train_utils.py 6.69 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
# @author lucasmiranda42
# encoding: utf-8
# module deepof

"""

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

"""
from datetime import datetime

from kerastuner import BayesianOptimization
13
from kerastuner import HyperParameters
14
from kerastuner_tensorboard_logger import TensorBoardLogger
15
from typing import Tuple, Union, Any, List
16
17
18
19
20
21
22
import deepof.hypermodels
import deepof.model_utils
import numpy as np
import os
import pickle
import tensorflow as tf

23
24
hp = HyperParameters()

25

26
def load_hparams(hparams):
27
28
29
30
31
32
33
34
35
36
37
38
    """Loads hyperparameters from a custom dictionary pickled on disc.
    Thought to be used with the output of hyperparameter_tuning.py"""

    if hparams is not None:
        with open(hparams, "rb") as handle:
            hparams = pickle.load(handle)
    else:
        hparams = {
            "units_conv": 256,
            "units_lstm": 256,
            "units_dense2": 64,
            "dropout_rate": 0.25,
39
            "encoding": 16,
40
41
42
43
44
45
46
47
48
49
50
51
            "learning_rate": 1e-3,
        }
    return hparams


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(
            os.path.join(
                train_path,
52
                [i for i in os.listdir(train_path) if i.endswith(".pkl")][0],
53
54
55
56
57
58
59
60
61
62
63
            ),
            "rb",
        ) as handle:
            treatment_dict = pickle.load(handle)
    except IndexError:
        treatment_dict = None

    return treatment_dict


def get_callbacks(
64
    X_train: np.array, batch_size: int, variational: bool, predictor: float, loss: str,
65
66
67
68
69
70
71
) -> Tuple:
    """Generates callbacks for model training, including:
        - run_ID: run name, with coarse parameter details;
        - tensorboard_callback: for real-time visualization;
        - cp_callback: for checkpoint saving,
        - onecycle: for learning rate scheduling"""

72
    run_ID = "{}{}{}_{}".format(
73
74
75
76
77
78
79
        ("GMVAE" if variational else "AE"),
        ("P" if predictor > 0 and variational else ""),
        ("_loss={}".format(loss) if variational else ""),
        datetime.now().strftime("%Y%m%d-%H%M%S"),
    )

    log_dir = os.path.abspath("logs/fit/{}".format(run_ID))
80
    tensorboard_callback = tf.keras.callbacks.TensorBoard(
81
82
83
        log_dir=log_dir, histogram_freq=1, profile_batch=2,
    )

84
85
86
87
88
89
    cp_callback = tf.keras.callbacks.ModelCheckpoint(
        "./logs/checkpoints/" + run_ID + "/cp-{epoch:04d}.ckpt",
        verbose=1,
        save_best_only=False,
        save_weights_only=True,
        save_freq="epoch",
90
91
92
93
94
95
96
97
98
99
    )

    onecycle = deepof.model_utils.one_cycle_scheduler(
        X_train.shape[0] // batch_size * 250, max_rate=0.005,
    )

    return run_ID, tensorboard_callback, cp_callback, onecycle


def tune_search(
100
    data: List[np.array],
101
102
103
104
105
    bayopt_trials: int,
    hypermodel: str,
    k: int,
    loss: str,
    overlap_loss: float,
106
    pheno_class: float,
107
108
    predictor: float,
    project_name: str,
109
    callbacks: List,
110
    n_epochs: int = 40,
111
    n_replicas: int = 1,
112
) -> Union[bool, Tuple[Any, Any]]:
113
114
115
116
117
118
    """Define the search space using keras-tuner and bayesian optimization

        Parameters:
            - train (np.array): dataset to train the model on
            - test (np.array): dataset to validate the model on
            - bayopt_trials (int): number of Bayesian optimization iterations to run
119
120
            - hypermodel (str): hypermodel to load. Must be one of S2SAE (plain autoencoder)
            or S2SGMVAE (Gaussian Mixture Variational autoencoder).
121
122
123
124
            - 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
125
126
            - pheno_class (float): adds an extra regularizing neural network to the model,
            which tries to predict the phenotype of the animal from which the sequence comes
127
128
129
130
            - 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
131
132
133
            - 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
134
135
136
137
138
139
140

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

    """

141
142
    X_train, y_train, X_val, y_val = data

lucas_miranda's avatar
lucas_miranda committed
143
    if hypermodel == "S2SAE":  # pragma: no cover
144
145
146
        assert (
            predictor == 0.0 and pheno_class == 0.0
        ), "Prediction branches are only available for variational models. See documentation for more details"
147
        hypermodel = deepof.hypermodels.SEQ_2_SEQ_AE(input_shape=X_train.shape)
148
149
150

    elif hypermodel == "S2SGMVAE":
        hypermodel = deepof.hypermodels.SEQ_2_SEQ_GMVAE(
151
            input_shape=X_train.shape,
152
153
            loss=loss,
            number_of_components=k,
154
            overlap_loss=overlap_loss,
155
            predictor=predictor,
156
        )
lucas_miranda's avatar
lucas_miranda committed
157

158
159
160
161
162
    else:
        return False

    tuner = BayesianOptimization(
        hypermodel,
163
        directory="BayesianOptx",
164
        executions_per_trial=n_replicas,
165
166
        logger=TensorBoardLogger(metrics=["val_mae"], logdir="./logs/hparams"),
        max_trials=bayopt_trials,
167
168
        objective="val_mae",
        project_name=project_name,
169
        seed=42,
170
        tune_new_entries=True,
171
172
173
174
    )

    print(tuner.search_space_summary())

175
176
177
178
179
180
181
182
183
184
185
    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:]]

    if pheno_class > 0.0:
        ys += [y_train]
        yvals += [y_val]

186
    tuner.search(
187
188
        Xs,
        ys,
189
        epochs=n_epochs,
190
        validation_data=(Xvals, yvals),
191
192
        verbose=1,
        batch_size=256,
lucas_miranda's avatar
lucas_miranda committed
193
        callbacks=callbacks,
194
195
196
197
198
    )

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

lucas_miranda's avatar
lucas_miranda committed
199
200
    print(tuner.results_summary())

201
202
203
204
205
206
207
208
    return best_hparams, best_run


# TODO:
#    - load_treatments should be part of the main data module. If available in the main directory,
#    a table (preferrable in csv) should be loaded as metadata of the coordinates automatically.
#    This becomes particularly important por the supervised models that include phenotype classification
#    alongside the encoding.