File size: 10,264 Bytes
18faf97 |
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 |
# ----------------------------------------------------------------------
# IMPORTS
# ----------------------------------------------------------------------
import io
import json
import re
import traceback
import time
import logging
import os
import psutil
import gc
from datetime import datetime, timezone, timedelta
from typing import Dict, List, Optional, Tuple, Any, Callable, Union
from PIL import Image
from pydantic import BaseModel, Field
import torch
import numpy as np
# ----------------------------------------------------------------------
# SYSTEM MONITORING
# ----------------------------------------------------------------------
def get_system_info():
cpu_percent = psutil.cpu_percent(interval=0.1)
memory = psutil.virtual_memory()
info = {
"cpu_percent": round(cpu_percent, 2),
"memory_percent": round(memory.percent, 2),
"memory_total_gb": round(memory.total / (1024**3), 2),
"memory_available_gb": round(memory.available / (1024**3), 2),
"memory_used_gb": round(memory.used / (1024**3), 2),
}
if torch.cuda.is_available() and os.getenv("SPACE_ID") is None:
try:
info["gpu_count"] = torch.cuda.device_count()
for i in range(torch.cuda.device_count()):
torch.cuda.set_device(i)
gpu_allocated = torch.cuda.memory_allocated(i) / (1024**3)
gpu_reserved = torch.cuda.memory_reserved(i) / (1024**3)
gpu_total = torch.cuda.get_device_properties(i).total_memory / (1024**3)
info[f"gpu_{i}_memory_allocated_gb"] = round(gpu_allocated, 2)
info[f"gpu_{i}_memory_reserved_gb"] = round(gpu_reserved, 2)
info[f"gpu_{i}_memory_total_gb"] = round(gpu_total, 2)
info[f"gpu_{i}_memory_allocated_percent"] = round((gpu_allocated / gpu_total * 100) if gpu_total > 0 else 0, 2)
info[f"gpu_{i}_memory_reserved_percent"] = round((gpu_reserved / gpu_total * 100) if gpu_total > 0 else 0, 2)
except Exception as e:
logging.warning(f"GPU memory monitoring failed: {e}")
info["gpu_count"] = 0
elif torch.cuda.is_available():
info["gpu_available"] = True
info["gpu_count"] = 0 # Will be set properly within GPU functions
return info
def cleanup_memory():
gc.collect()
if torch.cuda.is_available():
try:
torch.cuda.empty_cache()
if not (hasattr(torch, '_C') and hasattr(torch._C, '_cuda_getDeviceCount') and torch._C._cuda_getDeviceCount() > 0):
torch.cuda.synchronize()
except RuntimeError as e:
if "CUDA must not be initialized" not in str(e):
raise
def force_gpu_memory_update():
if torch.cuda.is_available() and torch.cuda.device_count() > 0 and os.getenv("SPACE_ID") is None:
try:
temp_tensor = torch.randn(1, device='cuda')
del temp_tensor
torch.cuda.empty_cache()
except Exception as e:
logging.debug(f"GPU memory update failed: {e}")
# ----------------------------------------------------------------------
# LOGGING CONFIGURATION
# ----------------------------------------------------------------------
LOG_LEVEL_MAP = {
"DEBUG": logging.DEBUG,
"INFO": logging.INFO,
"SUCCESS": logging.INFO,
"WARNING": logging.WARNING,
"ERROR": logging.ERROR,
"CRITICAL": logging.CRITICAL,
"PROCESSING": logging.INFO,
"RETRY": logging.WARNING,
"JSON": logging.INFO,
"PERFORMANCE": logging.INFO,
"MEMORY": logging.INFO
}
EMOJI_MAP = {
"DEBUG": "🔍",
"INFO": "ℹ️",
"SUCCESS": "✅",
"WARNING": "⚠️",
"ERROR": "❌",
"CRITICAL": "🔥",
"PROCESSING": "⚙️",
"RETRY": "🔄",
"JSON": "📊",
"PERFORMANCE": "⚡",
"MEMORY": "💾"
}
def setup_logging():
LOG_FORMAT = "%(asctime)s [%(levelname)s] %(module)s: %(message)s"
LOG_LEVEL = logging.INFO
if os.getenv("DEBUG", "").lower() == "true":
LOG_LEVEL = logging.DEBUG
for name, level in LOG_LEVEL_MAP.items():
if not hasattr(logging, name):
setattr(logging, name, level)
class EmojiFormatter(logging.Formatter):
def formatTime(self, record, datefmt=None):
dt = datetime.fromtimestamp(record.created, tz=timezone.utc)
dt = dt - timedelta(hours=2)
if datefmt:
return dt.strftime(datefmt)
return dt.strftime('%Y-%m-%d %H:%M:%S,%f')[:-3]
def format(self, record):
if not getattr(record, 'emoji_prefixed', False):
for emoji in EMOJI_MAP.values():
if str(record.msg).startswith(emoji):
record.emoji_prefixed = True
break
if not getattr(record, 'emoji_prefixed', False):
for name, level in LOG_LEVEL_MAP.items():
if record.levelno == level:
record.msg = f"{EMOJI_MAP.get(name, '')} {record.msg}"
record.emoji_prefixed = True
break
return super().format(record)
formatter = EmojiFormatter(LOG_FORMAT)
handler = logging.StreamHandler()
handler.setFormatter(formatter)
# Clear all existing loggers to avoid duplicates
logging.root.handlers = []
# Configure root logger
root = logging.getLogger()
root.setLevel(LOG_LEVEL)
root.addHandler(handler)
# Configure specific loggers with their own handlers to avoid duplicates
for logger_name in ["uvicorn", "uvicorn.access", "uvicorn.error", "uvicorn.asgi"]:
logger = logging.getLogger(logger_name)
logger.handlers = [] # Clear any existing handlers
logger.propagate = False # Prevent propagation to root logger
# Don't add handler - let them use default formatting
# Prevent duplicate logging from libraries
for logger_name in ["_client", "httpx._client", "httpcore._sync.connection_pool", "httpcore._sync.http11"]:
logger = logging.getLogger(logger_name)
logger.handlers = []
if logger_name == "_client":
# Keep _client logs but with our handler
logger.addHandler(handler)
logger.setLevel(LOG_LEVEL)
logger.propagate = False
# Silence overly verbose loggers
for logger_name in ["PIL", "PIL.Image", "transformers", "accelerate"]:
logger = logging.getLogger(logger_name)
logger.setLevel(logging.WARNING)
logging.info("Application logging configured successfully")
# Log system info at startup
system_info = get_system_info()
logging.info(f"System info: {system_info}")
# ----------------------------------------------------------------------
# PERFORMANCE DECORATORS
# ----------------------------------------------------------------------
def measure_performance(func: Callable) -> Callable:
def wrapper(*args, **kwargs):
start_time = time.perf_counter()
force_gpu_memory_update()
start_memory = get_system_info()
try:
result = func(*args, **kwargs)
end_time = time.perf_counter()
force_gpu_memory_update()
end_memory = get_system_info()
duration = end_time - start_time
memory_used_delta_gb = end_memory["memory_used_gb"] - start_memory["memory_used_gb"]
memory_percent_delta = end_memory["memory_percent"] - start_memory["memory_percent"]
total_memory_gb = end_memory.get("memory_total_gb", 0)
memory_delta = {
"memory_used_delta_gb": round(memory_used_delta_gb, 2),
"memory_percent_delta": round(memory_percent_delta, 2),
"memory_used_percent": round(end_memory["memory_percent"], 2),
"memory_total_gb": round(total_memory_gb, 2)
}
if torch.cuda.is_available() and torch.cuda.device_count() > 0 and not os.getenv("SPACE_ID"):
for i in range(torch.cuda.device_count()):
allocated_key = f"gpu_{i}_memory_allocated_gb"
reserved_key = f"gpu_{i}_memory_reserved_gb"
total_key = f"gpu_{i}_memory_total_gb"
start_allocated = start_memory.get(allocated_key, 0)
end_allocated = end_memory.get(allocated_key, 0)
start_reserved = start_memory.get(reserved_key, 0)
end_reserved = end_memory.get(reserved_key, 0)
if start_allocated > 0 or end_allocated > 0 or start_reserved > 0 or end_reserved > 0:
allocated_delta = end_allocated - start_allocated
reserved_delta = end_reserved - start_reserved
gpu_total = end_memory.get(total_key, 0)
memory_delta[f"gpu_{i}_allocated_delta_gb"] = round(allocated_delta, 2)
memory_delta[f"gpu_{i}_reserved_delta_gb"] = round(reserved_delta, 2)
memory_delta[f"gpu_{i}_allocated_percent"] = round(end_memory.get(f"gpu_{i}_memory_allocated_percent", 0), 2)
memory_delta[f"gpu_{i}_reserved_percent"] = round(end_memory.get(f"gpu_{i}_memory_reserved_percent", 0), 2)
memory_delta[f"gpu_{i}_total_gb"] = round(gpu_total, 2)
logging.log(
LOG_LEVEL_MAP["PERFORMANCE"],
f"{EMOJI_MAP['PERFORMANCE']} {func.__name__} completed in {duration:.3f}s | Memory delta: {memory_delta}"
)
return result
except Exception as e:
end_time = time.perf_counter()
duration = end_time - start_time
logging.error(f"{func.__name__} failed after {duration:.3f}s: {str(e)}")
raise
return wrapper |