declearn.main.FederatedClient
Client-side Federated Learning orchestrating class.
Source code in declearn/main/_client.py
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 |
|
__init__(netwk, train_data, valid_data=None, checkpoint=None, share_metrics=True, logger=None)
Instantiate a client to participate in a federated learning task.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
netwk |
Union[NetworkClient, NetworkClientConfig, Dict[str, Any], str]
|
NetworkClient communication endpoint instance, or configuration
dict, dataclass or path to a TOML file enabling its instantiation.
In the latter three cases, the object's default logger will be set
to that of this |
required |
train_data |
Union[Dataset, str]
|
Dataset instance wrapping the training data, or path to a JSON file from which it can be instantiated. |
required |
valid_data |
Optional[Union[Dataset, str]]
|
Optional Dataset instance wrapping validation data, or
path to a JSON file from which it can be instantiated.
If None, run evaluation rounds over |
None
|
checkpoint |
Union[Checkpointer, Dict[str, Any], str, None]
|
Optional Checkpointer instance or instantiation dict to be used so as to save round-wise model, optimizer and metrics. If a single string is provided, treat it as the checkpoint folder path and use default values for other parameters. |
None
|
share_metrics |
bool
|
Whether to share evaluation metrics with the server, or save them locally and only send the model's loss. This may prevent information leakage, e.g. as to the local distribution of target labels or values. |
True
|
logger |
Union[logging.Logger, str, None]
|
Logger to use, or name of a logger to set up with
|
None
|
Source code in declearn/main/_client.py
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 |
|
async_run()
async
Participate in the federated learning process.
Note: this method is the async backend of self.run
.
Source code in declearn/main/_client.py
143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 |
|
cancel_training(message)
async
Handle a server request to cancel training.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
message |
messaging.CancelTraining
|
CancelTraining message received from the server. |
required |
Source code in declearn/main/_client.py
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 |
|
evaluation_round(message)
async
Run a local evaluation round.
If an exception is raised during the local process, wrap it as an Error message and send it to the server instead of raising it.
If a checkpointer is set, record the local loss, and the model weights received from the server.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
message |
messaging.EvaluationRequest
|
Instructions from the server regarding the evaluation round. |
required |
Source code in declearn/main/_client.py
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 |
|
handle_message(message)
async
Handle an incoming message from the server.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
message |
messaging.Message
|
Message instance that needs triage and processing. |
required |
Returns:
Name | Type | Description |
---|---|---|
exit_loop |
bool
|
Whether to interrupt the client's message-receiving loop. |
Source code in declearn/main/_client.py
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 |
|
initialize()
async
Set up a Model and an Optimizer based on server instructions.
Await server instructions (as an InitRequest message) and conduct initialization.
Raises:
Type | Description |
---|---|
RuntimeError
|
If initialization failed, either because the message was not received or was of incorrect type, or because instantiation of the objects it specifies failed. |
Returns:
Name | Type | Description |
---|---|---|
model |
Model
|
Model that is to be trained (with shared initial parameters). |
optim |
Optimizer
|
Optimizer that is to be used locally to train the model. |
Source code in declearn/main/_client.py
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 |
|
make_private(message)
Set up differentially-private training, using DP-SGD.
Based on the server message, replace the wrapped trainmanager
attribute by an instance of a subclass that provides with DP-SGD.
Note that this method triggers the import of declearn.main.privacy
which may result in an error if the third-party dependency 'opacus'
is not available.
Parameters:
message: PrivacyRequest Instructions from the server regarding the DP-SGD setup.
Source code in declearn/main/_client.py
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 |
|
register()
async
Register for participation in the federated learning process.
Raises:
Type | Description |
---|---|
RuntimeError
|
If registration has failed 10 times (with a 1 minute delay between connection and registration attempts). |
Source code in declearn/main/_client.py
194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 |
|
run()
Participate in the federated learning process.
- Connect to the orchestrating
FederatedServer
and register for training, sharing some metadata aboutself.train_data
. - Await initialization instructions to spawn the Model that is to be trained and the local Optimizer used to do so.
- Participate in training and evaluation rounds based on the server's requests, checkpointing the model and local loss.
- Expect instructions to stop training, or to cancel it in case errors are reported during the process.
Source code in declearn/main/_client.py
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 |
|
stop_training(message)
async
Handle a server request to stop training.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
message |
messaging.StopTraining
|
StopTraining message received from the server. |
required |
Source code in declearn/main/_client.py
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 |
|
training_round(message)
async
Run a local training round.
If an exception is raised during the local process, wrap it as an Error message and send it to the server instead of raising it.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
message |
messaging.TrainRequest
|
Instructions from the server regarding the training round. |
required |
Source code in declearn/main/_client.py
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 |
|