gr.Column():
|
51 |
+
prompt = gr.Textbox(label="Prompt")
|
52 |
+
num_samples = gr.Number(label="Number of Samples", value=1)
|
53 |
+
guidance_scale = gr.Number(label="Guidance Scale", value=7.5)
|
54 |
+
height = gr.Number(label="Height", value=512)
|
55 |
+
width = gr.Number(label="Width", value=512)
|
56 |
+
num_inference_steps = gr.Slider(label="Steps", value=50, minimum=1, maximum=100)
|
57 |
+
generate_button = gr.Button("Generate Images")
|
58 |
+
with gr.Column():
|
59 |
+
gallery = gr.Gallery(label="Generated Images")
|
60 |
+
generate_button.click(generate_images, inputs=[prompt, num_samples, height, width, num_inference_steps, guidance_scale], outputs=gallery)
|
61 |
|
62 |
+
demo.launch()
|
63 |
+
|
64 |
+
if __name__ == "__main__":
|
65 |
+
gradio_app()
|
|
|
|
|
|
|
|
|
|
|
|
|
|