Create main.py
Browse files
main.py
ADDED
|
@@ -0,0 +1,443 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import torch
|
| 3 |
+
import boto3
|
| 4 |
+
import random
|
| 5 |
+
import string
|
| 6 |
+
import numpy as np
|
| 7 |
+
import logging
|
| 8 |
+
import datetime
|
| 9 |
+
from fastapi import FastAPI, HTTPException, Request, Response
|
| 10 |
+
from fastapi.middleware.cors import CORSMiddleware
|
| 11 |
+
from pydantic import BaseModel, constr, conint
|
| 12 |
+
from diffusers import (FluxPipeline, FluxControlNetPipeline,
|
| 13 |
+
FluxControlNetModel, FluxImg2ImgPipeline,
|
| 14 |
+
FluxInpaintPipeline, CogVideoXImageToVideoPipeline)
|
| 15 |
+
from diffusers.utils import load_image
|
| 16 |
+
from PIL import Image
|
| 17 |
+
|
| 18 |
+
# Setup logging
|
| 19 |
+
logging.basicConfig(level=logging.INFO,
|
| 20 |
+
format='%(asctime)s - %(levelname)s - %(message)s',
|
| 21 |
+
handlers=[
|
| 22 |
+
logging.FileHandler("error.txt"),
|
| 23 |
+
logging.StreamHandler()
|
| 24 |
+
])
|
| 25 |
+
|
| 26 |
+
app = FastAPI()
|
| 27 |
+
|
| 28 |
+
# Allow CORS for specific origins if needed
|
| 29 |
+
app.add_middleware(
|
| 30 |
+
CORSMiddleware,
|
| 31 |
+
allow_origins=["*"], # Update with specific domains as necessary
|
| 32 |
+
allow_credentials=True,
|
| 33 |
+
allow_methods=["*"],
|
| 34 |
+
allow_headers=["*"],
|
| 35 |
+
)
|
| 36 |
+
|
| 37 |
+
MAX_SEED = np.iinfo(np.int32).max
|
| 38 |
+
|
| 39 |
+
# AWS S3 Configuration
|
| 40 |
+
AWS_ACCESS_KEY_ID = "your-access-key-id"
|
| 41 |
+
AWS_SECRET_ACCESS_KEY = "your-secret-access-key"
|
| 42 |
+
AWS_REGION = "your-region"
|
| 43 |
+
S3_BUCKET_NAME = "your-bucket-name"
|
| 44 |
+
|
| 45 |
+
# Initialize S3 client
|
| 46 |
+
s3_client = boto3.client(
|
| 47 |
+
's3',
|
| 48 |
+
aws_access_key_id=AWS_ACCESS_KEY_ID,
|
| 49 |
+
aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
|
| 50 |
+
region_name=AWS_REGION
|
| 51 |
+
)
|
| 52 |
+
|
| 53 |
+
def log_requests(user_key: str, prompt: str):
|
| 54 |
+
timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
| 55 |
+
log_entry = f"{timestamp}, {user_key}, {prompt}\n"
|
| 56 |
+
with open("key_requests.txt", "a") as log_file:
|
| 57 |
+
log_file.write(log_entry)
|
| 58 |
+
|
| 59 |
+
# Function to upload image to S3
|
| 60 |
+
def upload_image_to_s3(image_path: str, s3_path: str):
|
| 61 |
+
try:
|
| 62 |
+
s3_client.upload_file(image_path, S3_BUCKET_NAME, s3_path)
|
| 63 |
+
return f"https://{S3_BUCKET_NAME}.s3.{AWS_REGION}.amazonaws.com/{s3_path}"
|
| 64 |
+
except Exception as e:
|
| 65 |
+
logging.error(f"Error uploading image to S3: {e}")
|
| 66 |
+
raise HTTPException(status_code=500, detail=f"Image upload failed: {str(e)}")
|
| 67 |
+
|
| 68 |
+
# Generate a random sequence of 12 numbers and 11 words
|
| 69 |
+
def generate_random_sequence():
|
| 70 |
+
random_numbers = ''.join(random.choices(string.digits, k=12)) # 12 random digits
|
| 71 |
+
random_words = ''.join(random.choices(string.ascii_lowercase, k=11)) # 11 random letters
|
| 72 |
+
return f"{random_numbers}_{random_words}"
|
| 73 |
+
|
| 74 |
+
# Load the default pipeline once globally for efficiency
|
| 75 |
+
|
| 76 |
+
# Load the default pipeline once globally for efficiency
|
| 77 |
+
try:
|
| 78 |
+
flux_pipe = FluxPipeline.from_pretrained("pranavajay/flow", torch_dtype=torch.bfloat16)
|
| 79 |
+
flux_pipe.enable_model_cpu_offload()
|
| 80 |
+
logging.info("FluxPipeline loaded successfully.")
|
| 81 |
+
except Exception as e:
|
| 82 |
+
logging.error(f"Failed to load FluxPipeline: {e}")
|
| 83 |
+
raise HTTPException(status_code=500, detail=f"Failed to load the model: {str(e)}")
|
| 84 |
+
|
| 85 |
+
try:
|
| 86 |
+
img_pipe = FluxImg2ImgPipeline.from_pretrained("pranavajay/flow", torch_dtype=torch.bfloat16)
|
| 87 |
+
img_pipe.enable_model_cpu_offload()
|
| 88 |
+
logging.info("FluxImg2ImgPipeline loaded successfully.")
|
| 89 |
+
except Exception as e:
|
| 90 |
+
logging.error(f"Failed to load FluxPipeline: {e}")
|
| 91 |
+
raise HTTPException(status_code=500, detail=f"Failed to load the model: {str(e)}")
|
| 92 |
+
|
| 93 |
+
try:
|
| 94 |
+
inpainting_pipe = FluxInpaintPipeline.from_pretrained("pranavajay/flow", torch_dtype=torch.bfloat16)
|
| 95 |
+
inpainting_pipe.enable_model_cpu_offload()
|
| 96 |
+
logging.info("FluxInpaintPipeline loaded successfully.")
|
| 97 |
+
except Exception as e:
|
| 98 |
+
logging.error(f"Failed to load FluxInpaintPipeline: {e}")
|
| 99 |
+
raise HTTPException(status_code=500, detail=f"Failed to load the model: {str(e)}")
|
| 100 |
+
|
| 101 |
+
try:
|
| 102 |
+
video = CogVideoXImageToVideoPipeline.from_pretrained(
|
| 103 |
+
"THUDM/CogVideoX-5b-I2V",
|
| 104 |
+
torch_dtype=torch.bfloat16
|
| 105 |
+
)
|
| 106 |
+
video.enable_sequential_cpu_offload()
|
| 107 |
+
video.vae.enable_tiling()
|
| 108 |
+
video.vae.enable_slicing()
|
| 109 |
+
logging.info("CogVideoXImageToVideoPipeline loaded successfully.")
|
| 110 |
+
except Exception as e:
|
| 111 |
+
logging.error(f"Failed to load CogVideoXImageToVideoPipeline: {e}")
|
| 112 |
+
raise HTTPException(status_code=500, detail=f"Failed to load the model: {str(e)}")
|
| 113 |
+
|
| 114 |
+
|
| 115 |
+
flux_controlnet_pipe = None
|
| 116 |
+
|
| 117 |
+
|
| 118 |
+
|
| 119 |
+
# Rate limiting variables
|
| 120 |
+
request_timestamps = defaultdict(list) # Store timestamps of requests per user key
|
| 121 |
+
RATE_LIMIT = 30 # Maximum requests allowed
|
| 122 |
+
TIME_WINDOW = 5 # Time window in seconds
|
| 123 |
+
|
| 124 |
+
# Available LoRA styles and ControlNet adapters
|
| 125 |
+
style_lora_mapping = {
|
| 126 |
+
"Uncensored": {"path": "enhanceaiteam/Flux-uncensored", "triggered_word": "nsfw"},
|
| 127 |
+
"Logo": {"path": "Shakker-Labs/FLUX.1-dev-LoRA-Logo-Design", "triggered_word": "logo"},
|
| 128 |
+
"Yarn": {"path": "Shakker-Labs/FLUX.1-dev-LoRA-MiaoKa-Yarn-World", "triggered_word": "mkym this is made of wool"},
|
| 129 |
+
"Anime": {"path": "prithivMLmods/Canopus-LoRA-Flux-Anime", "triggered_word": "anime"},
|
| 130 |
+
"Comic": {"path": "wkplhc/comic", "triggered_word": "comic"}
|
| 131 |
+
}
|
| 132 |
+
|
| 133 |
+
adapter_controlnet_mapping = {
|
| 134 |
+
"Canny": "InstantX/FLUX.1-dev-controlnet-canny",
|
| 135 |
+
"Depth": "Shakker-Labs/FLUX.1-dev-ControlNet-Depth",
|
| 136 |
+
"Pose": "Shakker-Labs/FLUX.1-dev-ControlNet-Pose",
|
| 137 |
+
"Upscale": "jasperai/Flux.1-dev-Controlnet-Upscaler"
|
| 138 |
+
}
|
| 139 |
+
|
| 140 |
+
# Request model for query parameters
|
| 141 |
+
class GenerateImageRequest(BaseModel):
|
| 142 |
+
prompt: constr(min_length=1) # Ensures prompt is not empty
|
| 143 |
+
guidance_scale: float = 7.5
|
| 144 |
+
seed: conint(ge=0, le=MAX_SEED) = 42
|
| 145 |
+
randomize_seed: bool = False
|
| 146 |
+
height: conint(gt=0) = 768
|
| 147 |
+
width: conint(gt=0) = 1360
|
| 148 |
+
control_image_url: str = "https://enhanceai.s3.amazonaws.com/792e2322-77fe-4070-aac4-7fa8d9e29c11_1.png"
|
| 149 |
+
controlnet_conditioning_scale: float = 0.6
|
| 150 |
+
num_inference_steps: conint(gt=0) = 50
|
| 151 |
+
num_images_per_prompt: conint(gt=0, le=5) = 1 # Limit to max 5 images per request
|
| 152 |
+
style: str = None # Optional LoRA style
|
| 153 |
+
adapter: str = None # Optional ControlNet adapter
|
| 154 |
+
user_key: str # API user key
|
| 155 |
+
|
| 156 |
+
def log_request(key: str, query: str):
|
| 157 |
+
with open("key.txt", "a") as f:
|
| 158 |
+
f.write(f"{datetime.datetime.now()} - Key: {key} - Query: {query}\n")
|
| 159 |
+
|
| 160 |
+
def apply_lora_style(pipe, style, prompt):
|
| 161 |
+
""" Apply the specified LoRA style to the prompt and load weights. """
|
| 162 |
+
if style in style_lora_mapping:
|
| 163 |
+
lora_path = style_lora_mapping[style]["path"]
|
| 164 |
+
triggered_word = style_lora_mapping[style]["triggered_word"]
|
| 165 |
+
pipe.load_lora_weights(lora_path)
|
| 166 |
+
return f"{triggered_word} {prompt}" # Add triggered word to prompt
|
| 167 |
+
return prompt
|
| 168 |
+
|
| 169 |
+
def set_controlnet_adapter(adapter: str, is_inpainting: bool = False):
|
| 170 |
+
"""
|
| 171 |
+
Set the ControlNet adapter for the pipeline.
|
| 172 |
+
|
| 173 |
+
Parameters:
|
| 174 |
+
adapter (str): The key to identify which ControlNet adapter to load.
|
| 175 |
+
is_inpainting (bool, optional): Whether to use the inpainting pipeline. Defaults to False.
|
| 176 |
+
|
| 177 |
+
Raises:
|
| 178 |
+
ValueError: If the adapter is not found in the adapter_controlnet_mapping.
|
| 179 |
+
"""
|
| 180 |
+
global flux_controlnet_pipe
|
| 181 |
+
|
| 182 |
+
# Check if the adapter is valid
|
| 183 |
+
if adapter not in adapter_controlnet_mapping:
|
| 184 |
+
raise ValueError(f"Invalid ControlNet adapter: {adapter}")
|
| 185 |
+
|
| 186 |
+
# Get the ControlNet model path based on the adapter
|
| 187 |
+
controlnet_model_path = adapter_controlnet_mapping[adapter]
|
| 188 |
+
|
| 189 |
+
# Load the ControlNet model with the specified torch_dtype
|
| 190 |
+
controlnet = FluxControlNetModel.from_pretrained(controlnet_model_path, torch_dtype=torch.bfloat16)
|
| 191 |
+
|
| 192 |
+
# Select the appropriate pipeline (inpainting or standard)
|
| 193 |
+
pipeline_cls = FluxControlNetInpaintPipeline if is_inpainting else FluxControlNetPipeline
|
| 194 |
+
|
| 195 |
+
# Load the pipeline
|
| 196 |
+
flux_controlnet_pipe = pipeline_cls.from_pretrained(
|
| 197 |
+
"pranavajay/flow", controlnet=controlnet, torch_dtype=torch.bfloat16
|
| 198 |
+
)
|
| 199 |
+
|
| 200 |
+
# Move the pipeline to the GPU
|
| 201 |
+
flux_controlnet_pipe.to("cuda")
|
| 202 |
+
|
| 203 |
+
logging.info(f"ControlNet adapter '{adapter}' loaded successfully.")
|
| 204 |
+
|
| 205 |
+
|
| 206 |
+
|
| 207 |
+
|
| 208 |
+
|
| 209 |
+
|
| 210 |
+
|
| 211 |
+
def rate_limit(user_key: str):
|
| 212 |
+
""" Check if the user is exceeding the rate limit. """
|
| 213 |
+
current_time = time.time()
|
| 214 |
+
|
| 215 |
+
# Clean up old timestamps
|
| 216 |
+
request_timestamps[user_key] = [t for t in request_timestamps[user_key] if current_time - t < TIME_WINDOW]
|
| 217 |
+
|
| 218 |
+
if len(request_timestamps[user_key]) >= RATE_LIMIT:
|
| 219 |
+
logging.info(f"Rate limit exceeded for user_key: {user_key}")
|
| 220 |
+
return False
|
| 221 |
+
|
| 222 |
+
# Record the new request timestamp
|
| 223 |
+
request_timestamps[user_key].append(current_time)
|
| 224 |
+
return True
|
| 225 |
+
|
| 226 |
+
@app.post("/text_to_image/")
|
| 227 |
+
async def generate_image(req: GenerateImageRequest):
|
| 228 |
+
seed = req.seed
|
| 229 |
+
if not rate_limit(req.user_key):
|
| 230 |
+
log_requests(req.user_key, req.prompt) # Log the request when rate limit is exceeded
|
| 231 |
+
|
| 232 |
+
retries = 3 # Number of retries for transient errors
|
| 233 |
+
|
| 234 |
+
for attempt in range(retries):
|
| 235 |
+
try:
|
| 236 |
+
# Check if prompt is None or empty
|
| 237 |
+
if not req.prompt or req.prompt.strip() == "":
|
| 238 |
+
raise ValueError("Prompt cannot be empty.")
|
| 239 |
+
|
| 240 |
+
original_prompt = req.prompt # Save the original prompt
|
| 241 |
+
|
| 242 |
+
# Set ControlNet if adapter is provided
|
| 243 |
+
if req.adapter:
|
| 244 |
+
try:
|
| 245 |
+
set_controlnet_adapter(req.adapter)
|
| 246 |
+
except Exception as e:
|
| 247 |
+
logging.error(f"Error setting ControlNet adapter: {e}")
|
| 248 |
+
raise HTTPException(status_code=400, detail=f"Failed to load ControlNet adapter: {str(e)}")
|
| 249 |
+
apply_lora_style(flux_controlnet_pipe, req.style, req.prompt)
|
| 250 |
+
|
| 251 |
+
|
| 252 |
+
# Load control image
|
| 253 |
+
try:
|
| 254 |
+
control_image = load_image(req.control_image_url)
|
| 255 |
+
except Exception as e:
|
| 256 |
+
logging.error(f"Error loading control image from URL: {e}")
|
| 257 |
+
raise HTTPException(status_code=400, detail="Invalid control image URL or image could not be loaded.")
|
| 258 |
+
|
| 259 |
+
# Image generation with ControlNet
|
| 260 |
+
try:
|
| 261 |
+
if req.randomize_seed:
|
| 262 |
+
seed = random.randint(0, MAX_SEED)
|
| 263 |
+
generator = torch.Generator().manual_seed(seed)
|
| 264 |
+
|
| 265 |
+
images = flux_controlnet_pipe(
|
| 266 |
+
prompt=modified_prompt,
|
| 267 |
+
guidance_scale=req.guidance_scale,
|
| 268 |
+
height=req.height,
|
| 269 |
+
width=req.width,
|
| 270 |
+
num_inference_steps=req.num_inference_steps,
|
| 271 |
+
num_images_per_prompt=req.num_images_per_prompt,
|
| 272 |
+
control_image=control_image,
|
| 273 |
+
generator=generator,
|
| 274 |
+
controlnet_conditioning_scale=req.controlnet_conditioning_scale
|
| 275 |
+
).images
|
| 276 |
+
except torch.cuda.OutOfMemoryError:
|
| 277 |
+
logging.error("GPU out of memory error while generating images with ControlNet.")
|
| 278 |
+
raise HTTPException(status_code=500, detail="GPU overload occurred while generating images. Try reducing the resolution or number of steps.")
|
| 279 |
+
except Exception as e:
|
| 280 |
+
logging.error(f"Error during image generation with ControlNet: {e}")
|
| 281 |
+
raise HTTPException(status_code=500, detail=f"Error during image generation: {str(e)}")
|
| 282 |
+
else:
|
| 283 |
+
# Image generation without ControlNet
|
| 284 |
+
try:
|
| 285 |
+
apply_lora_style(flux_pipe, req.style, req.prompt)
|
| 286 |
+
if req.randomize_seed:
|
| 287 |
+
seed = random.randint(0, MAX_SEED)
|
| 288 |
+
generator = torch.Generator().manual_seed(seed)
|
| 289 |
+
|
| 290 |
+
images = flux_pipe(
|
| 291 |
+
prompt=modified_prompt,
|
| 292 |
+
guidance_scale=req.guidance_scale,
|
| 293 |
+
height=req.height,
|
| 294 |
+
width=req.width,
|
| 295 |
+
num_inference_steps=req.num_inference_steps,
|
| 296 |
+
num_images_per_prompt=req.num_images_per_prompt,
|
| 297 |
+
generator=generator
|
| 298 |
+
).images
|
| 299 |
+
except torch.cuda.OutOfMemoryError:
|
| 300 |
+
logging.error("GPU out of memory error while generating images without ControlNet.")
|
| 301 |
+
raise HTTPException(status_code=500, detail="GPU overload occurred while generating images. Try reducing the resolution or number of steps.")
|
| 302 |
+
except Exception as e:
|
| 303 |
+
logging.error(f"Error during image generation without ControlNet: {e}")
|
| 304 |
+
raise HTTPException(status_code=500, detail=f"Error during image generation: {str(e)}")
|
| 305 |
+
|
| 306 |
+
# Saving images and uploading to S3
|
| 307 |
+
image_urls = []
|
| 308 |
+
for i, img in enumerate(images):
|
| 309 |
+
image_path = f"generated_images/{generate_random_sequence()}.png"
|
| 310 |
+
img.save(image_path)
|
| 311 |
+
image_url = upload_image_to_s3(image_path, image_path)
|
| 312 |
+
image_urls.append(image_url)
|
| 313 |
+
os.remove(image_path) # Clean up local files after upload
|
| 314 |
+
|
| 315 |
+
return {"status": "success", "output": image_url, "prompt": original_prompt, "height": req.height, "width": req.width, "scale": req.guidance_scale, "step": step, "sytle": req.sytle, "adapter": req.adapter}
|
| 316 |
+
|
| 317 |
+
except Exception as e:
|
| 318 |
+
logging.error(f"Attempt {attempt + 1} failed: {e}")
|
| 319 |
+
if attempt == retries - 1: # Last attempt
|
| 320 |
+
raise HTTPException(status_code=500, detail=f"Failed to generate image after multiple attempts: {str(e)}")
|
| 321 |
+
continue # Retry on transient errors
|
| 322 |
+
|
| 323 |
+
|
| 324 |
+
# Image-to-Image request model
|
| 325 |
+
class GenerateImageToImageRequest(BaseModel):
|
| 326 |
+
prompt: str = None # Prompt can be None
|
| 327 |
+
image: str = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png"
|
| 328 |
+
strength: float = 0.7
|
| 329 |
+
guidance_scale: float = 7.5
|
| 330 |
+
seed: conint(ge=0, le=MAX_SEED) = 42
|
| 331 |
+
randomize_seed: bool = False
|
| 332 |
+
height: conint(gt=0) = 768
|
| 333 |
+
width: conint(gt=0) = 1360
|
| 334 |
+
control_image_url: str = None # Optional ControlNet image
|
| 335 |
+
controlnet_conditioning_scale: float = 0.6
|
| 336 |
+
num_inference_steps: conint(gt=0) = 50
|
| 337 |
+
num_images_per_prompt: conint(gt=0, le=5) = 1
|
| 338 |
+
style: str = None # Optional LoRA style
|
| 339 |
+
adapter: str = None # Optional ControlNet adapter
|
| 340 |
+
user_key: str # API user key
|
| 341 |
+
|
| 342 |
+
@app.post("/image_to_image/")
|
| 343 |
+
async def generate_image_to_image(req: GenerateImageToImageRequest):
|
| 344 |
+
seed = req.seed
|
| 345 |
+
original_prompt = req.prompt
|
| 346 |
+
modified_prompt = original_prompt
|
| 347 |
+
|
| 348 |
+
# Check if user is exceeding rate limit
|
| 349 |
+
if not rate_limit(req.user_key):
|
| 350 |
+
log_requests(req.user_key, req.prompt if req.prompt else "No prompt")
|
| 351 |
+
raise HTTPException(status_code=429, detail="Rate limit exceeded")
|
| 352 |
+
|
| 353 |
+
retries = 3 # Number of retries for transient errors
|
| 354 |
+
|
| 355 |
+
for attempt in range(retries):
|
| 356 |
+
try:
|
| 357 |
+
# Check if prompt is None or empty
|
| 358 |
+
if not req.prompt or req.prompt.strip() == "":
|
| 359 |
+
raise ValueError("Prompt cannot be empty.")
|
| 360 |
+
|
| 361 |
+
original_prompt = req.prompt # Save the original prompt
|
| 362 |
+
|
| 363 |
+
# Set ControlNet if adapter is provided
|
| 364 |
+
if req.adapter:
|
| 365 |
+
try:
|
| 366 |
+
set_controlnet_adapter(req.adapter)
|
| 367 |
+
except Exception as e:
|
| 368 |
+
logging.error(f"Error setting ControlNet adapter: {e}")
|
| 369 |
+
raise HTTPException(status_code=400, detail=f"Failed to load ControlNet adapter: {str(e)}")
|
| 370 |
+
apply_lora_style(flux_controlnet_pipe, req.style, req.prompt)
|
| 371 |
+
|
| 372 |
+
|
| 373 |
+
# Load control image
|
| 374 |
+
try:
|
| 375 |
+
control_image = load_image(req.control_image_url)
|
| 376 |
+
except Exception as e:
|
| 377 |
+
logging.error(f"Error loading control image from URL: {e}")
|
| 378 |
+
raise HTTPException(status_code=400, detail="Invalid control image URL or image could not be loaded.")
|
| 379 |
+
|
| 380 |
+
# Image generation with ControlNet
|
| 381 |
+
try:
|
| 382 |
+
if req.randomize_seed:
|
| 383 |
+
seed = random.randint(0, MAX_SEED)
|
| 384 |
+
generator = torch.Generator().manual_seed(seed)
|
| 385 |
+
|
| 386 |
+
images = flux_controlnet_pipe(
|
| 387 |
+
prompt=modified_prompt,
|
| 388 |
+
guidance_scale=req.guidance_scale,
|
| 389 |
+
height=req.height,
|
| 390 |
+
width=req.width,
|
| 391 |
+
num_inference_steps=req.num_inference_steps,
|
| 392 |
+
num_images_per_prompt=req.num_images_per_prompt,
|
| 393 |
+
control_image=control_image,
|
| 394 |
+
generator=generator,
|
| 395 |
+
controlnet_conditioning_scale=req.controlnet_conditioning_scale
|
| 396 |
+
).images
|
| 397 |
+
except torch.cuda.OutOfMemoryError:
|
| 398 |
+
logging.error("GPU out of memory error while generating images with ControlNet.")
|
| 399 |
+
raise HTTPException(status_code=500, detail="GPU overload occurred while generating images. Try reducing the resolution or number of steps.")
|
| 400 |
+
except Exception as e:
|
| 401 |
+
logging.error(f"Error during image generation with ControlNet: {e}")
|
| 402 |
+
raise HTTPException(status_code=500, detail=f"Error during image generation: {str(e)}")
|
| 403 |
+
else:
|
| 404 |
+
# Image generation without ControlNet
|
| 405 |
+
try:
|
| 406 |
+
apply_lora_style(img_pipe, req.style, req.prompt)
|
| 407 |
+
if req.randomize_seed:
|
| 408 |
+
seed = random.randint(0, MAX_SEED)
|
| 409 |
+
generator = torch.Generator().manual_seed(seed)
|
| 410 |
+
source = load_image(req.image)
|
| 411 |
+
images = img_pipe(
|
| 412 |
+
prompt=modified_prompt,
|
| 413 |
+
image=source,
|
| 414 |
+
strength=req.strength,
|
| 415 |
+
guidance_scale=req.guidance_scale,
|
| 416 |
+
height=req.height,
|
| 417 |
+
width=req.width,
|
| 418 |
+
num_inference_steps=req.num_inference_steps,
|
| 419 |
+
num_images_per_prompt=req.num_images_per_prompt,
|
| 420 |
+
generator=generator
|
| 421 |
+
).images
|
| 422 |
+
except torch.cuda.OutOfMemoryError:
|
| 423 |
+
logging.error("GPU out of memory error while generating images without ControlNet.")
|
| 424 |
+
raise HTTPException(status_code=500, detail="GPU overload occurred while generating images. Try reducing the resolution or number of steps.")
|
| 425 |
+
except Exception as e:
|
| 426 |
+
logging.error(f"Error during image generation without ControlNet: {e}")
|
| 427 |
+
raise HTTPException(status_code=500, detail=f"Error during image generation: {str(e)}")
|
| 428 |
+
|
| 429 |
+
# Saving images and uploading to S3
|
| 430 |
+
image_urls = []
|
| 431 |
+
for i, img in enumerate(images):
|
| 432 |
+
image_path = f"generated_images/{generate_random_sequence()}.png"
|
| 433 |
+
img.save(image_path)
|
| 434 |
+
image_url = upload_image_to_s3(image_path, image_path)
|
| 435 |
+
image_urls.append(image_url)
|
| 436 |
+
os.remove(image_path) # Clean up local files after upload
|
| 437 |
+
|
| 438 |
+
return {"status": "success", "output": image_url, "prompt": original_prompt, "height": req.height, "width": width, "image": req.image, "strength": req.strength, "scale": req.guidance_scale, "step": step, "sytle": req.sytle, "adapter": req.adapter}
|
| 439 |
+
|
| 440 |
+
except Exception as e:
|
| 441 |
+
logging.error(f"Attempt {attempt + 1} failed: {e}")
|
| 442 |
+
if attempt == retries - 1: # Last attempt
|
| 443 |
+
raise HTTPException(status_code=500, detail=f"Failed to generate image after m
|