File size: 5,185 Bytes
284f445
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import spaces
import gradio as gr
import numpy as np
import PIL.Image
from PIL import Image
import random
from diffusers import StableDiffusionXLPipeline
from diffusers import EulerAncestralDiscreteScheduler
import torch
import gradio_client.utils

# もっと徹底的なモンキーパッチ
old_get_type = gradio_client.utils.get_type
def new_get_type(schema):
    if isinstance(schema, bool):
        return "bool"
    return old_get_type(schema)

gradio_client.utils.get_type = new_get_type

# _json_schema_to_python_typeの修正も追加
old_json_schema_to_python_type = gradio_client.utils._json_schema_to_python_type
def new_json_schema_to_python_type(schema, defs=None):
    if isinstance(schema, bool):
        return "bool"
    try:
        return old_json_schema_to_python_type(schema, defs)
    except Exception as e:
        # エラーが発生した場合は汎用的な型を返す
        return "any"

gradio_client.utils._json_schema_to_python_type = new_json_schema_to_python_type

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Make sure to use torch.float16 consistently throughout the pipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
    "votepurchase/novaAnimeXL_ilV60",
    torch_dtype=torch.float16,
    variant="fp16",  # Explicitly use fp16 variant
    use_safetensors=True  # Use safetensors if available
)

pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
pipe.to(device)

# Force all components to use the same dtype
pipe.text_encoder.to(torch.float16)
pipe.text_encoder_2.to(torch.float16)
pipe.vae.to(torch.float16)
pipe.unet.to(torch.float16)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1216
    
@spaces.GPU
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
    # Check and truncate prompt if too long (CLIP can only handle 77 tokens)
    if len(prompt.split()) > 60:  # Rough estimate to avoid exceeding token limit
        print("Warning: Prompt may be too long and will be truncated by the model")
        
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    generator = torch.Generator(device=device).manual_seed(seed)
    
    try:
        output_image = pipe(
            prompt=prompt,
            negative_prompt=negative_prompt,
            guidance_scale=guidance_scale,
            num_inference_steps=num_inference_steps,
            width=width,
            height=height,
            generator=generator
        ).images[0]
        
        return output_image
    except RuntimeError as e:
        print(f"Error during generation: {e}")
        # Return a blank image with error message
        error_img = Image.new('RGB', (width, height), color=(0, 0, 0))
        return error_img


css = """
#col-container {
    margin: 0 auto;
    max-width: 520px;
}
"""

with gr.Blocks(css=css) as demo:

    with gr.Column(elem_id="col-container"):

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt (keep it under 60 words for best results)",
                container=False,
            )

            run_button = gr.Button("Run", scale=0)

        result = gr.Image(label="Result", show_label=False)
        
        with gr.Accordion("Advanced Settings", open=False):

            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
                value="nsfw, (low quality, worst quality:1.2), very displeasing, 3d, watermark, signature, ugly, poorly drawn"
            )

            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
            )

            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)

            with gr.Row():
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                )

                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                )

            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=20.0,
                    step=0.1,
                    value=7,
                )

                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=28,
                    step=1,
                    value=28,
                )

    run_button.click(
        fn=infer,
        inputs=[prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
        outputs=[result]
    )

demo.queue().launch()