train_utils.py 18.1 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
# @author lucasmiranda42
# encoding: utf-8
# module deepof

"""

Simple utility functions used in deepof example scripts. These are not part of the main package

"""

11
from datetime import date, datetime
12
from kerastuner import BayesianOptimization, Hyperband
13
from kerastuner import HyperParameters
14
from kerastuner_tensorboard_logger import TensorBoardLogger
lucas_miranda's avatar
lucas_miranda committed
15
from sklearn.metrics import roc_auc_score
16
from tensorboard.plugins.hparams import api as hp
17
from typing import Tuple, Union, Any, List
18
19
20
21
22
23
24
import deepof.hypermodels
import deepof.model_utils
import numpy as np
import os
import pickle
import tensorflow as tf

25
26
27
28
29
# Ignore warning with no downstream effect
tf.get_logger().setLevel("ERROR")
tf.autograph.set_verbosity(0)


30
class CustomStopper(tf.keras.callbacks.EarlyStopping):
31
    """ Custom early stopping callback. Prevents the model from stopping before warmup is over """
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

    def __init__(self, start_epoch, *args, **kwargs):
        super(CustomStopper, self).__init__(*args, **kwargs)
        self.start_epoch = start_epoch

    def get_config(self):  # pragma: no cover
        """Updates callback metadata"""

        config = super().get_config().copy()
        config.update({"start_epoch": self.start_epoch})
        return config

    def on_epoch_end(self, epoch, logs=None):
        if epoch > self.start_epoch:
            super().on_epoch_end(epoch, logs)


49
50
51
52
53
def load_treatments(train_path):
    """Loads a dictionary containing the treatments per individual,
    to be loaded as metadata in the coordinates class"""
    try:
        with open(
54
55
56
57
58
            os.path.join(
                train_path,
                [i for i in os.listdir(train_path) if i.endswith(".pkl")][0],
            ),
            "rb",
59
60
61
62
63
64
65
66
67
        ) as handle:
            treatment_dict = pickle.load(handle)
    except IndexError:
        treatment_dict = None

    return treatment_dict


def get_callbacks(
68
69
70
71
72
73
74
75
76
77
78
    X_train: np.array,
    batch_size: int,
    variational: bool,
    phenotype_class: float,
    predictor: float,
    loss: str,
    X_val: np.array = None,
    cp: bool = False,
    reg_cat_clusters: bool = False,
    reg_cluster_variance: bool = False,
    entropy_samples: int = 15000,
79
    entropy_knn: int = 100,
80
81
    logparam: dict = None,
    outpath: str = ".",
82
) -> List[Union[Any]]:
83
    """Generates callbacks for model training, including:
84
85
86
87
    - run_ID: run name, with coarse parameter details;
    - tensorboard_callback: for real-time visualization;
    - cp_callback: for checkpoint saving,
    - onecycle: for learning rate scheduling"""
88

89
90
91
92
93
94
95
96
    latreg = "none"
    if reg_cat_clusters and not reg_cluster_variance:
        latreg = "categorical"
    elif reg_cluster_variance and not reg_cat_clusters:
        latreg = "variance"
    elif reg_cat_clusters and reg_cluster_variance:
        latreg = "categorical+variance"

97
    run_ID = "{}{}{}{}{}{}{}_{}".format(
98
        ("GMVAE" if variational else "AE"),
lucas_miranda's avatar
lucas_miranda committed
99
        ("_Pred={}".format(predictor) if predictor > 0 and variational else ""),
100
        ("_Pheno={}".format(phenotype_class) if phenotype_class > 0 else ""),
101
        ("_loss={}".format(loss) if variational else ""),
102
103
        ("_encoding={}".format(logparam["encoding"]) if logparam is not None else ""),
        ("_k={}".format(logparam["k"]) if logparam is not None else ""),
104
        ("_latreg={}".format(latreg)),
105
        ("entknn={}".format(entropy_knn)),
106
        (datetime.now().strftime("%Y%m%d-%H%M%S")),
107
108
    )

109
    log_dir = os.path.abspath(os.path.join(outpath, "fit", run_ID))
110
    tensorboard_callback = tf.keras.callbacks.TensorBoard(
111
112
113
        log_dir=log_dir,
        histogram_freq=1,
        profile_batch=2,
114
115
    )

116
    entropy = deepof.model_utils.neighbor_latent_entropy(
117
        encoding_dim=logparam["encoding"],
118
        k=entropy_knn,
119
        samples=entropy_samples,
lucas_miranda's avatar
lucas_miranda committed
120
        validation_data=X_val,
121
        log_dir=os.path.join(outpath, "metrics", run_ID),
122
        variational=variational,
lucas_miranda's avatar
lucas_miranda committed
123
124
    )

125
    onecycle = deepof.model_utils.one_cycle_scheduler(
126
127
        X_train.shape[0] // batch_size * 250,
        max_rate=0.005,
128
        log_dir=os.path.join(outpath, "metrics", run_ID),
129
130
    )

131
    callbacks = [run_ID, tensorboard_callback, entropy, onecycle]
132
133
134

    if cp:
        cp_callback = tf.keras.callbacks.ModelCheckpoint(
135
            os.path.join(outpath, "checkpoints", run_ID + "/cp-{epoch:04d}.ckpt"),
136
137
138
139
140
141
142
143
            verbose=1,
            save_best_only=False,
            save_weights_only=True,
            save_freq="epoch",
        )
        callbacks.append(cp_callback)

    return callbacks
144
145


lucas_miranda's avatar
lucas_miranda committed
146
def log_hyperparameters(phenotype_class: float, rec: str):
lucas_miranda's avatar
lucas_miranda committed
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
    """Blueprint for hyperparameter and metric logging in tensorboard during hyperparameter tuning"""

    logparams = [
        hp.HParam(
            "encoding",
            hp.Discrete([2, 4, 6, 8, 12, 16]),
            display_name="encoding",
            description="encoding size dimensionality",
        ),
        hp.HParam(
            "k",
            hp.IntInterval(min_value=1, max_value=25),
            display_name="k",
            description="cluster_number",
        ),
        hp.HParam(
            "loss",
            hp.Discrete(["ELBO", "MMD", "ELBO+MMD"]),
            display_name="loss function",
            description="loss function",
        ),
    ]

    metrics = [
        hp.Metric("val_{}mae".format(rec), display_name="val_{}mae".format(rec)),
        hp.Metric("val_{}mse".format(rec), display_name="val_{}mse".format(rec)),
    ]
    if phenotype_class:
        logparams.append(
            hp.HParam(
                "pheno_weight",
                hp.RealInterval(min_value=0.0, max_value=1000.0),
                display_name="pheno weight",
                description="weight applied to phenotypic classifier from the latent space",
            )
        )
        metrics += [
            hp.Metric(
                "phenotype_prediction_accuracy",
                display_name="phenotype_prediction_accuracy",
            ),
            hp.Metric(
                "phenotype_prediction_auc",
                display_name="phenotype_prediction_auc",
            ),
        ]

    return logparams, metrics


# noinspection PyUnboundLocalVariable
lucas_miranda's avatar
lucas_miranda committed
198
def tensorboard_metric_logging(
199
200
201
202
203
204
205
206
    run_dir: str,
    hpms: Any,
    ae: Any,
    X_val: np.ndarray,
    y_val: np.ndarray,
    phenotype_class: float,
    predictor: float,
    rec: str,
lucas_miranda's avatar
lucas_miranda committed
207
):
lucas_miranda's avatar
lucas_miranda committed
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
    """Autoencoder metric logging in tensorboard"""

    output = ae.predict(X_val)
    if phenotype_class or predictor:
        reconstruction = output[0]
        prediction = output[1]
        pheno = output[-1]
    else:
        reconstruction = output

    with tf.summary.create_file_writer(run_dir).as_default():
        hp.hparams(hpms)  # record the values used in this trial
        val_mae = tf.reduce_mean(
            tf.keras.metrics.mean_absolute_error(X_val, reconstruction)
        )
        val_mse = tf.reduce_mean(
            tf.keras.metrics.mean_squared_error(X_val, reconstruction)
        )
        tf.summary.scalar("val_{}mae".format(rec), val_mae, step=1)
        tf.summary.scalar("val_{}mse".format(rec), val_mse, step=1)

        if predictor:
            pred_mae = tf.reduce_mean(
                tf.keras.metrics.mean_absolute_error(X_val, prediction)
            )
            pred_mse = tf.reduce_mean(
                tf.keras.metrics.mean_squared_error(X_val, prediction)
            )
            tf.summary.scalar("val_prediction_mae".format(rec), pred_mae, step=1)
            tf.summary.scalar("val_prediction_mse".format(rec), pred_mse, step=1)

        if phenotype_class:
            pheno_acc = tf.keras.metrics.binary_accuracy(y_val, tf.squeeze(pheno))
            pheno_auc = roc_auc_score(y_val, pheno)

            tf.summary.scalar("phenotype_prediction_accuracy", pheno_acc, step=1)
            tf.summary.scalar("phenotype_prediction_auc", pheno_auc, step=1)


247
def autoencoder_fitting(
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
    preprocessed_object: Tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray],
    batch_size: int,
    encoding_size: int,
    epochs: int,
    hparams: dict,
    kl_warmup: int,
    log_history: bool,
    log_hparams: bool,
    loss: str,
    mmd_warmup: int,
    montecarlo_kl: int,
    n_components: int,
    output_path: str,
    phenotype_class: float,
    predictor: float,
    pretrained: str,
    save_checkpoints: bool,
    save_weights: bool,
    variational: bool,
    reg_cat_clusters: bool,
    reg_cluster_variance: bool,
    entropy_samples: int,
270
    entropy_knn: int,
271
):
272
273
    """Implementation function for deepof.data.coordinates.deep_unsupervised_embedding"""

274
    # Load data
275
276
277
278
279
    X_train, y_train, X_val, y_val = preprocessed_object

    # To avoid stability issues
    tf.keras.backend.clear_session()

280
    # Defines what to log on tensorboard (useful for trying out different models)
281
282
    logparam = {
        "encoding": encoding_size,
283
        "k": n_components,
284
285
286
287
288
        "loss": loss,
    }
    if phenotype_class:
        logparam["pheno_weight"] = phenotype_class

289
    # Load callbacks
290
    run_ID, *cbacks = get_callbacks(
291
        X_train=X_train,
lucas_miranda's avatar
lucas_miranda committed
292
        X_val=(X_val if X_val.shape != (0,) else None),
293
294
295
        batch_size=batch_size,
        cp=save_checkpoints,
        variational=variational,
296
        phenotype_class=phenotype_class,
297
298
        predictor=predictor,
        loss=loss,
299
        entropy_samples=entropy_samples,
300
        entropy_knn=entropy_knn,
301
        reg_cat_clusters=reg_cat_clusters,
302
        reg_cluster_variance=reg_cluster_variance,
303
304
305
        logparam=logparam,
        outpath=output_path,
    )
306
307
    if not log_history:
        cbacks = cbacks[1:]
308

309
    # Logs hyperparameters to tensorboard
lucas_miranda's avatar
lucas_miranda committed
310
    rec = "reconstruction_" if phenotype_class else ""
311
    if log_hparams:
lucas_miranda's avatar
lucas_miranda committed
312
        logparams, metrics = log_hyperparameters(phenotype_class, rec)
313
314

        with tf.summary.create_file_writer(
315
            os.path.join(output_path, "hparams", run_ID)
316
317
318
319
320
        ).as_default():
            hp.hparams_config(
                hparams=logparams,
                metrics=metrics,
            )
321

322
    # Build models
323
324
325
326
327
328
329
    if not variational:
        encoder, decoder, ae = deepof.models.SEQ_2_SEQ_AE(
            ({} if hparams is None else hparams)
        ).build(X_train.shape)
        return_list = (encoder, decoder, ae)

    else:
330
        (encoder, generator, grouper, ae,) = deepof.models.SEQ_2_SEQ_GMVAE(
331
332
333
334
335
336
337
338
339
340
341
342
343
            architecture_hparams=({} if hparams is None else hparams),
            batch_size=batch_size,
            compile_model=True,
            encoding=encoding_size,
            kl_warmup_epochs=kl_warmup,
            loss=loss,
            mmd_warmup_epochs=mmd_warmup,
            montecarlo_kl=montecarlo_kl,
            neuron_control=False,
            number_of_components=n_components,
            overlap_loss=False,
            phenotype_prediction=phenotype_class,
            predictor=predictor,
344
345
            reg_cat_clusters=reg_cat_clusters,
            reg_cluster_variance=reg_cluster_variance,
346
        ).build(X_train.shape)
347
348
349
        return_list = (encoder, generator, grouper, ae)

    if pretrained:
350
        # If pretrained models are specified, load weights and return
351
352
353
354
355
356
357
358
359
        ae.load_weights(pretrained)
        return return_list

    else:
        if not variational:

            ae.fit(
                x=X_train,
                y=X_train,
360
                epochs=epochs,
361
362
363
                batch_size=batch_size,
                verbose=1,
                validation_data=(X_val, X_val),
364
                callbacks=cbacks
365
366
367
368
369
370
371
372
                + [
                    CustomStopper(
                        monitor="val_loss",
                        patience=5,
                        restore_best_weights=True,
                        start_epoch=max(kl_warmup, mmd_warmup),
                    ),
                ],
373
374
            )

375
376
377
            if save_weights:
                ae.save_weights("{}_final_weights.h5".format(run_ID))

378
379
        else:

380
            callbacks_ = cbacks + [
381
382
383
384
                CustomStopper(
                    monitor="val_loss",
                    patience=5,
                    restore_best_weights=True,
385
                    start_epoch=max(kl_warmup, mmd_warmup),
386
387
388
                ),
            ]

389
390
391
392
393
394
395
            Xs, ys = [X_train], [X_train]
            Xvals, yvals = [X_val], [X_val]

            if predictor > 0.0:
                Xs, ys = X_train[:-1], [X_train[:-1], X_train[1:]]
                Xvals, yvals = X_val[:-1], [X_val[:-1], X_val[1:]]

396
            if phenotype_class > 0.0:
397
398
399
                ys += [y_train]
                yvals += [y_val]

400
            ae.fit(
401
402
                x=Xs,
                y=ys,
403
                epochs=epochs,
404
405
406
407
408
409
410
411
412
                batch_size=batch_size,
                verbose=1,
                validation_data=(
                    Xvals,
                    yvals,
                ),
                callbacks=callbacks_,
            )

413
            if not os.path.exists(os.path.join(output_path, "trained_weights")):
414
415
                os.makedirs("trained_weights")

416
            if save_weights:
417
418
                ae.save_weights(
                    os.path.join(
419
420
421
                        "{}".format(output_path),
                        "trained_weights",
                        "{}_final_weights.h5".format(run_ID),
422
423
                    )
                )
424

425
426
427
428
429
            if log_hparams:
                # Logparams to tensorboard
                tensorboard_metric_logging(
                    os.path.join(output_path, "hparams", run_ID),
                    logparam,
lucas_miranda's avatar
lucas_miranda committed
430
                    ae,
lucas_miranda's avatar
lucas_miranda committed
431
432
                    Xvals,
                    yvals[-1],
lucas_miranda's avatar
lucas_miranda committed
433
434
435
                    phenotype_class,
                    predictor,
                    rec,
436
                )
437

438
439
440
    return return_list


441
def tune_search(
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
    data: List[np.array],
    encoding_size: int,
    hypertun_trials: int,
    hpt_type: str,
    hypermodel: str,
    k: int,
    kl_warmup_epochs: int,
    loss: str,
    mmd_warmup_epochs: int,
    overlap_loss: float,
    phenotype_class: float,
    predictor: float,
    project_name: str,
    callbacks: List,
    n_epochs: int = 30,
    n_replicas: int = 1,
    outpath: str = ".",
459
) -> Union[bool, Tuple[Any, Any]]:
460
461
    """Define the search space using keras-tuner and bayesian optimization

462
463
464
465
466
467
468
469
470
471
472
    Parameters:
        - train (np.array): dataset to train the model on
        - test (np.array): dataset to validate the model on
        - hypertun_trials (int): number of Bayesian optimization iterations to run
        - hpt_type (str): specify one of Bayesian Optimization (bayopt) and Hyperband (hyperband)
        - hypermodel (str): hypermodel to load. Must be one of S2SAE (plain autoencoder)
        or S2SGMVAE (Gaussian Mixture Variational autoencoder).
        - k (int) number of components of the Gaussian Mixture
        - loss (str): one of [ELBO, MMD, ELBO+MMD]
        - overlap_loss (float): assigns as weight to an extra loss term which
        penalizes overlap between GM components
473
        - phenotype_class (float): adds an extra regularizing neural network to the model,
474
475
476
477
478
479
480
481
482
483
484
485
        which tries to predict the phenotype of the animal from which the sequence comes
        - predictor (float): adds an extra regularizing neural network to the model,
        which tries to predict the next frame from the current one
        - project_name (str): ID of the current run
        - callbacks (list): list of callbacks for the training loop
        - n_epochs (int): optional. Number of epochs to train each run for
        - n_replicas (int): optional. Number of replicas per parameter set. Higher values
         will yield more robust results, but will affect performance severely

    Returns:
        - best_hparams (dict): dictionary with the best retrieved hyperparameters
        - best_run (tf.keras.Model): trained instance of the best model found
486
487
488

    """

489
490
    X_train, y_train, X_val, y_val = data

491
492
493
494
    assert hpt_type in ["bayopt", "hyperband"], (
        "Invalid hyperparameter tuning framework. " "Select one of bayopt and hyperband"
    )

lucas_miranda's avatar
lucas_miranda committed
495
    if hypermodel == "S2SAE":  # pragma: no cover
496
        assert (
497
            predictor == 0.0 and phenotype_class == 0.0
498
        ), "Prediction branches are only available for variational models. See documentation for more details"
499
        batch_size = 1
500
        hypermodel = deepof.hypermodels.SEQ_2_SEQ_AE(input_shape=X_train.shape)
501
502

    elif hypermodel == "S2SGMVAE":
503
        batch_size = 64
504
        hypermodel = deepof.hypermodels.SEQ_2_SEQ_GMVAE(
505
            input_shape=X_train.shape,
506
            encoding=encoding_size,
507
            kl_warmup_epochs=kl_warmup_epochs,
508
            loss=loss,
509
            mmd_warmup_epochs=mmd_warmup_epochs,
510
            number_of_components=k,
511
            overlap_loss=overlap_loss,
512
            phenotype_predictor=phenotype_class,
513
            predictor=predictor,
514
        )
lucas_miranda's avatar
lucas_miranda committed
515

516
517
518
    else:
        return False

519
520
521
    hpt_params = {
        "hypermodel": hypermodel,
        "executions_per_trial": n_replicas,
522
523
524
        "logger": TensorBoardLogger(
            metrics=["val_mae"], logdir=os.path.join(outpath, "logged_hparams")
        ),
525
526
527
528
529
530
531
        "objective": "val_mae",
        "project_name": project_name,
        "tune_new_entries": True,
    }

    if hpt_type == "hyperband":
        tuner = Hyperband(
532
533
534
            directory=os.path.join(
                outpath, "HyperBandx_{}_{}".format(loss, str(date.today()))
            ),
535
536
            max_epochs=35,
            hyperband_iterations=hypertun_trials,
537
            factor=3,
538
539
540
541
            **hpt_params
        )
    else:
        tuner = BayesianOptimization(
542
543
544
            directory=os.path.join(
                outpath, "BayOpt_{}_{}".format(loss, str(date.today()))
            ),
545
546
547
            max_trials=hypertun_trials,
            **hpt_params
        )
548
549
550

    print(tuner.search_space_summary())

551
552
553
554
555
556
557
    Xs, ys = [X_train], [X_train]
    Xvals, yvals = [X_val], [X_val]

    if predictor > 0.0:
        Xs, ys = X_train[:-1], [X_train[:-1], X_train[1:]]
        Xvals, yvals = X_val[:-1], [X_val[:-1], X_val[1:]]

558
    if phenotype_class > 0.0:
559
560
561
        ys += [y_train]
        yvals += [y_val]

562
    tuner.search(
563
564
        Xs,
        ys,
565
        epochs=n_epochs,
566
        validation_data=(Xvals, yvals),
567
        verbose=1,
568
        batch_size=batch_size,
lucas_miranda's avatar
lucas_miranda committed
569
        callbacks=callbacks,
570
571
572
573
574
    )

    best_hparams = tuner.get_best_hyperparameters(num_trials=1)[0]
    best_run = tuner.hypermodel.build(best_hparams)

lucas_miranda's avatar
lucas_miranda committed
575
576
    print(tuner.results_summary())

577
    return best_hparams, best_run