Spaces:
Sleeping
Sleeping
File size: 19,131 Bytes
e9bc512 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 |
import websockets
import json
import numpy as np
from typing import Dict, Any, Optional, Union
import threading
from queue import Queue
import time
import asyncio
import hashlib
class WebSocketGPUStorage:
# Singleton instance
_instance = None
_lock = threading.Lock()
def __new__(cls, url: str = "wss://factorst-wbs1.hf.space/ws"):
with cls._lock:
if cls._instance is None:
cls._instance = super().__new__(cls)
cls._instance._init_singleton(url)
return cls._instance
def _init_singleton(self, url: str):
"""Initialize the singleton instance"""
if hasattr(self, 'initialized'):
return
self.url = url
self.websocket = None
self.connected = False
self.message_queue = Queue()
self.response_queues: Dict[str, Queue] = {}
self.lock = threading.Lock()
self._closing = False
self._loop = None
self.error_count = 0
self.last_error_time = 0
self.max_retries = 5
self.tensor_registry: Dict[str, Dict[str, Any]] = {} # Track tensor metadata
self.model_registry: Dict[str, Dict[str, Any]] = {} # Track loaded models
self.resource_monitor = {
'vram_used': 0,
'active_tensors': 0,
'loaded_models': set()
}
# Start WebSocket connection in a separate thread
self.ws_thread = threading.Thread(target=self._run_websocket_loop, daemon=True)
self.ws_thread.start()
self.initialized = True
def __init__(self, url: str = "wss://factorst-wbs1.hf.space/ws"):
"""This will actually just return the singleton instance"""
pass
def _run_websocket_loop(self):
self._loop = asyncio.new_event_loop()
asyncio.set_event_loop(self._loop)
self._loop.run_until_complete(self._websocket_handler())
async def _websocket_handler(self):
while not self._closing:
try:
async with websockets.connect(self.url) as websocket:
self.websocket = websocket
self.connected = True
self.error_count = 0 # Reset error count on successful connection
print("Connected to GPU storage server")
while True:
# Handle outgoing messages
try:
while not self.message_queue.empty():
msg_id, operation = self.message_queue.get()
await websocket.send(json.dumps(operation))
# Wait for response with timeout
try:
response = await asyncio.wait_for(websocket.recv(), timeout=30)
response_data = json.loads(response)
# Put response in corresponding queue
if msg_id in self.response_queues:
self.response_queues[msg_id].put(response_data)
except asyncio.TimeoutError:
if msg_id in self.response_queues:
self.response_queues[msg_id].put({
"status": "error",
"message": "Operation timed out"
})
except Exception as e:
if msg_id in self.response_queues:
self.response_queues[msg_id].put({
"status": "error",
"message": f"Error processing response: {str(e)}"
})
except Exception as e:
print(f"Error processing message: {str(e)}")
# Keep connection alive with heartbeat
try:
await websocket.ping()
except:
break # Break inner loop on ping failure
await asyncio.sleep(0.001) # 1ms sleep for electron-speed response
except Exception as e:
print(f"WebSocket connection error: {e}")
self.connected = False
await asyncio.sleep(1) # Wait before reconnecting
def _send_operation(self, operation: Dict[str, Any]) -> Dict[str, Any]:
if self._closing:
return {"status": "error", "message": "WebSocket is closing"}
if not self.wait_for_connection(timeout=10):
return {"status": "error", "message": "Not connected to GPU storage server"}
msg_id = str(time.time())
response_queue = Queue()
with self.lock:
self.response_queues[msg_id] = response_queue
self.message_queue.put((msg_id, operation))
try:
# Wait for response with configurable timeout
response = response_queue.get(timeout=30) # Extended timeout for large models
if response.get("status") == "error" and "model_size" in operation:
# Retry once for model loading operations
self.message_queue.put((msg_id, operation))
response = response_queue.get(timeout=30)
except Exception as e:
response = {"status": "error", "message": f"Operation failed: {str(e)}"}
finally:
with self.lock:
if msg_id in self.response_queues:
del self.response_queues[msg_id]
return response
def store_tensor(self, tensor_id: str, data: np.ndarray, model_size: Optional[int] = None) -> bool:
try:
if data is None:
raise ValueError("Cannot store None tensor")
# Calculate tensor metadata
tensor_shape = data.shape
tensor_dtype = str(data.dtype)
tensor_size = data.nbytes
operation = {
'operation': 'vram',
'type': 'write',
'block_id': tensor_id,
'data': data.tolist(),
'model_size': model_size if model_size is not None else -1, # -1 indicates unlimited
'metadata': {
'shape': tensor_shape,
'dtype': tensor_dtype,
'size': tensor_size,
'timestamp': time.time()
}
}
response = self._send_operation(operation)
if response.get('status') == 'success':
# Update tensor registry
with self.lock:
self.tensor_registry[tensor_id] = {
'shape': tensor_shape,
'dtype': tensor_dtype,
'size': tensor_size,
'timestamp': time.time()
}
self.resource_monitor['vram_used'] += tensor_size
self.resource_monitor['active_tensors'] += 1
return True
else:
print(f"Failed to store tensor {tensor_id}: {response.get('message', 'Unknown error')}")
return False
except Exception as e:
print(f"Error storing tensor {tensor_id}: {str(e)}")
return False
def load_tensor(self, tensor_id: str) -> Optional[np.ndarray]:
try:
# Check tensor registry first
if tensor_id not in self.tensor_registry:
print(f"Tensor {tensor_id} not registered in VRAM")
return None
operation = {
'operation': 'vram',
'type': 'read',
'block_id': tensor_id,
'expected_metadata': self.tensor_registry.get(tensor_id, {})
}
response = self._send_operation(operation)
if response.get('status') == 'success':
data = response.get('data')
if data is None:
print(f"No data found for tensor {tensor_id}")
return None
# Verify tensor metadata
metadata = response.get('metadata', {})
expected_metadata = self.tensor_registry.get(tensor_id, {})
if metadata.get('shape') != expected_metadata.get('shape'):
print(f"Warning: Tensor {tensor_id} shape mismatch")
try:
# Convert to numpy array with correct dtype
arr = np.array(data, dtype=np.dtype(expected_metadata.get('dtype', 'float32')))
if arr.shape != expected_metadata.get('shape'):
arr = arr.reshape(expected_metadata.get('shape'))
return arr
except Exception as e:
print(f"Error converting tensor data: {str(e)}")
return None
else:
print(f"Failed to load tensor {tensor_id}: {response.get('message', 'Unknown error')}")
return None
except Exception as e:
print(f"Error loading tensor {tensor_id}: {str(e)}")
return None
def store_state(self, component: str, state_id: str, state_data: Dict[str, Any]) -> bool:
try:
operation = {
'operation': 'state',
'type': 'save',
'component': component,
'state_id': state_id,
'data': state_data,
'timestamp': time.time()
}
response = self._send_operation(operation)
if response.get('status') != 'success':
print(f"Failed to store state for {component}/{state_id}: {response.get('message', 'Unknown error')}")
return False
return True
except Exception as e:
print(f"Error storing state for {component}/{state_id}: {str(e)}")
return False
def load_state(self, component: str, state_id: str) -> Optional[Dict[str, Any]]:
try:
operation = {
'operation': 'state',
'type': 'load',
'component': component,
'state_id': state_id
}
response = self._send_operation(operation)
if response.get('status') == 'success':
data = response.get('data')
if data is None:
print(f"No state found for {component}/{state_id}")
return None
return data
else:
print(f"Failed to load state for {component}/{state_id}: {response.get('message', 'Unknown error')}")
return None
except Exception as e:
print(f"Error loading state for {component}/{state_id}: {str(e)}")
return None
def is_model_loaded(self, model_name: str) -> bool:
"""Check if a model is already loaded in VRAM"""
return model_name in self.resource_monitor['loaded_models']
def load_model(self, model_name: str, model_path: Optional[str] = None, model_data: Optional[Dict] = None) -> bool:
"""Load a model into VRAM if not already loaded"""
try:
# Check if model is already loaded
if self.is_model_loaded(model_name):
print(f"Model {model_name} already loaded in VRAM")
return True
# Calculate model hash if path provided
model_hash = None
if model_path:
model_hash = self._calculate_model_hash(model_path)
operation = {
'operation': 'model',
'type': 'load',
'model_name': model_name,
'model_hash': model_hash,
'model_data': model_data
}
response = self._send_operation(operation)
if response.get('status') == 'success':
with self.lock:
self.model_registry[model_name] = {
'hash': model_hash,
'timestamp': time.time(),
'tensors': response.get('tensor_ids', [])
}
self.resource_monitor['loaded_models'].add(model_name)
print(f"Successfully loaded model {model_name}")
return True
else:
print(f"Failed to load model {model_name}: {response.get('message', 'Unknown error')}")
return False
except Exception as e:
print(f"Error loading model {model_name}: {str(e)}")
return False
def _calculate_model_hash(self, model_path: str) -> str:
"""Calculate SHA256 hash of model file"""
try:
sha256_hash = hashlib.sha256()
with open(model_path, "rb") as f:
for byte_block in iter(lambda: f.read(4096), b""):
sha256_hash.update(byte_block)
return sha256_hash.hexdigest()
except Exception as e:
print(f"Error calculating model hash: {str(e)}")
return ""
def cache_data(self, key: str, data: Any) -> bool:
operation = {
'operation': 'cache',
'type': 'set',
'key': key,
'data': data
}
response = self._send_operation(operation)
return response.get('status') == 'success'
def get_cached_data(self, key: str) -> Optional[Any]:
operation = {
'operation': 'cache',
'type': 'get',
'key': key
}
response = self._send_operation(operation)
if response.get('status') == 'success':
return response['data']
return None
def wait_for_connection(self, timeout: float = 30.0) -> bool:
"""Wait for WebSocket connection to be established"""
start_time = time.time()
while not self._closing and not self.connected:
if time.time() - start_time > timeout:
print("Connection timeout exceeded")
return False
time.sleep(0.1)
return self.connected
def is_connected(self) -> bool:
"""Check if WebSocket connection is active"""
return self.connected and not self._closing
def get_connection_status(self) -> Dict[str, Any]:
"""Get detailed connection status"""
return {
"connected": self.connected,
"closing": self._closing,
"error_count": self.error_count,
"url": self.url,
"last_error_time": self.last_error_time,
"loaded_models": list(self.resource_monitor['loaded_models'])
}
def start_inference(self, model_name: str, input_data: np.ndarray) -> Optional[Dict[str, Any]]:
"""Start inference with a loaded model"""
try:
if not self.is_model_loaded(model_name):
print(f"Model {model_name} not loaded. Please load the model first.")
return None
operation = {
'operation': 'inference',
'type': 'run',
'model_name': model_name,
'input_data': input_data.tolist() if isinstance(input_data, np.ndarray) else input_data
}
response = self._send_operation(operation)
if response.get('status') == 'success':
return {
'output': np.array(response['output']) if 'output' in response else None,
'metrics': response.get('metrics', {}),
'model_info': self.model_registry.get(model_name, {})
}
else:
print(f"Inference failed: {response.get('message', 'Unknown error')}")
return None
except Exception as e:
print(f"Error during inference: {str(e)}")
return None
def close(self):
"""Close WebSocket connection and cleanup resources."""
if not self._closing:
self._closing = True
if self.websocket and self._loop:
async def cleanup():
try:
# Clean up registries
with self.lock:
self.tensor_registry.clear()
self.model_registry.clear()
self.resource_monitor['vram_used'] = 0
self.resource_monitor['active_tensors'] = 0
self.resource_monitor['loaded_models'].clear()
# Notify server about cleanup
if self.connected:
try:
await self.websocket.send(json.dumps({
'operation': 'cleanup',
'type': 'full'
}))
except:
pass
await self.websocket.close()
except Exception as e:
print(f"Error during cleanup: {str(e)}")
finally:
self.connected = False
if self._loop.is_running():
self._loop.create_task(cleanup())
else:
asyncio.run(cleanup())
async def aclose(self):
"""Asynchronously close WebSocket connection."""
if not self._closing:
self._closing = True
if self.websocket:
try:
await self.websocket.close()
except:
pass
finally:
self.connected = False
def __del__(self):
"""Ensure cleanup on deletion."""
self.close()
|