chaine 3.13.1__cp312-cp312-musllinux_1_2_i686.whl

chaine/_core/crfsuite/lib/crf/src/crfsuite_internal.h

@@ -0,0 +1,233 @@
+/*
+ *      CRFsuite internal interface.
+ *
+ * Copyright (c) 2007-2010, Naoaki Okazaki
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in the
+ *       documentation and/or other materials provided with the distribution.
+ *     * Neither the names of the authors nor the names of its contributors
+ *       may be used to endorse or promote products derived from this
+ *       software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
+ * OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* $Id$ */
+
+#ifndef __CRFSUITE_INTERNAL_H__
+#define __CRFSUITE_INTERNAL_H__
+
+#include <crfsuite.h>
+#include "logging.h"
+
+enum
+{
+    FTYPE_NONE = 0, /**< Unselected. */
+    FTYPE_CRF1D,    /**< 1st-order tyad features. */
+    FTYPE_CRF1T,    /**< 1st-order triad features. */
+};
+
+enum
+{
+    TRAIN_NONE = 0,            /**< Unselected. */
+    TRAIN_LBFGS,               /**< L-BFGS batch training. */
+    TRAIN_L2SGD,               /**< Pegasos online training. */
+    TRAIN_AVERAGED_PERCEPTRON, /**< Averaged perceptron. */
+    TRAIN_PASSIVE_AGGRESSIVE,
+    TRAIN_AROW,
+};
+
+struct tag_crfsuite_train_internal;
+typedef struct tag_crfsuite_train_internal crfsuite_train_internal_t;
+
+struct tag_encoder;
+typedef struct tag_encoder encoder_t;
+
+typedef struct
+{
+    crfsuite_data_t *data;
+    int *perm;
+    int num_instances;
+} dataset_t;
+
+void dataset_init_trainset(dataset_t *ds, crfsuite_data_t *data, int holdout);
+void dataset_init_testset(dataset_t *ds, crfsuite_data_t *data, int holdout);
+void dataset_finish(dataset_t *ds);
+void dataset_shuffle(dataset_t *ds);
+crfsuite_instance_t *dataset_get(dataset_t *ds, int i);
+
+typedef void (*crfsuite_encoder_features_on_path_callback)(void *instance, int fid, floatval_t value);
+
+/**
+ * Internal data structure for
+ */
+struct tag_crfsuite_train_internal
+{
+    encoder_t *gm;             /** Interface to the graphical model. */
+    crfsuite_params_t *params; /**< Parameter interface. */
+    logging_t *lg;             /**< Logging interface. */
+    int feature_type;          /**< Feature type. */
+    int algorithm;             /**< Training algorithm. */
+};
+
+/**
+ * Interface for a graphical model.
+ */
+struct tag_encoder
+{
+    void *internal;
+
+    const floatval_t *w;
+    floatval_t scale;
+
+    dataset_t *ds;
+    const crfsuite_instance_t *inst;
+    int level;
+
+    int num_features;
+    int cap_items;
+
+    /**
+     * Exchanges options.
+     *  @param  self        The encoder instance.
+     *  @param  params      The parameter interface.
+     *  @param  mode        The direction of parameter exchange.
+     *  @return             A status code.
+     */
+    int (*exchange_options)(encoder_t *self, crfsuite_params_t *params, int mode);
+
+    /**
+     * Initializes the encoder with a training data set.
+     *  @param  self        The encoder instance.
+     *  @param  ds          The data set for training.
+     *  @param  lg          The logging interface.
+     *  @return             A status code.
+     */
+    int (*initialize)(encoder_t *self, dataset_t *ds, logging_t *lg);
+
+    /**
+     * Compute the objective value and gradients for the whole data set.
+     *  @param  self        The encoder instance.
+     *  @param  ds          The data set.
+     *  @param  w           The feature weights.
+     *  @param  f           The pointer to a floatval_t variable to which the
+     *                      objective value is stored by this function.
+     *  @param  g           The pointer to the array that receives gradients.
+     *  @return             A status code.
+     */
+    int (*objective_and_gradients_batch)(encoder_t *self, dataset_t *ds, const floatval_t *w, floatval_t *f, floatval_t *g);
+
+    int (*features_on_path)(encoder_t *self, const crfsuite_instance_t *inst, const int *path, crfsuite_encoder_features_on_path_callback func, void *instance);
+
+    /**
+     * Sets the feature weights (and their scale factor).
+     *  @param  self        The encoder instance.
+     *  @param  w           The array of feature weights.
+     *  @param  scale       The scale factor that should be applied to the
+     *                      feature weights.
+     *  @return             A status code.
+     */
+    int (*set_weights)(encoder_t *self, const floatval_t *w, floatval_t scale);
+
+    /* Instance-wise operations. */
+    int (*set_instance)(encoder_t *self, const crfsuite_instance_t *inst);
+
+    /* Level 0. */
+
+    /* Level 1 (feature weights). */
+    int (*score)(encoder_t *self, const int *path, floatval_t *ptr_score);
+    int (*viterbi)(encoder_t *self, int *path, floatval_t *ptr_score);
+
+    /* Level 2 (forward-backward). */
+    int (*partition_factor)(encoder_t *self, floatval_t *ptr_pf);
+
+    /* Level 3 (marginals). */
+    int (*objective_and_gradients)(encoder_t *self, floatval_t *f, floatval_t *g, floatval_t gain, floatval_t weight);
+
+    int (*save_model)(encoder_t *self, const char *filename, const floatval_t *w, logging_t *lg);
+
+    void (*release)(encoder_t *self);
+};
+
+/**
+ * \defgroup crf1d_encode.c
+ */
+/** @{ */
+
+encoder_t *crf1d_create_encoder();
+
+/** @} */
+
+void holdout_evaluation(
+    encoder_t *gm,
+    dataset_t *testset,
+    const floatval_t *w,
+    logging_t *lg);
+
+int crfsuite_train_lbfgs(
+    encoder_t *gm,
+    dataset_t *trainset,
+    dataset_t *testset,
+    crfsuite_params_t *params,
+    logging_t *lg,
+    floatval_t **ptr_w);
+
+void crfsuite_train_lbfgs_init(crfsuite_params_t *params);
+
+void crfsuite_train_averaged_perceptron_init(crfsuite_params_t *params);
+
+int crfsuite_train_averaged_perceptron(
+    encoder_t *gm,
+    dataset_t *trainset,
+    dataset_t *testset,
+    crfsuite_params_t *params,
+    logging_t *lg,
+    floatval_t **ptr_w);
+
+void crfsuite_train_l2sgd_init(crfsuite_params_t *params);
+
+int crfsuite_train_l2sgd(
+    encoder_t *gm,
+    dataset_t *trainset,
+    dataset_t *testset,
+    crfsuite_params_t *params,
+    logging_t *lg,
+    floatval_t **ptr_w);
+
+void crfsuite_train_passive_aggressive_init(crfsuite_params_t *params);
+
+int crfsuite_train_passive_aggressive(
+    encoder_t *gm,
+    dataset_t *trainset,
+    dataset_t *testset,
+    crfsuite_params_t *params,
+    logging_t *lg,
+    floatval_t **ptr_w);
+
+void crfsuite_train_arow_init(crfsuite_params_t *params);
+
+int crfsuite_train_arow(
+    encoder_t *gm,
+    dataset_t *trainset,
+    dataset_t *testset,
+    crfsuite_params_t *params,
+    logging_t *lg,
+    floatval_t **ptr_w);
+
+#endif /*__CRFSUITE_INTERNAL_H__*/

chaine/_core/crfsuite/lib/crf/src/crfsuite_train.c

@@ -0,0 +1,302 @@
+/*
+ *      Implementation of the training interface (crfsuite_trainer_t).
+ *
+ * Copyright (c) 2007-2010, Naoaki Okazaki
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in the
+ *       documentation and/or other materials provided with the distribution.
+ *     * Neither the names of the authors nor the names of its contributors
+ *       may be used to endorse or promote products derived from this
+ *       software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
+ * OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* $Id$ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif /*HAVE_CONFIG_H*/
+
+#include <os.h>
+
+#include <stdlib.h>
+#include <string.h>
+
+#include <crfsuite.h>
+#include "crfsuite_internal.h"
+#include "params.h"
+#include "logging.h"
+#include "crf1d.h"
+
+static crfsuite_train_internal_t *crfsuite_train_new(int ftype, int algorithm)
+{
+    crfsuite_train_internal_t *tr = (crfsuite_train_internal_t *)calloc(1, sizeof(crfsuite_train_internal_t));
+    if (tr != NULL)
+    {
+        tr->lg = (logging_t *)calloc(1, sizeof(logging_t));
+        tr->params = params_create_instance();
+        tr->feature_type = ftype;
+        tr->algorithm = algorithm;
+
+        tr->gm = crf1d_create_encoder();
+        tr->gm->exchange_options(tr->gm, tr->params, 0);
+
+        /* Initialize parameters for the training algorithm. */
+        switch (algorithm)
+        {
+        case TRAIN_LBFGS:
+            crfsuite_train_lbfgs_init(tr->params);
+            break;
+        case TRAIN_L2SGD:
+            crfsuite_train_l2sgd_init(tr->params);
+            break;
+        case TRAIN_AVERAGED_PERCEPTRON:
+            crfsuite_train_averaged_perceptron_init(tr->params);
+            break;
+        case TRAIN_PASSIVE_AGGRESSIVE:
+            crfsuite_train_passive_aggressive_init(tr->params);
+            break;
+        case TRAIN_AROW:
+            crfsuite_train_arow_init(tr->params);
+            break;
+        }
+    }
+
+    return tr;
+}
+
+static void crfsuite_train_delete(crfsuite_trainer_t *self)
+{
+    crfsuite_train_internal_t *tr = (crfsuite_train_internal_t *)self->internal;
+    if (tr != NULL)
+    {
+        if (tr->gm != NULL)
+        {
+            tr->gm->release(tr->gm);
+        }
+        if (tr->params != NULL)
+        {
+            tr->params->release(tr->params);
+        }
+        free(tr->lg);
+        free(tr);
+    }
+    free(self);
+}
+
+static int crfsuite_train_addref(crfsuite_trainer_t *tr)
+{
+    return crfsuite_interlocked_increment(&tr->nref);
+}
+
+static int crfsuite_train_release(crfsuite_trainer_t *self)
+{
+    int count = crfsuite_interlocked_decrement(&self->nref);
+    if (count == 0)
+    {
+        crfsuite_train_delete(self);
+    }
+    return count;
+}
+
+static void crfsuite_train_set_message_callback(crfsuite_trainer_t *self, void *instance, crfsuite_logging_callback cbm)
+{
+    crfsuite_train_internal_t *tr = (crfsuite_train_internal_t *)self->internal;
+    tr->lg->func = cbm;
+    tr->lg->instance = instance;
+}
+
+static crfsuite_params_t *crfsuite_train_params(crfsuite_trainer_t *self)
+{
+    crfsuite_train_internal_t *tr = (crfsuite_train_internal_t *)self->internal;
+    crfsuite_params_t *params = tr->params;
+    params->addref(params);
+    return params;
+}
+
+static int crfsuite_train_train(
+    crfsuite_trainer_t *self,
+    const crfsuite_data_t *data,
+    const char *filename,
+    int holdout)
+{
+    char *algorithm = NULL;
+    crfsuite_train_internal_t *tr = (crfsuite_train_internal_t *)self->internal;
+    logging_t *lg = tr->lg;
+    encoder_t *gm = tr->gm;
+    floatval_t *w = NULL;
+    dataset_t trainset;
+    dataset_t testset;
+
+    /* Prepare the data set(s) for training (and holdout evaluation). */
+    dataset_init_trainset(&trainset, (crfsuite_data_t *)data, holdout);
+    if (0 <= holdout)
+    {
+        dataset_init_testset(&testset, (crfsuite_data_t *)data, holdout);
+    }
+
+    /* Set the training set to the CRF, and generate features. */
+    gm->exchange_options(gm, tr->params, -1);
+    gm->initialize(gm, &trainset, lg);
+
+    /* Call the training algorithm. */
+    switch (tr->algorithm)
+    {
+    case TRAIN_LBFGS:
+        crfsuite_train_lbfgs(
+            gm,
+            &trainset,
+            (holdout != -1 ? &testset : NULL),
+            tr->params,
+            lg,
+            &w);
+        break;
+    case TRAIN_L2SGD:
+        crfsuite_train_l2sgd(
+            gm,
+            &trainset,
+            (holdout != -1 ? &testset : NULL),
+            tr->params,
+            lg,
+            &w);
+        break;
+    case TRAIN_AVERAGED_PERCEPTRON:
+        crfsuite_train_averaged_perceptron(
+            gm,
+            &trainset,
+            (holdout != -1 ? &testset : NULL),
+            tr->params,
+            lg,
+            &w);
+        break;
+    case TRAIN_PASSIVE_AGGRESSIVE:
+        crfsuite_train_passive_aggressive(
+            gm,
+            &trainset,
+            (holdout != -1 ? &testset : NULL),
+            tr->params,
+            lg,
+            &w);
+        break;
+    case TRAIN_AROW:
+        crfsuite_train_arow(
+            gm,
+            &trainset,
+            (holdout != -1 ? &testset : NULL),
+            tr->params,
+            lg,
+            &w);
+        break;
+    }
+
+    /* Store the model file. */
+    if (filename != NULL && *filename != '\0')
+    {
+        gm->save_model(gm, filename, w, lg);
+    }
+
+    if (0 <= holdout)
+    {
+        dataset_finish(&testset);
+    }
+    dataset_finish(&trainset);
+    free(w);
+
+    return 0;
+}
+
+int crf1de_create_instance(const char *interface, void **ptr)
+{
+    int ftype = FTYPE_NONE;
+    int algorithm = TRAIN_NONE;
+
+    /* Check if the interface name begins with "train/". */
+    if (strncmp(interface, "train/", 6) != 0)
+    {
+        return 1;
+    }
+    interface += 6;
+
+    /* Obtain the feature type. */
+    if (strncmp(interface, "crf1d/", 6) == 0)
+    {
+        ftype = FTYPE_CRF1D;
+        interface += 6;
+    }
+    else
+    {
+        return 1;
+    }
+
+    /* Obtain the training algorithm. */
+    if (strcmp(interface, "lbfgs") == 0)
+    {
+        algorithm = TRAIN_LBFGS;
+    }
+    else if (strcmp(interface, "l2sgd") == 0)
+    {
+        algorithm = TRAIN_L2SGD;
+    }
+    else if (strcmp(interface, "averaged-perceptron") == 0)
+    {
+        algorithm = TRAIN_AVERAGED_PERCEPTRON;
+    }
+    else if (strcmp(interface, "passive-aggressive") == 0)
+    {
+        algorithm = TRAIN_PASSIVE_AGGRESSIVE;
+    }
+    else if (strcmp(interface, "arow") == 0)
+    {
+        algorithm = TRAIN_AROW;
+    }
+    else
+    {
+        return 1;
+    }
+
+    /* Create an instance. */
+    if (ftype != FTYPE_NONE && algorithm != TRAIN_NONE)
+    {
+        crfsuite_trainer_t *trainer = (crfsuite_trainer_t *)calloc(1, sizeof(crfsuite_trainer_t));
+        if (trainer != NULL)
+        {
+            trainer->internal = crfsuite_train_new(ftype, algorithm);
+            if (trainer->internal != NULL)
+            {
+                trainer->nref = 1;
+                trainer->addref = crfsuite_train_addref;
+                trainer->release = crfsuite_train_release;
+                trainer->params = crfsuite_train_params;
+                trainer->set_message_callback = crfsuite_train_set_message_callback;
+                trainer->train = crfsuite_train_train;
+
+                *ptr = trainer;
+                return 0;
+            }
+            else
+            {
+                free(trainer);
+                trainer = NULL;
+            }
+        }
+    }
+
+    return 1;
+}

chaine/_core/crfsuite/lib/crf/src/dataset.c

@@ -0,0 +1,115 @@
+/*
+ *      Implementation for data sets (dataset_t).
+ *
+ * Copyright (c) 2007-2010, Naoaki Okazaki
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in the
+ *       documentation and/or other materials provided with the distribution.
+ *     * Neither the names of the authors nor the names of its contributors
+ *       may be used to endorse or promote products derived from this
+ *       software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
+ * OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* $Id$ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif /*HAVE_CONFIG_H*/
+
+#include <os.h>
+
+#include <stdlib.h>
+#include <crfsuite.h>
+#include "crfsuite_internal.h"
+
+void dataset_init_trainset(dataset_t *ds, crfsuite_data_t *data, int holdout)
+{
+    int i, n = 0;
+
+    for (i = 0; i < data->num_instances; ++i)
+    {
+        if (data->instances[i].group != holdout)
+        {
+            ++n;
+        }
+    }
+
+    ds->data = data;
+    ds->num_instances = n;
+    ds->perm = (int *)malloc(sizeof(int) * n);
+
+    n = 0;
+    for (i = 0; i < data->num_instances; ++i)
+    {
+        if (data->instances[i].group != holdout)
+        {
+            ds->perm[n++] = i;
+        }
+    }
+}
+
+void dataset_init_testset(dataset_t *ds, crfsuite_data_t *data, int holdout)
+{
+    int i, n = 0;
+
+    for (i = 0; i < data->num_instances; ++i)
+    {
+        if (data->instances[i].group == holdout)
+        {
+            ++n;
+        }
+    }
+
+    ds->data = data;
+    ds->num_instances = n;
+    ds->perm = (int *)malloc(sizeof(int) * n);
+
+    n = 0;
+    for (i = 0; i < data->num_instances; ++i)
+    {
+        if (data->instances[i].group == holdout)
+        {
+            ds->perm[n++] = i;
+        }
+    }
+}
+
+void dataset_finish(dataset_t *ds)
+{
+    free(ds->perm);
+}
+
+void dataset_shuffle(dataset_t *ds)
+{
+    int i;
+    for (i = 0; i < ds->num_instances; ++i)
+    {
+        int j = rand() % ds->num_instances;
+        int tmp = ds->perm[j];
+        ds->perm[j] = ds->perm[i];
+        ds->perm[i] = tmp;
+    }
+}
+
+crfsuite_instance_t *dataset_get(dataset_t *ds, int i)
+{
+    return &ds->data->instances[ds->perm[i]];
+}