Spaces:
Running
on
Zero
Running
on
Zero
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,132 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import numpy as np
|
3 |
+
import spaces
|
4 |
+
import torch
|
5 |
+
import random
|
6 |
+
from PIL import Image
|
7 |
+
|
8 |
+
# Import the pipeline from diffusers
|
9 |
+
from diffusers import FluxKontextPipeline
|
10 |
+
|
11 |
+
# --- Constants and Model Loading ---
|
12 |
+
MAX_SEED = np.iinfo(np.int32).max
|
13 |
+
|
14 |
+
# Load the pretrained model
|
15 |
+
# Note: This requires a CUDA-enabled GPU. Error handling is added for environments without it.
|
16 |
+
try:
|
17 |
+
pipe = FluxKontextPipeline.from_pretrained(
|
18 |
+
"black-forest-labs/FLUX.1-Kontext-dev",
|
19 |
+
torch_dtype=torch.bfloat16
|
20 |
+
).to("cuda")
|
21 |
+
except Exception as e:
|
22 |
+
pipe = None
|
23 |
+
print(f"Warning: Could not load the model on CUDA. GPU is required. Error: {e}")
|
24 |
+
|
25 |
+
# --- Core Inference Function for ChatInterface ---
|
26 |
+
|
27 |
+
@spaces.GPU
|
28 |
+
def chat_fn(message, chat_history, seed, randomize_seed, guidance_scale, steps, progress=gr.Progress(track_tqdm=True)):
|
29 |
+
"""
|
30 |
+
Performs image generation or editing based on user input from the chat interface.
|
31 |
+
|
32 |
+
Args:
|
33 |
+
message (dict): A dictionary from gr.MultimodalTextbox, containing:
|
34 |
+
- "text" (str): The user's text prompt.
|
35 |
+
- "files" (list): A list of paths to uploaded files.
|
36 |
+
chat_history (list): The history of the conversation (managed by ChatInterface).
|
37 |
+
seed (int): The random seed for generation.
|
38 |
+
randomize_seed (bool): If True, a random seed is used.
|
39 |
+
guidance_scale (float): Controls adherence to the prompt.
|
40 |
+
steps (int): Number of inference steps.
|
41 |
+
progress (gr.Progress): Gradio progress tracker.
|
42 |
+
|
43 |
+
Returns:
|
44 |
+
PIL.Image.Image: The generated or edited image to be displayed in the chat.
|
45 |
+
"""
|
46 |
+
if pipe is None:
|
47 |
+
raise gr.Error("Model could not be loaded. A CUDA-enabled GPU is required to run this application.")
|
48 |
+
|
49 |
+
prompt = message["text"]
|
50 |
+
files = message["files"]
|
51 |
+
|
52 |
+
# Input validation
|
53 |
+
if not prompt and not files:
|
54 |
+
raise gr.Error("Please provide a prompt and/or upload an image.")
|
55 |
+
|
56 |
+
if randomize_seed:
|
57 |
+
seed = random.randint(0, MAX_SEED)
|
58 |
+
|
59 |
+
# Set up a PyTorch generator for reproducible results
|
60 |
+
generator = torch.Generator(device="cuda").manual_seed(seed)
|
61 |
+
|
62 |
+
input_image = None
|
63 |
+
if files:
|
64 |
+
# User has uploaded an image for editing (image-to-image)
|
65 |
+
print(f"Received image: {files[0]}")
|
66 |
+
input_image = Image.open(files[0]).convert("RGB")
|
67 |
+
image = pipe(
|
68 |
+
image=input_image,
|
69 |
+
prompt=prompt,
|
70 |
+
guidance_scale=guidance_scale,
|
71 |
+
num_inference_steps=steps,
|
72 |
+
generator=generator,
|
73 |
+
).images[0]
|
74 |
+
else:
|
75 |
+
# No image uploaded, perform text-to-image generation
|
76 |
+
print(f"Received prompt for text-to-image: {prompt}")
|
77 |
+
image = pipe(
|
78 |
+
prompt=prompt,
|
79 |
+
guidance_scale=guidance_scale,
|
80 |
+
num_inference_steps=steps,
|
81 |
+
generator=generator,
|
82 |
+
).images[0]
|
83 |
+
|
84 |
+
# To also inform the user of the seed, you could optionally return a tuple,
|
85 |
+
# but for a clean image output, we just return the image.
|
86 |
+
# For example: return (image, f"Seed: {seed}")
|
87 |
+
return image
|
88 |
+
|
89 |
+
# --- UI Definition using gr.ChatInterface ---
|
90 |
+
|
91 |
+
# Define the components for "Advanced Settings" that will be passed to `additional_inputs`
|
92 |
+
seed_slider = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=42)
|
93 |
+
randomize_checkbox = gr.Checkbox(label="Randomize seed", value=False)
|
94 |
+
guidance_slider = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=10.0, step=0.1, value=2.5)
|
95 |
+
steps_slider = gr.Slider(label="Steps", minimum=1, maximum=30, value=28, step=1)
|
96 |
+
|
97 |
+
# Create the ChatInterface
|
98 |
+
demo = gr.ChatInterface(
|
99 |
+
fn=chat_fn,
|
100 |
+
title="FLUX.1 Kontext [dev]",
|
101 |
+
description="""<p style='text-align: center;'>
|
102 |
+
A simple chat UI for the <b>FLUX.1 Kontext</b> model.
|
103 |
+
<br>
|
104 |
+
To edit an image, upload it and type your instructions (e.g., "Add a hat").
|
105 |
+
<br>
|
106 |
+
To generate an image, just type a prompt (e.g., "A photo of an astronaut on a horse").
|
107 |
+
<br>
|
108 |
+
Find the model on <a href='https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev' target='_blank'>Hugging Face</a>.
|
109 |
+
</p>""",
|
110 |
+
# Use a multimodal textbox to allow both text and image uploads
|
111 |
+
textbox=gr.MultimodalTextbox(
|
112 |
+
file_types=["image"],
|
113 |
+
placeholder="Type a prompt and/or upload an image...",
|
114 |
+
render=False # Important: Let ChatInterface render the textbox
|
115 |
+
),
|
116 |
+
additional_inputs=[
|
117 |
+
seed_slider,
|
118 |
+
randomize_checkbox,
|
119 |
+
guidance_slider,
|
120 |
+
steps_slider
|
121 |
+
],
|
122 |
+
examples=[
|
123 |
+
{"text": "A cute robot reading a book", "files": []},
|
124 |
+
{"text": "change his shirt to a hawaiian shirt", "files": ["https://gradio-builds.s3.amazonaws.com/demo-files/chewbacca.png"]},
|
125 |
+
{"text": "make it a wooden house", "files": ["https://gradio-builds.s3.amazonaws.com/demo-files/house.png"]},
|
126 |
+
],
|
127 |
+
theme="soft"
|
128 |
+
)
|
129 |
+
|
130 |
+
# Launch the application
|
131 |
+
if __name__ == "__main__":
|
132 |
+
demo.launch()
|