Update app.py
Browse files
app.py
CHANGED
|
@@ -1,26 +1,37 @@
|
|
| 1 |
import os
|
| 2 |
-
import requests
|
| 3 |
import gradio as gr
|
| 4 |
from PIL import Image, ImageDraw, ImageFont
|
| 5 |
import io
|
| 6 |
-
import
|
| 7 |
-
from
|
| 8 |
|
| 9 |
# ===== CONFIGURATION =====
|
| 10 |
-
|
| 11 |
-
MODEL_NAME = "stabilityai/stable-diffusion-xl-base-1.0"
|
| 12 |
-
API_URL = f"https://api-inference.huggingface.co/models/{MODEL_NAME}"
|
| 13 |
-
headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
|
| 14 |
WATERMARK_TEXT = "SelamGPT"
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
# ===== WATERMARK FUNCTION =====
|
| 20 |
-
def add_watermark(
|
| 21 |
"""Add watermark with optimized PNG output"""
|
| 22 |
try:
|
| 23 |
-
image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
|
| 24 |
draw = ImageDraw.Draw(image)
|
| 25 |
|
| 26 |
font_size = 24
|
|
@@ -43,49 +54,29 @@ def add_watermark(image_bytes):
|
|
| 43 |
return Image.open(img_byte_arr)
|
| 44 |
except Exception as e:
|
| 45 |
print(f"Watermark error: {str(e)}")
|
| 46 |
-
return
|
| 47 |
|
| 48 |
# ===== IMAGE GENERATION =====
|
| 49 |
def generate_image(prompt):
|
| 50 |
if not prompt.strip():
|
| 51 |
return None, "⚠️ Please enter a prompt"
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
"options": {"wait_for_model": True}
|
| 65 |
-
},
|
| 66 |
-
timeout=TIMEOUT
|
| 67 |
-
)
|
| 68 |
-
|
| 69 |
-
for attempt in range(MAX_RETRIES):
|
| 70 |
-
try:
|
| 71 |
-
future = EXECUTOR.submit(api_call)
|
| 72 |
-
response = future.result()
|
| 73 |
-
|
| 74 |
-
if response.status_code == 200:
|
| 75 |
-
return add_watermark(response.content), "✔️ Generation successful"
|
| 76 |
-
elif response.status_code == 503:
|
| 77 |
-
wait_time = (attempt + 1) * 15
|
| 78 |
-
print(f"Model loading, waiting {wait_time}s...")
|
| 79 |
-
time.sleep(wait_time)
|
| 80 |
-
continue
|
| 81 |
-
else:
|
| 82 |
-
return None, f"⚠️ API Error: {response.text[:200]}"
|
| 83 |
-
except requests.Timeout:
|
| 84 |
-
return None, f"⚠️ Timeout: Model took >{TIMEOUT}s to respond"
|
| 85 |
-
except Exception as e:
|
| 86 |
-
return None, f"⚠️ Unexpected error: {str(e)[:200]}"
|
| 87 |
|
| 88 |
-
|
|
|
|
|
|
|
|
|
|
| 89 |
|
| 90 |
# ===== GRADIO THEME =====
|
| 91 |
theme = gr.themes.Default(
|
|
@@ -98,7 +89,7 @@ theme = gr.themes.Default(
|
|
| 98 |
with gr.Blocks(theme=theme, title="SelamGPT Image Generator") as demo:
|
| 99 |
gr.Markdown("""
|
| 100 |
# 🎨 SelamGPT Image Generator
|
| 101 |
-
*Powered by
|
| 102 |
""")
|
| 103 |
|
| 104 |
with gr.Row():
|
|
|
|
| 1 |
import os
|
|
|
|
| 2 |
import gradio as gr
|
| 3 |
from PIL import Image, ImageDraw, ImageFont
|
| 4 |
import io
|
| 5 |
+
import torch
|
| 6 |
+
from diffusers import DiffusionPipeline
|
| 7 |
|
| 8 |
# ===== CONFIGURATION =====
|
| 9 |
+
MODEL_NAME = "HiDream-ai/HiDream-I1-Full"
|
|
|
|
|
|
|
|
|
|
| 10 |
WATERMARK_TEXT = "SelamGPT"
|
| 11 |
+
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
| 12 |
+
TORCH_DTYPE = torch.float16 if DEVICE == "cuda" else torch.float32
|
| 13 |
+
|
| 14 |
+
# ===== MODEL LOADING =====
|
| 15 |
+
@gr.Cache() # Cache model between generations
|
| 16 |
+
def load_model():
|
| 17 |
+
pipe = DiffusionPipeline.from_pretrained(
|
| 18 |
+
MODEL_NAME,
|
| 19 |
+
torch_dtype=TORCH_DTYPE
|
| 20 |
+
).to(DEVICE)
|
| 21 |
+
|
| 22 |
+
# Optimizations
|
| 23 |
+
if DEVICE == "cuda":
|
| 24 |
+
pipe.enable_xformers_memory_efficient_attention()
|
| 25 |
+
pipe.enable_attention_slicing()
|
| 26 |
+
|
| 27 |
+
return pipe
|
| 28 |
+
|
| 29 |
+
pipe = load_model()
|
| 30 |
|
| 31 |
# ===== WATERMARK FUNCTION =====
|
| 32 |
+
def add_watermark(image):
|
| 33 |
"""Add watermark with optimized PNG output"""
|
| 34 |
try:
|
|
|
|
| 35 |
draw = ImageDraw.Draw(image)
|
| 36 |
|
| 37 |
font_size = 24
|
|
|
|
| 54 |
return Image.open(img_byte_arr)
|
| 55 |
except Exception as e:
|
| 56 |
print(f"Watermark error: {str(e)}")
|
| 57 |
+
return image
|
| 58 |
|
| 59 |
# ===== IMAGE GENERATION =====
|
| 60 |
def generate_image(prompt):
|
| 61 |
if not prompt.strip():
|
| 62 |
return None, "⚠️ Please enter a prompt"
|
| 63 |
|
| 64 |
+
try:
|
| 65 |
+
# Generate image (1024x1024 by default)
|
| 66 |
+
image = pipe(
|
| 67 |
+
prompt,
|
| 68 |
+
num_inference_steps=30,
|
| 69 |
+
guidance_scale=7.5
|
| 70 |
+
).images[0]
|
| 71 |
+
|
| 72 |
+
# Add watermark
|
| 73 |
+
watermarked = add_watermark(image)
|
| 74 |
+
return watermarked, "✔️ Generation successful"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
+
except torch.cuda.OutOfMemoryError:
|
| 77 |
+
return None, "⚠️ Out of memory! Try a simpler prompt"
|
| 78 |
+
except Exception as e:
|
| 79 |
+
return None, f"⚠️ Error: {str(e)[:200]}"
|
| 80 |
|
| 81 |
# ===== GRADIO THEME =====
|
| 82 |
theme = gr.themes.Default(
|
|
|
|
| 89 |
with gr.Blocks(theme=theme, title="SelamGPT Image Generator") as demo:
|
| 90 |
gr.Markdown("""
|
| 91 |
# 🎨 SelamGPT Image Generator
|
| 92 |
+
*Powered by HiDream-I1-Full (1024x1024 PNG output)*
|
| 93 |
""")
|
| 94 |
|
| 95 |
with gr.Row():
|