Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
import gradio as gr
|
2 |
-
import
|
3 |
-
import
|
4 |
import random
|
5 |
import os
|
6 |
from PIL import Image
|
@@ -8,11 +8,12 @@ from deep_translator import GoogleTranslator
|
|
8 |
import json
|
9 |
from langdetect import detect
|
10 |
|
11 |
-
|
12 |
-
|
13 |
-
|
|
|
14 |
|
15 |
-
def query(prompt, is_negative=False, steps=30, cfg_scale=7,
|
16 |
if prompt == "" or prompt == None:
|
17 |
return None
|
18 |
|
@@ -52,8 +53,6 @@ def query(prompt, is_negative=False, steps=30, cfg_scale=7, sampler="DPM++ 2M Ka
|
|
52 |
else:
|
53 |
print(f"Error: {response.status_code} - {response.text}")
|
54 |
|
55 |
-
API_TOKEN = random.choice([os.getenv("HF_READ_TOKEN"), os.getenv("HF_READ_TOKEN_2"), os.getenv("HF_READ_TOKEN_3"), os.getenv("HF_READ_TOKEN_4"), os.getenv("HF_READ_TOKEN_5")])
|
56 |
-
headers = {"Authorization": f"Bearer {API_TOKEN}"}
|
57 |
language = detect(prompt)
|
58 |
|
59 |
if language != 'en':
|
@@ -63,37 +62,13 @@ def query(prompt, is_negative=False, steps=30, cfg_scale=7, sampler="DPM++ 2M Ka
|
|
63 |
prompt = f"{prompt} | ultra detail, ultra elaboration, ultra quality, perfect."
|
64 |
print(f'\033[1mГенерация {key}:\033[0m {prompt}')
|
65 |
|
66 |
-
|
|
|
67 |
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
"cfg_scale": cfg_scale,
|
73 |
-
"seed": seed if seed != -1 else random.randint(1, 1000000000),
|
74 |
-
"guidance_scale": cfg_scale,
|
75 |
-
"num_inference_steps": steps,
|
76 |
-
"negative_prompt": is_negative
|
77 |
-
}
|
78 |
-
|
79 |
-
response = requests.post(f"{api_base}{API_URL}", headers=headers, json=payload, timeout=timeout)
|
80 |
-
if response.status_code != 200:
|
81 |
-
print(f"Ошибка: Не удалось получить изображение. Статус ответа: {response.status_code}")
|
82 |
-
print(f"Содержимое ответа: {response.text}")
|
83 |
-
if response.status_code == 503:
|
84 |
-
raise gr.Error(f"{response.status_code} : The model is being loaded")
|
85 |
-
return None
|
86 |
-
raise gr.Error(f"{response.status_code}")
|
87 |
-
return None
|
88 |
-
|
89 |
-
try:
|
90 |
-
image_bytes = response.content
|
91 |
-
image = Image.open(io.BytesIO(image_bytes))
|
92 |
-
print(f'\033[1mГенерация {key} завершена!\033[0m ({prompt})')
|
93 |
-
return image
|
94 |
-
except Exception as e:
|
95 |
-
print(f"Ошибка при попытке открыть изображение: {e}")
|
96 |
-
return None
|
97 |
|
98 |
css = """
|
99 |
* {}
|
@@ -113,8 +88,6 @@ with gr.Blocks(css=css) as dalle:
|
|
113 |
steps = gr.Slider(label="Sampling steps", value=35, minimum=1, maximum=70, step=1)
|
114 |
with gr.Row():
|
115 |
cfg = gr.Slider(label="CFG Scale", value=7, minimum=1, maximum=20, step=0.1)
|
116 |
-
with gr.Row():
|
117 |
-
method = gr.Radio(label="Sampling method", value="Euler a", choices=["DPM++ 2M Karras", "DPM++ SDE Karras", "Euler", "Euler a", "Heun", "DDIM"])
|
118 |
with gr.Row():
|
119 |
seed = gr.Slider(label="Seed", value=-1, minimum=-1, maximum=1000000000, step=1)
|
120 |
with gr.Row():
|
@@ -131,6 +104,6 @@ with gr.Blocks(css=css) as dalle:
|
|
131 |
with gr.Row():
|
132 |
image_output = gr.Image(type="pil", label="Изображение", elem_id="gallery")
|
133 |
|
134 |
-
text_button.click(query, inputs=[text_prompt, negative_prompt, steps, cfg,
|
135 |
|
136 |
-
dalle.queue(max_size=
|
|
|
1 |
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from diffusers import DiffusionPipeline, EulerDiscreteScheduler
|
4 |
import random
|
5 |
import os
|
6 |
from PIL import Image
|
|
|
8 |
import json
|
9 |
from langdetect import detect
|
10 |
|
11 |
+
model_id = "cagliostrolab/animagine-xl-3.1"
|
12 |
+
pipe = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16, revision="main")
|
13 |
+
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config)
|
14 |
+
pipe = pipe.to("cuda" if torch.cuda.is_available() else "cpu")
|
15 |
|
16 |
+
def query(prompt, is_negative=False, steps=30, cfg_scale=7, seed=-1, gpt=False):
|
17 |
if prompt == "" or prompt == None:
|
18 |
return None
|
19 |
|
|
|
53 |
else:
|
54 |
print(f"Error: {response.status_code} - {response.text}")
|
55 |
|
|
|
|
|
56 |
language = detect(prompt)
|
57 |
|
58 |
if language != 'en':
|
|
|
62 |
prompt = f"{prompt} | ultra detail, ultra elaboration, ultra quality, perfect."
|
63 |
print(f'\033[1mГенерация {key}:\033[0m {prompt}')
|
64 |
|
65 |
+
if seed == -1:
|
66 |
+
seed = random.randint(1, 1000000000)
|
67 |
|
68 |
+
generator = torch.Generator("cuda" if torch.cuda.is_available() else "cpu").manual_seed(seed)
|
69 |
+
image = pipe(prompt, negative_prompt=is_negative, guidance_scale=cfg_scale, num_inference_steps=steps, generator=generator).images[0]
|
70 |
+
print(f'\033[1mГенерация {key} завершена!\033[0m ({prompt})')
|
71 |
+
return image
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
|
73 |
css = """
|
74 |
* {}
|
|
|
88 |
steps = gr.Slider(label="Sampling steps", value=35, minimum=1, maximum=70, step=1)
|
89 |
with gr.Row():
|
90 |
cfg = gr.Slider(label="CFG Scale", value=7, minimum=1, maximum=20, step=0.1)
|
|
|
|
|
91 |
with gr.Row():
|
92 |
seed = gr.Slider(label="Seed", value=-1, minimum=-1, maximum=1000000000, step=1)
|
93 |
with gr.Row():
|
|
|
104 |
with gr.Row():
|
105 |
image_output = gr.Image(type="pil", label="Изображение", elem_id="gallery")
|
106 |
|
107 |
+
text_button.click(query, inputs=[text_prompt, negative_prompt, steps, cfg, seed, gpt], outputs=image_output)
|
108 |
|
109 |
+
dalle.queue(max_size=100).launch(show_api=False, share=False)
|