annif.backend.nn_ensemble.NNEnsembleBackend.get_hp_optimizer() - Code Metrics - Inspection of "Hyperparameter optimization for NN ensemble" - NatLibFi/Annif - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Pull Request — master (#569)

by Osma

created 2022-02-11 13:50 UTC

NNEnsembleBackend.get_hp_optimizer() A

↳ Parent: annif.backend.nn_ensemble

Complexity

Conditions

Size

Total Lines	2
Code Lines	2

Duplication

Lines	2
Ratio	100 %

Importance

Changes

Metric	Value
cc	1
eloc	2
nop	3
dl	2
loc	2
rs	10
c	0
b	0
f	0

"""Neural network based ensemble backend that combines results from multiple
projects."""


from io import BytesIO
import shutil
import os.path
import numpy as np
from scipy.sparse import csr_matrix, csc_matrix
import joblib
import lmdb
from tensorflow.keras.layers import Input, Dense, Add, Flatten, Dropout, Layer
from tensorflow.keras.models import Model, load_model
from tensorflow.keras.utils import Sequence
import tensorflow.keras.backend as K
import annif.corpus
import annif.parallel
import annif.util
from annif.exception import NotInitializedException, NotSupportedException
from annif.suggestion import VectorSuggestionResult
from . import backend
from . import ensemble
from . import hyperopt


def idx_to_key(idx):
    """convert an integer index to a binary key for use in LMDB"""
    return b'%08d' % idx


def key_to_idx(key):
    """convert a binary LMDB key to an integer index"""
    return int(key)


class LMDBSequence(Sequence):

    """A sequence of samples stored in a LMDB database."""

    def __init__(self, txn, batch_size):
        self._txn = txn
        cursor = txn.cursor()
        if cursor.last():
            # Counter holds the number of samples in the database
            self._counter = key_to_idx(cursor.key()) + 1
        else:  # empty database
            self._counter = 0
        self._batch_size = batch_size

    def add_sample(self, inputs, targets):
        # use zero-padded 8-digit key
        key = idx_to_key(self._counter)
        self._counter += 1
        # convert the sample into a sparse matrix and serialize it as bytes
        sample = (csc_matrix(inputs), csr_matrix(targets))
        buf = BytesIO()
        joblib.dump(sample, buf)
        buf.seek(0)
        self._txn.put(key, buf.read())

    def __getitem__(self, idx):
        """get a particular batch of samples"""
        cursor = self._txn.cursor()
        first_key = idx * self._batch_size
        cursor.set_key(idx_to_key(first_key))
        input_arrays = []
        target_arrays = []
        for key, value in cursor.iternext():
            if key_to_idx(key) >= (first_key + self._batch_size):
                break
            input_csr, target_csr = joblib.load(BytesIO(value))
            input_arrays.append(input_csr.toarray())
            target_arrays.append(target_csr.toarray().flatten())
        return np.array(input_arrays), np.array(target_arrays)

    def __len__(self):
        """return the number of available batches"""
        return int(np.ceil(self._counter / self._batch_size))


class NNEnsembleOptimizer(hyperopt.HyperparameterOptimizer):

    """Hyperparameter optimizer for the NN ensemble backend"""

    def _prepare(self, n_jobs=1):
        sources = dict(
            annif.util.parse_sources(self._backend.params['sources']))

        # initialize the source projects before forking, to save memory
        for project_id in sources.keys():
            project = self._backend.project.registry.get_project(project_id)
            project.initialize(parallel=True)

        psmap = annif.parallel.ProjectSuggestMap(
            self._backend.project.registry,
            list(sources.keys()),
            backend_params=None,
            limit=None,
            threshold=0.0)

        jobs, pool_class = annif.parallel.get_pool(n_jobs)

        self._score_vectors = []
        self._gold_subjects = []

        with pool_class(jobs) as pool:
            for hits, uris, labels in pool.imap_unordered(
                    psmap.suggest, self._corpus.documents):
                doc_scores = []
                for project_id, p_hits in hits.items():
                    vector = p_hits.as_vector(self._backend.project.subjects)
                    doc_scores.append(np.sqrt(vector)
                                      * sources[project_id]
                                      * len(sources))
                score_vector = np.array(doc_scores,
                                        dtype=np.float32).transpose()
                subjects = annif.corpus.SubjectSet((uris, labels))
                self._score_vectors.append(score_vector)
                self._gold_subjects.append(subjects)

    def _objective(self, trial):
        sources = annif.util.parse_sources(self._backend.params['sources'])
        params = {
            'nodes': trial.suggest_int('nodes', 50, 200),
            'dropout_rate': trial.suggest_float('dropout_rate', 0.0, 0.5),
            'epochs': trial.suggest_int('epochs', 5, 20),
            'optimizer': 'adam'
        }
        model = self._backend._create_model(sources, params)

        env = self._backend._open_lmdb(True,
                                       self._backend.params['lmdb_map_size'])
        with env.begin(buffers=True) as txn:
            seq = LMDBSequence(txn, batch_size=32)
            model.fit(seq, verbose=0, epochs=params['epochs'])

        batch = annif.eval.EvaluationBatch(self._backend.project.subjects)
        for goldsubj, score_vector in zip(self._gold_subjects,
                                          self._score_vectors):

            results = model.predict(
                np.expand_dims(score_vector, 0))
            output = VectorSuggestionResult(results[0])
            batch.evaluate(output, goldsubj)
        eval_results = batch.results(metrics=[self._metric])
        return eval_results[self._metric]

    def _postprocess(self, study):
        bp = study.best_params
        lines = [
            f"nodes={bp['nodes']}",
            f"dropout_rate={bp['dropout_rate']}",
            f"epochs={bp['epochs']}"
        ]
        return hyperopt.HPRecommendation(lines=lines, score=study.best_value)


class MeanLayer(Layer):
    """Custom Keras layer that calculates mean values along the 2nd axis."""
    def call(self, inputs):
        return K.mean(inputs, axis=2)


class NNEnsembleBackend(

        backend.AnnifLearningBackend,
        ensemble.BaseEnsembleBackend,
        hyperopt.AnnifHyperoptBackend):
    """Neural network ensemble backend that combines results from multiple
    projects"""

    name = "nn_ensemble"

    MODEL_FILE = "nn-model.h5"
    LMDB_FILE = 'nn-train.mdb'

    DEFAULT_PARAMETERS = {
        'nodes': 100,
        'dropout_rate': 0.2,
        'optimizer': 'adam',
        'epochs': 10,
        'learn-epochs': 1,
        'lmdb_map_size': 1024 * 1024 * 1024
    }

    # defaults for uninitialized instances
    _model = None

    def get_hp_optimizer(self, corpus, metric):
        return NNEnsembleOptimizer(self, corpus, metric)

    def default_params(self):
        params = backend.AnnifBackend.DEFAULT_PARAMETERS.copy()
        params.update(self.DEFAULT_PARAMETERS)
        return params

    def initialize(self, parallel=False):
        super().initialize(parallel)
        if self._model is not None:
            return  # already initialized
        if parallel:
            # Don't load TF model just before parallel execution,
            # since it won't work after forking worker processes
            return
        model_filename = os.path.join(self.datadir, self.MODEL_FILE)
        if not os.path.exists(model_filename):
            raise NotInitializedException(
                'model file {} not found'.format(model_filename),
                backend_id=self.backend_id)
        self.debug('loading Keras model from {}'.format(model_filename))
        self._model = load_model(model_filename,
                                 custom_objects={'MeanLayer': MeanLayer})

    def _merge_hits_from_sources(self, hits_from_sources, params):
        score_vector = np.array([np.sqrt(hits.as_vector(subjects))
                                 * weight * len(hits_from_sources)
                                 for hits, weight, subjects
                                 in hits_from_sources],
                                dtype=np.float32)
        results = self._model.predict(
            np.expand_dims(score_vector.transpose(), 0))
        return VectorSuggestionResult(results[0])

    def _create_model(self, sources, params):
        inputs = Input(shape=(len(self.project.subjects), len(sources)))

        flat_input = Flatten()(inputs)
        drop_input = Dropout(
            rate=float(params['dropout_rate']))(flat_input)
        hidden = Dense(int(params['nodes']),
                       activation="relu")(drop_input)
        drop_hidden = Dropout(rate=float(params['dropout_rate']))(hidden)
        delta = Dense(len(self.project.subjects),
                      kernel_initializer='zeros',
                      bias_initializer='zeros')(drop_hidden)

        mean = MeanLayer()(inputs)

        predictions = Add()([mean, delta])

        model = Model(inputs=inputs, outputs=predictions)
        model.compile(optimizer=params['optimizer'],
                      loss='binary_crossentropy',
                      metrics=['top_k_categorical_accuracy'])
        if 'lr' in params:
            model.optimizer.learning_rate.assign(
                float(params['lr']))

        summary = []
        model.summary(print_fn=summary.append)
        self.debug("Created model: \n" + "\n".join(summary))
        return model

    def _train(self, corpus, params, jobs=0):
        sources = annif.util.parse_sources(params['sources'])
        self.info("creating NN ensemble model")
        self._model = self._create_model(sources, params)
        self._fit_model(
            corpus,
            epochs=int(params['epochs']),
            lmdb_map_size=int(params['lmdb_map_size']),
            n_jobs=jobs)

    def _corpus_to_vectors(self, corpus, seq, n_jobs):
        # pass corpus through all source projects
        sources = dict(
            annif.util.parse_sources(self.params['sources']))

        # initialize the source projects before forking, to save memory
        self.info(
            f"Initializing source projects: {', '.join(sources.keys())}")
        for project_id in sources.keys():
            project = self.project.registry.get_project(project_id)
            project.initialize(parallel=True)

        psmap = annif.parallel.ProjectSuggestMap(
            self.project.registry,
            list(sources.keys()),
            backend_params=None,
            limit=None,
            threshold=0.0)

        jobs, pool_class = annif.parallel.get_pool(n_jobs)

        self.info("Processing training documents...")
        with pool_class(jobs) as pool:
            for hits, uris, labels in pool.imap_unordered(
                    psmap.suggest, corpus.documents):
                doc_scores = []
                for project_id, p_hits in hits.items():
                    vector = p_hits.as_vector(self.project.subjects)
                    doc_scores.append(np.sqrt(vector)
                                      * sources[project_id]
                                      * len(sources))
                score_vector = np.array(doc_scores,
                                        dtype=np.float32).transpose()
                subjects = annif.corpus.SubjectSet((uris, labels))
                true_vector = subjects.as_vector(self.project.subjects)
                seq.add_sample(score_vector, true_vector)

    def _open_lmdb(self, cached, lmdb_map_size):
        lmdb_path = os.path.join(self.datadir, self.LMDB_FILE)
        if not cached and os.path.exists(lmdb_path):
            shutil.rmtree(lmdb_path)
        return lmdb.open(lmdb_path, map_size=lmdb_map_size, writemap=True)

    def _fit_model(self, corpus, epochs, lmdb_map_size, n_jobs=1):
        env = self._open_lmdb(corpus == 'cached', lmdb_map_size)
        if corpus != 'cached':
            if corpus.is_empty():
                raise NotSupportedException(
                    'Cannot train nn_ensemble project with no documents')
            with env.begin(write=True, buffers=True) as txn:
                seq = LMDBSequence(txn, batch_size=32)
                self._corpus_to_vectors(corpus, seq, n_jobs)
        else:
            self.info("Reusing cached training data from previous run.")
        # fit the model using a read-only view of the LMDB
        self.info("Training neural network model...")
        with env.begin(buffers=True) as txn:
            seq = LMDBSequence(txn, batch_size=32)
            self._model.fit(seq, verbose=1, epochs=epochs)

        annif.util.atomic_save(
            self._model,
            self.datadir,
            self.MODEL_FILE)

    def _learn(self, corpus, params):
        self.initialize()
        self._fit_model(
            corpus,
            int(params['learn-epochs']),
            int(params['lmdb_map_size']))


1		"""Neural network based ensemble backend that combines results from multiple
2		projects."""
3
4
5		from io import BytesIO
6		import shutil
7		import os.path
8		import numpy as np
9		from scipy.sparse import csr_matrix, csc_matrix
10		import joblib
11		import lmdb
12		from tensorflow.keras.layers import Input, Dense, Add, Flatten, Dropout, Layer
13		from tensorflow.keras.models import Model, load_model
14		from tensorflow.keras.utils import Sequence
15		import tensorflow.keras.backend as K
16		import annif.corpus
17		import annif.parallel
18		import annif.util
19		from annif.exception import NotInitializedException, NotSupportedException
20		from annif.suggestion import VectorSuggestionResult
21		from . import backend
22		from . import ensemble
23		from . import hyperopt
24
25
26		def idx_to_key(idx):
27		"""convert an integer index to a binary key for use in LMDB"""
28		return b'%08d' % idx
29
30
31		def key_to_idx(key):
32		"""convert a binary LMDB key to an integer index"""
33		return int(key)
34
35
36	View Code Duplication	class LMDBSequence(Sequence):
		0 ignored issues – show Duplication introduced 2022-01-18 08:57 UTC by Report Bug Copy Issue Report This code seems to be duplicated in your project. Loading history...
37		"""A sequence of samples stored in a LMDB database."""
38
39		def __init__(self, txn, batch_size):
40		self._txn = txn
41		cursor = txn.cursor()
42		if cursor.last():
43		# Counter holds the number of samples in the database
44		self._counter = key_to_idx(cursor.key()) + 1
45		else: # empty database
46		self._counter = 0
47		self._batch_size = batch_size
48
49		def add_sample(self, inputs, targets):
50		# use zero-padded 8-digit key
51		key = idx_to_key(self._counter)
52		self._counter += 1
53		# convert the sample into a sparse matrix and serialize it as bytes
54		sample = (csc_matrix(inputs), csr_matrix(targets))
55		buf = BytesIO()
56		joblib.dump(sample, buf)
57		buf.seek(0)
58		self._txn.put(key, buf.read())
59
60		def __getitem__(self, idx):
61		"""get a particular batch of samples"""
62		cursor = self._txn.cursor()
63		first_key = idx * self._batch_size
64		cursor.set_key(idx_to_key(first_key))
65		input_arrays = []
66		target_arrays = []
67		for key, value in cursor.iternext():
68		if key_to_idx(key) >= (first_key + self._batch_size):
69		break
70		input_csr, target_csr = joblib.load(BytesIO(value))
71		input_arrays.append(input_csr.toarray())
72		target_arrays.append(target_csr.toarray().flatten())
73		return np.array(input_arrays), np.array(target_arrays)
74
75		def __len__(self):
76		"""return the number of available batches"""
77		return int(np.ceil(self._counter / self._batch_size))
78
79
80	View Code Duplication	class NNEnsembleOptimizer(hyperopt.HyperparameterOptimizer):
		0 ignored issues – show Duplication introduced 2022-01-18 08:57 UTC by Report Bug Copy Issue Report This code seems to be duplicated in your project. Loading history...
81		"""Hyperparameter optimizer for the NN ensemble backend"""
82
83		def _prepare(self, n_jobs=1):
84		sources = dict(
85		annif.util.parse_sources(self._backend.params['sources']))
86
87		# initialize the source projects before forking, to save memory
88		for project_id in sources.keys():
89		project = self._backend.project.registry.get_project(project_id)
90		project.initialize(parallel=True)
91
92		psmap = annif.parallel.ProjectSuggestMap(
93		self._backend.project.registry,
94		list(sources.keys()),
95		backend_params=None,
96		limit=None,
97		threshold=0.0)
98
99		jobs, pool_class = annif.parallel.get_pool(n_jobs)
100
101		self._score_vectors = []
102		self._gold_subjects = []
103
104		with pool_class(jobs) as pool:
105		for hits, uris, labels in pool.imap_unordered(
106		psmap.suggest, self._corpus.documents):
107		doc_scores = []
108		for project_id, p_hits in hits.items():
109		vector = p_hits.as_vector(self._backend.project.subjects)
110		doc_scores.append(np.sqrt(vector)
111		* sources[project_id]
112		* len(sources))
113		score_vector = np.array(doc_scores,
114		dtype=np.float32).transpose()
115		subjects = annif.corpus.SubjectSet((uris, labels))
116		self._score_vectors.append(score_vector)
117		self._gold_subjects.append(subjects)
118
119		def _objective(self, trial):
120		sources = annif.util.parse_sources(self._backend.params['sources'])
121		params = {
122		'nodes': trial.suggest_int('nodes', 50, 200),
123		'dropout_rate': trial.suggest_float('dropout_rate', 0.0, 0.5),
124		'epochs': trial.suggest_int('epochs', 5, 20),
125		'optimizer': 'adam'
126		}
127		model = self._backend._create_model(sources, params)
128
129		env = self._backend._open_lmdb(True,
130		self._backend.params['lmdb_map_size'])
131		with env.begin(buffers=True) as txn:
132		seq = LMDBSequence(txn, batch_size=32)
133		model.fit(seq, verbose=0, epochs=params['epochs'])
134
135		batch = annif.eval.EvaluationBatch(self._backend.project.subjects)
136		for goldsubj, score_vector in zip(self._gold_subjects,
137		self._score_vectors):
138
139		results = model.predict(
140		np.expand_dims(score_vector, 0))
141		output = VectorSuggestionResult(results[0])
142		batch.evaluate(output, goldsubj)
143		eval_results = batch.results(metrics=[self._metric])
144		return eval_results[self._metric]
145
146		def _postprocess(self, study):
147		bp = study.best_params
148		lines = [
149		f"nodes={bp['nodes']}",
150		f"dropout_rate={bp['dropout_rate']}",
151		f"epochs={bp['epochs']}"
152		]
153		return hyperopt.HPRecommendation(lines=lines, score=study.best_value)
154
155
156		class MeanLayer(Layer):
157		"""Custom Keras layer that calculates mean values along the 2nd axis."""
158		def call(self, inputs):
159		return K.mean(inputs, axis=2)
160
161
162	View Code Duplication	class NNEnsembleBackend(
		0 ignored issues – show Duplication introduced 2022-02-11 08:55 UTC by Report Bug Copy Issue Report This code seems to be duplicated in your project. Loading history...
163		backend.AnnifLearningBackend,
164		ensemble.BaseEnsembleBackend,
165		hyperopt.AnnifHyperoptBackend):
166		"""Neural network ensemble backend that combines results from multiple
167		projects"""
168
169		name = "nn_ensemble"
170
171		MODEL_FILE = "nn-model.h5"
172		LMDB_FILE = 'nn-train.mdb'
173
174		DEFAULT_PARAMETERS = {
175		'nodes': 100,
176		'dropout_rate': 0.2,
177		'optimizer': 'adam',
178		'epochs': 10,
179		'learn-epochs': 1,
180		'lmdb_map_size': 1024 * 1024 * 1024
181		}
182
183		# defaults for uninitialized instances
184		_model = None
185
186		def get_hp_optimizer(self, corpus, metric):
187		return NNEnsembleOptimizer(self, corpus, metric)
188
189		def default_params(self):
190		params = backend.AnnifBackend.DEFAULT_PARAMETERS.copy()
191		params.update(self.DEFAULT_PARAMETERS)
192		return params
193
194		def initialize(self, parallel=False):
195		super().initialize(parallel)
196		if self._model is not None:
197		return # already initialized
198		if parallel:
199		# Don't load TF model just before parallel execution,
200		# since it won't work after forking worker processes
201		return
202		model_filename = os.path.join(self.datadir, self.MODEL_FILE)
203		if not os.path.exists(model_filename):
204		raise NotInitializedException(
205		'model file {} not found'.format(model_filename),
206		backend_id=self.backend_id)
207		self.debug('loading Keras model from {}'.format(model_filename))
208		self._model = load_model(model_filename,
209		custom_objects={'MeanLayer': MeanLayer})
210
211		def _merge_hits_from_sources(self, hits_from_sources, params):
212		score_vector = np.array([np.sqrt(hits.as_vector(subjects))
213		* weight * len(hits_from_sources)
214		for hits, weight, subjects
215		in hits_from_sources],
216		dtype=np.float32)
217		results = self._model.predict(
218		np.expand_dims(score_vector.transpose(), 0))
219		return VectorSuggestionResult(results[0])
220
221		def _create_model(self, sources, params):
222		inputs = Input(shape=(len(self.project.subjects), len(sources)))
223
224		flat_input = Flatten()(inputs)
225		drop_input = Dropout(
226		rate=float(params['dropout_rate']))(flat_input)
227		hidden = Dense(int(params['nodes']),
228		activation="relu")(drop_input)
229		drop_hidden = Dropout(rate=float(params['dropout_rate']))(hidden)
230		delta = Dense(len(self.project.subjects),
231		kernel_initializer='zeros',
232		bias_initializer='zeros')(drop_hidden)
233
234		mean = MeanLayer()(inputs)
235
236		predictions = Add()([mean, delta])
237
238		model = Model(inputs=inputs, outputs=predictions)
239		model.compile(optimizer=params['optimizer'],
240		loss='binary_crossentropy',
241		metrics=['top_k_categorical_accuracy'])
242		if 'lr' in params:
243		model.optimizer.learning_rate.assign(
244		float(params['lr']))
245
246		summary = []
247		model.summary(print_fn=summary.append)
248		self.debug("Created model: \n" + "\n".join(summary))
249		return model
250
251		def _train(self, corpus, params, jobs=0):
252		sources = annif.util.parse_sources(params['sources'])
253		self.info("creating NN ensemble model")
254		self._model = self._create_model(sources, params)
255		self._fit_model(
256		corpus,
257		epochs=int(params['epochs']),
258		lmdb_map_size=int(params['lmdb_map_size']),
259		n_jobs=jobs)
260
261		def _corpus_to_vectors(self, corpus, seq, n_jobs):
262		# pass corpus through all source projects
263		sources = dict(
264		annif.util.parse_sources(self.params['sources']))
265
266		# initialize the source projects before forking, to save memory
267		self.info(
268		f"Initializing source projects: {', '.join(sources.keys())}")
269		for project_id in sources.keys():
270		project = self.project.registry.get_project(project_id)
271		project.initialize(parallel=True)
272
273		psmap = annif.parallel.ProjectSuggestMap(
274		self.project.registry,
275		list(sources.keys()),
276		backend_params=None,
277		limit=None,
278		threshold=0.0)
279
280		jobs, pool_class = annif.parallel.get_pool(n_jobs)
281
282		self.info("Processing training documents...")
283		with pool_class(jobs) as pool:
284		for hits, uris, labels in pool.imap_unordered(
285		psmap.suggest, corpus.documents):
286		doc_scores = []
287		for project_id, p_hits in hits.items():
288		vector = p_hits.as_vector(self.project.subjects)
289		doc_scores.append(np.sqrt(vector)
290		* sources[project_id]
291		* len(sources))
292		score_vector = np.array(doc_scores,
293		dtype=np.float32).transpose()
294		subjects = annif.corpus.SubjectSet((uris, labels))
295		true_vector = subjects.as_vector(self.project.subjects)
296		seq.add_sample(score_vector, true_vector)
297
298		def _open_lmdb(self, cached, lmdb_map_size):
299		lmdb_path = os.path.join(self.datadir, self.LMDB_FILE)
300		if not cached and os.path.exists(lmdb_path):
301		shutil.rmtree(lmdb_path)
302		return lmdb.open(lmdb_path, map_size=lmdb_map_size, writemap=True)
303
304		def _fit_model(self, corpus, epochs, lmdb_map_size, n_jobs=1):
305		env = self._open_lmdb(corpus == 'cached', lmdb_map_size)
306		if corpus != 'cached':
307		if corpus.is_empty():
308		raise NotSupportedException(
309		'Cannot train nn_ensemble project with no documents')
310		with env.begin(write=True, buffers=True) as txn:
311		seq = LMDBSequence(txn, batch_size=32)
312		self._corpus_to_vectors(corpus, seq, n_jobs)
313		else:
314		self.info("Reusing cached training data from previous run.")
315		# fit the model using a read-only view of the LMDB
316		self.info("Training neural network model...")
317		with env.begin(buffers=True) as txn:
318		seq = LMDBSequence(txn, batch_size=32)
319		self._model.fit(seq, verbose=1, epochs=epochs)
320
321		annif.util.atomic_save(
322		self._model,
323		self.datadir,
324		self.MODEL_FILE)
325
326		def _learn(self, corpus, params):
327		self.initialize()
328		self._fit_model(
329		corpus,
330		int(params['learn-epochs']),
331		int(params['lmdb_map_size']))
332

NatLibFi / Annif

Pull Request — master (#569)

NNEnsembleBackend.get_hp_optimizer() A

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like