annif.backend.nn_ensemble.NNEnsembleBackend._fit_model() - Code Metrics - Inspection of "adapt to annif.simplemma_util and newer Connexion" - NatLibFi/Annif - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Push — issue631-rest-api-language-det... ( 34c253...1cd800 )

by Osma

created 2024-09-16 10:53 UTC

NNEnsembleBackend._fit_model() B

↳ Parent: annif.backend.nn_ensemble

Complexity

Conditions

Size

Total Lines	25
Code Lines	20

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
cc	5
eloc	20
nop	5
dl	0
loc	25
rs	8.9332
c	0
b	0
f	0

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

from __future__ import annotations

import importlib
import json
import os.path
import shutil
import zipfile
from io import BytesIO
from typing import TYPE_CHECKING, Any

import joblib
import keras.backend as K
import lmdb
import numpy as np
from keras.layers import Add, Dense, Dropout, Flatten, Input, Layer
from keras.models import Model
from keras.saving import load_model
from keras.utils import Sequence
from scipy.sparse import csc_matrix, csr_matrix

import annif.corpus
import annif.parallel
import annif.util
from annif.exception import (
    NotInitializedException,
    NotSupportedException,
    OperationFailedException,
)
from annif.suggestion import SuggestionBatch, vector_to_suggestions

from . import backend, ensemble

if TYPE_CHECKING:
    from tensorflow.python.framework.ops import EagerTensor

    from annif.corpus.document import DocumentCorpus

logger = annif.logger


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


def key_to_idx(key: memoryview | bytes) -> int:
    """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: np.ndarray, targets: np.ndarray) -> None:
        # 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: int) -> tuple[np.ndarray, np.ndarray]:
        """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) -> int:
        """return the number of available batches"""
        return int(np.ceil(self._counter / self._batch_size))


class MeanLayer(Layer):
    """Custom Keras layer that calculates mean values along the 2nd axis."""

    def call(self, inputs: EagerTensor) -> EagerTensor:
        return K.mean(inputs, axis=2)


class NNEnsembleBackend(backend.AnnifLearningBackend, ensemble.BaseEnsembleBackend):
    """Neural network ensemble backend that combines results from multiple
    projects"""

    name = "nn_ensemble"

    MODEL_FILE = "nn-model.keras"
    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 initialize(self, parallel: bool = False) -> None:
        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))
        try:
            self._model = load_model(
                model_filename, custom_objects={"MeanLayer": MeanLayer}
            )
        except Exception as err:
            metadata = self.get_model_metadata(model_filename)
            keras_version = importlib.metadata.version("keras")
            message = (
                f"loading Keras model from {model_filename}; "
                f"model metadata: {metadata}; "
                f"you have Keras version {keras_version}. "
                f'Original error message: "{err}"'
            )
            raise OperationFailedException(message, backend_id=self.backend_id)

    def _merge_source_batches(
        self,
        batch_by_source: dict[str, SuggestionBatch],
        sources: list[tuple[str, float]],
        params: dict[str, Any],
    ) -> SuggestionBatch:
        src_weight = dict(sources)
        score_vectors = np.array(
            [
                [
                    np.sqrt(suggestions.as_vector())
                    * src_weight[project_id]
                    * len(batch_by_source)
                    for suggestions in batch
                ]
                for project_id, batch in batch_by_source.items()
            ],
            dtype=np.float32,
        ).transpose(1, 2, 0)
        prediction = self._model(score_vectors).numpy()
        return SuggestionBatch.from_sequence(
            [
                vector_to_suggestions(row, limit=int(params["limit"]))
                for row in prediction
            ],
            self.project.subjects,
        )

    def _create_model(self, sources: list[tuple[str, float]]) -> None:
        self.info("creating NN ensemble model")

        inputs = Input(shape=(len(self.project.subjects), len(sources)))

        flat_input = Flatten()(inputs)
        drop_input = Dropout(rate=float(self.params["dropout_rate"]))(flat_input)
        hidden = Dense(int(self.params["nodes"]), activation="relu")(drop_input)
        drop_hidden = Dropout(rate=float(self.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])

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

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

    def _train(
        self,
        corpus: DocumentCorpus,
        params: dict[str, Any],
        jobs: int = 0,
    ) -> None:
        sources = annif.util.parse_sources(self.params["sources"])
        self._create_model(sources)
        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: DocumentCorpus,
        seq: LMDBSequence,
        n_jobs: int,
    ) -> None:
        # 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, subject_set in pool.imap_unordered(
                psmap.suggest, corpus.documents
            ):
                doc_scores = []
                for project_id, p_hits in hits.items():
                    vector = p_hits.as_vector()
                    doc_scores.append(
                        np.sqrt(vector) * sources[project_id] * len(sources)
                    )
                score_vector = np.array(doc_scores, dtype=np.float32).transpose()
                true_vector = subject_set.as_vector(len(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: DocumentCorpus,
        epochs: int,
        lmdb_map_size: int,
        n_jobs: int = 1,
    ) -> None:
        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=True, epochs=epochs)

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

    def _learn(
        self,
        corpus: DocumentCorpus,
        params: dict[str, Any],
    ) -> None:
        self.initialize()
        self._fit_model(
            corpus, int(params["learn-epochs"]), int(params["lmdb_map_size"])
        )

    def get_model_metadata(self, model_filename: str) -> dict | None:
        """Read metadata from Keras model files."""

        try:
            with zipfile.ZipFile(model_filename, "r") as zip:
                with zip.open("metadata.json") as metadata_file:
                    metadata_str = metadata_file.read().decode("utf-8")
                    metadata = json.loads(metadata_str)
                    return metadata
        except Exception:
            self.warning(f"Failed to read metadata from {model_filename}")
            return None


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

NatLibFi / Annif

Push — issue631-rest-api-language-det... ( 34c253...1cd800 )

NNEnsembleBackend._fit_model() B

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like