File size: 8,579 Bytes
2292ec1
93977a4
2292ec1
 
 
1bb3dc0
2292ec1
93977a4
1bb3dc0
93977a4
2292ec1
93977a4
 
1bb3dc0
93977a4
 
2292ec1
93977a4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2292ec1
 
 
93977a4
 
 
 
 
 
 
 
 
 
 
 
 
2292ec1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
93977a4
 
 
2292ec1
 
 
 
 
 
 
 
 
93977a4
2292ec1
93977a4
 
 
 
2292ec1
 
 
 
 
93977a4
 
2292ec1
 
 
 
 
 
 
 
93977a4
 
2292ec1
 
 
 
 
 
93977a4
 
 
2292ec1
 
93977a4
2292ec1
93977a4
 
 
2292ec1
 
 
93977a4
 
2292ec1
93977a4
 
2292ec1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
93977a4
 
2292ec1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
93977a4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
import torch
from diffusers import Flux2Pipeline
from huggingface_hub import get_token
import requests
import io
import gradio as gr
from PIL import Image
import os

# Configuration
repo_id = "diffusers/FLUX.2-dev-bnb-4bit"
device = "cuda" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32

print(f"Using device: {device}")
print(f"Using dtype: {torch_dtype}")

def remote_text_encoder(prompts):
    """Encode prompts using remote text encoder API."""
    try:
        token = get_token()
        if not token:
            raise ValueError("HuggingFace token not found. Please login using 'huggingface-cli login'")
        
        response = requests.post(
            "https://remote-text-encoder-flux-2.huggingface.co/predict",
            json={"prompt": prompts},
            headers={
                "Authorization": f"Bearer {token}",
                "Content-Type": "application/json"
            },
            timeout=60
        )
        response.raise_for_status()
        prompt_embeds = torch.load(io.BytesIO(response.content))
        return prompt_embeds.to(device)
    except Exception as e:
        raise Exception(f"Failed to encode prompt: {str(e)}")

# Load the pipeline
print("Loading Flux2 pipeline...")
try:
    pipe = Flux2Pipeline.from_pretrained(
        repo_id,
        text_encoder=None,
        torch_dtype=torch_dtype,
        device_map="cuda"
    )
    if not torch.cuda.is_available():
        pipe = pipe.to(device)
    print("Pipeline loaded successfully!")
except Exception as e:
    print(f"Error loading pipeline: {e}")
    raise

def generate_image(
    prompt: str,
    input_image: Image.Image = None,
    num_inference_steps: int = 28,
    guidance_scale: float = 4.0,
    seed: int = 42,
    progress=gr.Progress()
):
    """
    Generate an image using Flux2 based on text prompt and optional input image.
    
    Args:
        prompt: Text description of the desired image
        input_image: Optional input image for image-to-image generation
        num_inference_steps: Number of denoising steps (higher = better quality but slower)
        guidance_scale: How closely to follow the prompt (higher = more strict)
        seed: Random seed for reproducibility (-1 for random)
    """
    if not prompt or prompt.strip() == "":
        raise gr.Error("Please enter a prompt!")
    
    if not torch.cuda.is_available():
        raise gr.Error("This Space requires a GPU to run. Please try a different Space or upgrade your hardware.")
    
    progress(0, desc="Encoding prompt...")
    
    try:
        # Get prompt embeddings from remote encoder
        prompt_embeds = remote_text_encoder(prompt)
        
        progress(0.3, desc="Generating image...")
        
        # Set up generator
        generator_device = "cuda" if torch.cuda.is_available() else "cpu"
        if seed == -1:
            import random
            seed = random.randint(0, 2**32 - 1)
        
        generator = torch.Generator(device=generator_device).manual_seed(int(seed))
        
        # Prepare pipeline arguments
        pipe_kwargs = {
            "prompt_embeds": prompt_embeds,
            "generator": generator,
            "num_inference_steps": int(num_inference_steps),
            "guidance_scale": float(guidance_scale),
        }
        
        # Add input image if provided
        if input_image is not None:
            pipe_kwargs["image"] = input_image
            progress(0.4, desc="Processing input image...")
        
        # Generate image
        with torch.inference_mode():
            image = pipe(**pipe_kwargs).images[0]
        
        progress(1.0, desc="Done!")
        
        return image
    
    except Exception as e:
        import traceback
        error_msg = f"Error generating image: {str(e)}\n{traceback.format_exc()}"
        print(error_msg)
        raise gr.Error(f"Error generating image: {str(e)}")


# Create Gradio interface
with gr.Blocks(
    title="Flux2 Image Generator",
) as demo:
    gr.Markdown(
        """
        # 🎨 Flux2 Image Generator
        Generate stunning images using **FLUX.2-dev** with 4-bit quantization for efficient inference.
        
        Supports both **text-to-image** and **image-to-image** generation.
        
        ⚠️ **Note**: This Space requires a GPU to run.
        """
    )
    
    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("### 📝 Input")
            
            prompt_input = gr.Textbox(
                label="Prompt",
                placeholder="Describe the image you want to generate...",
                lines=4,
                value="Realistic macro photograph of a hermit crab using a soda can as its shell, partially emerging from the can, captured with sharp detail and natural colors, on a sunlit beach with soft shadows and a shallow depth of field, with blurred ocean waves in the background."
            )
            
            image_input = gr.Image(
                label="Input Image (Optional)",
                type="pil",
                sources=["upload", "clipboard"],
                height=300
            )
            
            gr.Markdown("### ⚙️ Parameters")
            
            with gr.Row():
                num_steps = gr.Slider(
                    minimum=1,
                    maximum=100,
                    value=28,
                    step=1,
                    label="Inference Steps",
                    info="More steps = better quality but slower"
                )
                
                guidance = gr.Slider(
                    minimum=1.0,
                    maximum=15.0,
                    value=4.0,
                    step=0.5,
                    label="Guidance Scale",
                    info="How closely to follow the prompt"
                )
            
            seed_input = gr.Number(
                label="Seed",
                value=42,
                precision=0,
                info="Use -1 for random seed"
            )
            
            generate_btn = gr.Button(
                "🚀 Generate Image",
                variant="primary",
                size="lg",
                elem_classes="generate-btn"
            )
            
            gr.Markdown(
                """
                ### 💡 Tips
                - **Text-to-Image**: Just enter a prompt and click generate
                - **Image-to-Image**: Upload an image and describe the changes
                - Start with 28 steps for a good balance of quality and speed
                - Higher guidance scale follows your prompt more strictly
                - Use the same seed to reproduce results
                """
            )
        
        with gr.Column(scale=1):
            gr.Markdown("### 🖼️ Output")
            
            output_image = gr.Image(
                label="Generated Image",
                type="pil",
                height=600
            )
            
            gr.Markdown(
                """
                ### 📊 Examples
                Try these prompts for inspiration!
                """
            )
    
    # Examples
    gr.Examples(
        examples=[
            [
                "A serene landscape with mountains at sunset, vibrant orange and pink sky, reflected in a calm lake, photorealistic",
                None,
                28,
                4.0,
                42
            ],
            [
                "A futuristic cityscape at night, neon lights, flying cars, cyberpunk style, highly detailed",
                None,
                28,
                4.0,
                123
            ],
            [
                "A cute robot reading a book in a cozy library, warm lighting, digital art style",
                None,
                28,
                4.0,
                456
            ],
            [
                "Macro photography of a dew drop on a leaf, morning light, sharp focus, bokeh background",
                None,
                28,
                4.0,
                789
            ],
        ],
        inputs=[prompt_input, image_input, num_steps, guidance, seed_input],
        outputs=output_image,
        cache_examples=False,
    )
    
    # Connect the generate button
    generate_btn.click(
        fn=generate_image,
        inputs=[prompt_input, image_input, num_steps, guidance, seed_input],
        outputs=output_image,
    )

if __name__ == "__main__":
    demo.queue(max_size=20).launch()