backward_learner
Bases: learner
The backward learner defined based on the error back propagation algorithm.
Attributes:
Name | Type | Description |
---|---|---|
n_epochs |
int, default = 100
|
Number of training epochs in the backward learner. |
loss |
Module
|
The loss function for RPN prediction evaluation. |
optimizer |
Optimizer
|
The optimizer for parameter gradient calculation |
lr_scheduler |
LRScheduler
|
The learning rate scheduler of the optimizer. |
Methods:
Name | Description |
---|---|
__init__ |
The backward learner initialization method. |
train |
The training method of the backward learner. |
test |
The testing method of the backward learner. |
Source code in tinybig/learner/backward_learner.py
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 |
|
__init__(name='error_backward_propagation_algorithm', n_epochs=100, lr_scheduler=None, optimizer=None, loss=None, lr_scheduler_configs=None, optimizer_configs=None, loss_configs=None, *args, **kwargs)
The initialization method of the backward learner.
It initializes the backward learner object, and initializes the loss function, optimizer, and lr_scheduler that will be used for the RPN model training.
Specifically, the loss function, optimizer, and lr_scheduler can be initialized with the provided parameter loss, optimizer and lr_scheduler directly. Another initialization approach is to define the corresponding configurations, and initialize them based on the configuration descriptions instead.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
name |
Name of the backward learner. |
'error_backward_propagation_algorithm'
|
|
n_epochs |
Number of training epochs in the backward learner. |
100
|
|
loss |
The loss function for RPN prediction evaluation. |
None
|
|
optimizer |
The optimizer for parameter gradient calculation |
None
|
|
lr_scheduler |
The learning rate scheduler of the optimizer. |
None
|
|
loss_configs |
The loss function configuration, which can also be used to initialize the loss function. |
None
|
|
optimizer_configs |
The optimizer configuration, which can also be used to initialize the optimizer. |
None
|
|
lr_scheduler_configs |
The configuration of the lr_scheduler, which can also be used to initialize the learning rate scheduler. |
None
|
Returns:
Type | Description |
---|---|
object
|
The backward learner object initialized with the parameters. |
Source code in tinybig/learner/backward_learner.py
test(model, test_loader, device='cpu', metric=None, return_full_result=True)
The testing method of the backward learning for RPN performance testing.
It applies the RPN model to the provided testing set, and return the generated prediction results on the testing set.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
model
|
The RPN model to be tested. |
required |
test_loader |
dataloader
|
The testing set dataloader. |
required |
device |
str
|
Device used for the testing method. |
'cpu'
|
metric |
metric
|
Evaluation metric used for evaluating the testing performance. |
None
|
return_full_result |
bool
|
The boolean tag indicating whether the full result should be returned. Since this test method will also be called in the train method for training performance display, which don't require the full testing results actually. |
True
|
Returns:
Type | Description |
---|---|
dict
|
The testing results together with testing performance records. |
Source code in tinybig/learner/backward_learner.py
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 |
|
train(model, data_loader, device='cpu', metric=None, test_check=True, disable_tqdm=False, display_step=1)
The backward learner training method for RPN model.
It trains the RPN model with the provided training dataset. Based on the provided parameters, this method will also display information about the training process for each of the epochs, like the current epochs, time cost, training loss, training scores, and testing loss and testing scores.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
model
|
The RPN model to be trained. |
required |
data_loader |
dataloader
|
The training data_loader. |
required |
device |
str
|
The device used for the model training. |
'cpu'
|
metric |
metric
|
The evaluation metric used to display the training process. |
None
|
test_check |
bool
|
Boolean tag indicating whether to display the testing performance or not during training. |
True
|
disable_tqdm |
bool
|
Boolean tag indicating whether to disable the tqdm progress bar or not. |
False
|
display_step |
int
|
How often this method will display the training progress information, e.g., display_step=10, the training information will be displayed every 10 epochs. |
1
|
Returns:
Type | Description |
---|---|
dict
|
The training record of the RPN model, covering information like the time cost, training loss, training scores, and testing loss and testing scores, etc. |
Source code in tinybig/learner/backward_learner.py
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 |
|