repo
stringlengths
7
90
file_url
stringlengths
81
315
file_path
stringlengths
4
228
content
stringlengths
0
32.8k
language
stringclasses
1 value
license
stringclasses
7 values
commit_sha
stringlengths
40
40
retrieved_at
stringdate
2026-01-04 14:38:15
2026-01-05 02:33:18
truncated
bool
2 classes
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperopt-sklearn/hpsklearn/tests/test_ts.py
training/helpers/hyperopt-sklearn/hpsklearn/tests/test_ts.py
""" Unit tests for time series forecast using sklearn and hyperopt In this file, a simulated time series dataset is used to demonstrate the use of hpsklearn for time series forecasting problems. More specifically, it shows: how a time series dataset can be converted into an sklearn compatible format; the use of the time series lag selector; and exogenous data for training machine learning models. Briefly, the following formula is used to generate the dataset: y[t] = a1 * y[t - 1] + a2 * y[t - 2] + b1 * X[t, 1] + b2 * X[t, 2] + b3 * X[t, 3] + c + e where y is the time series, X is an exogenous dataset, a1, a2, b1, b2, b3 and c are parameters and e is an error term with the following specifications: a1 = .666 a2 = -.333 c = -.5 b1 = 1.5 b2 = -1.5 b3 = .5 X[t, 1] ~ uniform(.5, 1.5) X[t, 2] ~ normal(2., 1.5) X[t, 3] ~ normal[3., 2.5] e ~ normal(0, 1.5) The purpose of learning is to correctly identify the lag size and the values of the parameters. """ from __future__ import print_function import sys try: import unittest2 as unittest except ImportError: import unittest import numpy as np import pandas as pd import warnings warnings.filterwarnings('ignore') from hyperopt import tpe from hpsklearn import HyperoptEstimator, svr_linear from hpsklearn.components import ts_lagselector class TimeSeriesForecast(unittest.TestCase): def setUp(self): '''Generate a simulated dataset and define utility functions ''' ts_size = 1000 y = np.random.normal(0, 1.5, ts_size) # white noises, i.e. errors. a = np.array([.666, -.333]) c = -.5 b = np.array([1.5, -1.5, .5]) x1 = np.random.uniform(.5, 1.5, ts_size) x2 = np.random.normal(2., 1.5, ts_size) x3 = np.random.normal(3., 2.5, ts_size) X = np.c_[x1, x2, x3] for i in range(2, ts_size): y[i] += np.dot(a, y[[i - 1, i - 2]]) y += np.dot(X, b) + c self.X = X self.y = y self.a = a self.b = b self.c = c #### A utility function to convert time series to tabular data #### def ts_to_lagged_df(ts, col_id, index=None, lag_size=5): '''Convert a time series to lagged dataframe Args: ts (DataFrame): a dataframe, series or array containing the time series. If it is a dataframe or array, the 2nd dimension must have a size of 1. col_id (str): if the col id is N, all columns will be named as N1, N2, N3, ... index ([DatetimeIndex]): if not provided, will use the index from the input dataframe or series, or integers if the input is array. lag_size ([int]): the lag size to use Returns: a dataframe containing the lagged values in columns. ''' # Sanity check and extract data and index. if isinstance(ts, pd.Series): dat = ts.values.ravel() if index is None: df_idx = ts.index[lag_size:] elif isinstance(ts, pd.DataFrame): assert ts.shape[1] == 1 dat = ts.values.ravel() if index is None: df_idx = ts.index[lag_size:] else: assert isinstance(ts, np.ndarray) assert ts.ndim <= 2 if ts.ndim == 2: assert ts.shape[1] == 1 dat = ts.ravel() if index is None: df_idx = range(len(dat) - lag_size) if index is not None: assert len(index) == len(dat) df_idx = index[lag_size:] # Convert to lagged dataframe. lagged_df = np.concatenate( map(lambda x: dat[range(x - 1, x - 1 - lag_size, -1)], range(lag_size, len(dat))) ).reshape((-1, lag_size)) lagnames = map(lambda x: col_id + str(x), range(1, lag_size + 1)) lagged_df = pd.DataFrame(lagged_df, index=df_idx, columns=lagnames) return lagged_df #### End definition #### self.ts_to_lagged_df = ts_to_lagged_df def test_svm_lagselector(self): """ Using SVM as regressors, I hope hyperopt can help me optimize its hyperparameters and decide the lag size as well. I also have an exogenous dataset that I want to use to help making predictions. """ # Convert time series to tabular. max_lag_size = 10 lagged_y_df = self.ts_to_lagged_df(self.y, 'L', lag_size=max_lag_size) # dim of lagged_y_df: [990, 10]. y_target = self.y[max_lag_size:] # Setup train/test predictors and targets. test_size = 300 X_train = lagged_y_df[:-test_size].values X_test = lagged_y_df[-test_size:].values y_train = y_target[:-test_size] y_test = y_target[-test_size:] EX_train = self.X[max_lag_size:-test_size, :] EX_test = self.X[-test_size:, :] # Optimize an SVM for forecasting. svr_opt = HyperoptEstimator( preprocessing=[ts_lagselector('lag', 1, 10)], ex_preprocs=[ [] ], # must explicitly set EX preprocessings. regressor=svr_linear('svm', max_iter=1e5), algo=tpe.suggest, max_evals=30, verbose=True ) svr_opt.fit(X_train, y_train, EX_list=[EX_train], valid_size=.2, cv_shuffle=False) # It's generally a good idea to turn off shuffling for time series # forecasting to avoid over-optimistic scores. ex_n_feas = EX_train.shape[1] bm_ = svr_opt.best_model() print('\n==== Time series forecast with SVM and lag selectors ====', file=sys.stderr) print('\nThe best model found:', bm_, file=sys.stderr) print('=' * 40, file=sys.stderr) print('Actual parameter values\n', 'lag size:', len(self.a), '\n', 'a:', self.a, 'b:', self.b, 'c:', self.c, file=sys.stderr) svr_mod = bm_['learner'] a_hat = np.round(svr_mod.coef_[0, :-ex_n_feas], 3) b_hat = np.round(svr_mod.coef_[0, -ex_n_feas:], 3) c_hat = np.round(svr_mod.intercept_, 3) print('-' * 4, file=sys.stderr) print('Estimated parameter values\n', 'lag size:', len(a_hat), '\n', 'a:', a_hat, 'b:', b_hat, 'c:', c_hat, file=sys.stderr) print('=' * 40, file=sys.stderr) print('Best trial validation R2:', 1 - svr_opt.trials.best_trial['result']['loss'], file=sys.stderr) print('Test R2:', svr_opt.score(X_test, y_test, EX_list=[EX_test]), file=sys.stderr) if __name__ == '__main__': unittest.main()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/scikit-small-ensemble/setup.py
training/helpers/scikit-small-ensemble/setup.py
#!/usr/bin/env python # -*- coding: utf-8 -*- from setuptools import find_packages, setup setup( name='scikit-small-ensemble', version='0.0.2', author='Stewart Park', url='https://github.com/stewartpark/scikit-small-ensemble', author_email='hello@stewartjpark.com', license='MIT', install_requires=['lz4', 'joblib'], packages=find_packages(), zip_safe=False )
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/scikit-small-ensemble/scikit_small_ensemble/scikit_ensemble.py
training/helpers/scikit-small-ensemble/scikit_small_ensemble/scikit_ensemble.py
import lz4 as zlib import tempfile import joblib import os try: import _pickle as pickle except ImportError: try: import cPickle as pickle except ImportError: print('cPickle is not installed. Using the builtin pickle instead.') import pickle class CompressedEstimators(object): def __init__(self, estimators, ratio): self.cutoff = int(len(estimators) * ratio) self.estimators = [ zlib.compress(pickle.dumps(x)) if i < self.cutoff else x for i, x in enumerate(estimators) ] def __getitem__(self, index): estimator = self.estimators[index] if index < self.cutoff: return pickle.loads(zlib.decompress(estimator)) else: return estimator def __len__(self): return len(self.estimators) class DiskEstimators(object): def __init__(self, estimators, ratio): self.cutoff = int(len(estimators) * ratio) self.saved_dir = tempfile.mkdtemp() for i in range(self.cutoff): joblib.dump(estimators[i], os.path.join(self.saved_dir, str(i)), compress=0) self.estimators = [ os.path.join(self.saved_dir, str(i)) if i < self.cutoff else x for i, x in enumerate(estimators) ] def __getitem__(self, index): estimator = self.estimators[index] if index < self.cutoff: return joblib.load(estimator, mmap_mode='r') else: return estimator def __len__(self): return len(self.estimators)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/scikit-small-ensemble/scikit_small_ensemble/__init__.py
training/helpers/scikit-small-ensemble/scikit_small_ensemble/__init__.py
from __future__ import absolute_import from scikit_small_ensemble.scikit_ensemble import CompressedEstimators, DiskEstimators def compress(model, ratio=0.5): if isinstance(model.estimators_, CompressedEstimators): raise Exception("The model is already compressed.") model.estimators_ = CompressedEstimators(model, ratio) def memory_map(model, ratio=1.0): if isinstance(model.estimators_, DiskEstimators): raise Exception("The model is already memory mapped.") model.estimators_ = DiskEstimators(model, ratio)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/devol/setup.py
training/helpers/devol/setup.py
from setuptools import setup, find_packages setup(name='devol', version='0.02', description='Genetic search for CNN classifier in Keras', url='https//github.com/joedav/devol', author='Joe Davison', author_email='josephddavison@gmail.com', license='MIT', classifiers=[ # How mature is this project? Common values are # 3 - Alpha # 4 - Beta # 5 - Production/Stable 'Development Status :: 3 - Alpha', # Indicate who your project is intended for 'Intended Audience :: Developers', 'Topic :: Scientific/Engineering :: Artificial Intelligence', # Pick your license as you wish (should match "license" above) 'License :: OSI Approved :: MIT License', # Specify the Python versions you support here. In particular, ensure # that you indicate whether you support Python 2, Python 3 or both. 'Programming Language :: Python :: 3.3', 'Programming Language :: Python :: 3.4', 'Programming Language :: Python :: 3.5', 'Programming Language :: Python :: 3.6', 'Programming Language :: Python :: 2.7', ], keywords='genetic algorithm', packages=['devol'], install_requires=['keras',], )
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/devol/devol/__init__.py
training/helpers/devol/devol/__init__.py
from .devol import DEvol from .genome_handler import GenomeHandler __all__ = ['DEvol', 'GenomeHandler']
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/devol/devol/genome_handler.py
training/helpers/devol/devol/genome_handler.py
import numpy as np import random as rand import math from keras.models import Sequential from keras.layers import Activation, Dense, Dropout, Flatten from keras.layers.convolutional import Convolution2D, MaxPooling2D from keras.layers.normalization import BatchNormalization class GenomeHandler: """ Defines the configuration and handles the conversion and mutation of individual genomes. Should be created and passed to a `DEvol` instance. --- Genomes are represented as fixed-with lists of integers corresponding to sequential layers and properties. A model with 2 convolutional layers and 1 dense layer would look like: [<conv layer><conv layer><dense layer><optimizer>] The makeup of the convolutional layers and dense layers is defined in the GenomeHandler below under self.convolutional_layer_shape and self.dense_layer_shape. <optimizer> consists of just one property. """ def __init__(self, max_conv_layers, max_dense_layers, max_filters, max_dense_nodes, input_shape, n_classes, batch_normalization=True, dropout=True, max_pooling=True, optimizers=None, activations=None): """ Creates a GenomeHandler according Args: max_conv_layers: The maximum number of convolutional layers max_conv_layers: The maximum number of dense (fully connected) layers, including output layer max_filters: The maximum number of conv filters (feature maps) in a convolutional layer max_dense_nodes: The maximum number of nodes in a dense layer input_shape: The shape of the input n_classes: The number of classes batch_normalization (bool): whether the GP should include batch norm dropout (bool): whether the GP should include dropout max_pooling (bool): whether the GP should include max pooling layers optimizers (list): list of optimizers to be tried by the GP. By default, the network uses Keras's built-in adam, rmsprop, adagrad, and adadelta activations (list): list of activation functions to be tried by the GP. By default, relu and sigmoid. """ if max_dense_layers < 1: raise ValueError( "At least one dense layer is required for softmax layer" ) if max_filters > 0: filter_range_max = int(math.log(max_filters, 2)) + 1 else: filter_range_max = 0 self.optimizer = optimizers or [ 'adam', 'rmsprop', 'adagrad', 'adadelta' ] self.activation = activations or [ 'relu', 'sigmoid', ] self.convolutional_layer_shape = [ "active", "num filters", "batch normalization", "activation", "dropout", "max pooling", ] self.dense_layer_shape = [ "active", "num nodes", "batch normalization", "activation", "dropout", ] self.layer_params = { "active": [0, 1], "num filters": [2**i for i in range(3, filter_range_max)], "num nodes": [2**i for i in range(4, int(math.log(max_dense_nodes, 2)) + 1)], "batch normalization": [0, (1 if batch_normalization else 0)], "activation": list(range(len(self.activation))), "dropout": [(i if dropout else 0) for i in range(11)], "max pooling": list(range(3)) if max_pooling else 0, } self.convolution_layers = max_conv_layers self.convolution_layer_size = len(self.convolutional_layer_shape) self.dense_layers = max_dense_layers - 1 # this doesn't include the softmax layer, so -1 self.dense_layer_size = len(self.dense_layer_shape) self.input_shape = input_shape self.n_classes = n_classes def convParam(self, i): key = self.convolutional_layer_shape[i] return self.layer_params[key] def denseParam(self, i): key = self.dense_layer_shape[i] return self.layer_params[key] def mutate(self, genome, num_mutations): num_mutations = np.random.choice(num_mutations) for i in range(num_mutations): index = np.random.choice(list(range(1, len(genome)))) if index < self.convolution_layer_size * self.convolution_layers: if genome[index - index % self.convolution_layer_size]: range_index = index % self.convolution_layer_size choice_range = self.convParam(range_index) genome[index] = np.random.choice(choice_range) elif rand.uniform(0, 1) <= 0.01: # randomly flip deactivated layers genome[index - index % self.convolution_layer_size] = 1 elif index != len(genome) - 1: offset = self.convolution_layer_size * self.convolution_layers new_index = (index - offset) present_index = new_index - new_index % self.dense_layer_size if genome[present_index + offset]: range_index = new_index % self.dense_layer_size choice_range = self.denseParam(range_index) genome[index] = np.random.choice(choice_range) elif rand.uniform(0, 1) <= 0.01: genome[present_index + offset] = 1 else: genome[index] = np.random.choice(list(range(len(self.optimizer)))) return genome def decode(self, genome): if not self.is_compatible_genome(genome): raise ValueError("Invalid genome for specified configs") model = Sequential() offset = 0 dim = min(self.input_shape[:-1]) # keep track of smallest dimension input_layer = True for i in range(self.convolution_layers): if genome[offset]: convolution = None if input_layer: convolution = Convolution2D( genome[offset + 1], (3, 3), padding='same', input_shape=self.input_shape ) input_layer = False else: convolution = Convolution2D( genome[offset + 1], (3, 3), padding='same' ) model.add(convolution) if genome[offset + 2]: model.add(BatchNormalization()) model.add(Activation(self.activation[genome[offset + 3]])) model.add(Dropout(float(genome[offset + 4] / 20.0))) max_pooling_type = genome[offset + 5] # must be large enough for a convolution if max_pooling_type == 1 and dim >= 5: model.add(MaxPooling2D(pool_size=(2, 2), padding="same")) dim = int(math.ceil(dim / 2)) offset += self.convolution_layer_size if not input_layer: model.add(Flatten()) for i in range(self.dense_layers): if genome[offset]: dense = None if input_layer: dense = Dense(genome[offset + 1], input_shape=self.input_shape) input_layer = False else: dense = Dense(genome[offset + 1]) model.add(dense) if genome[offset + 2]: model.add(BatchNormalization()) model.add(Activation(self.activation[genome[offset + 3]])) model.add(Dropout(float(genome[offset + 4] / 20.0))) offset += self.dense_layer_size model.add(Dense(self.n_classes, activation='softmax')) model.compile(loss='categorical_crossentropy', optimizer=self.optimizer[genome[offset]], metrics=["accuracy"]) return model def genome_representation(self): encoding = [] for i in range(self.convolution_layers): for key in self.convolutional_layer_shape: encoding.append("Conv" + str(i) + " " + key) for i in range(self.dense_layers): for key in self.dense_layer_shape: encoding.append("Dense" + str(i) + " " + key) encoding.append("Optimizer") return encoding def generate(self): genome = [] for i in range(self.convolution_layers): for key in self.convolutional_layer_shape: param = self.layer_params[key] genome.append(np.random.choice(param)) for i in range(self.dense_layers): for key in self.dense_layer_shape: param = self.layer_params[key] genome.append(np.random.choice(param)) genome.append(np.random.choice(list(range(len(self.optimizer))))) genome[0] = 1 return genome def is_compatible_genome(self, genome): expected_len = self.convolution_layers * self.convolution_layer_size \ + self.dense_layers * self.dense_layer_size + 1 if len(genome) != expected_len: return False ind = 0 for i in range(self.convolution_layers): for j in range(self.convolution_layer_size): if genome[ind + j] not in self.convParam(j): return False ind += self.convolution_layer_size for i in range(self.dense_layers): for j in range(self.dense_layer_size): if genome[ind + j] not in self.denseParam(j): return False ind += self.dense_layer_size if genome[ind] not in range(len(self.optimizer)): return False return True def best_genome(self, csv_path, metric="accuracy", include_metrics=True): best = max if metric is "accuracy" else min col = -1 if metric is "accuracy" else -2 data = np.genfromtxt(csv_path, delimiter=",") row = list(data[:, col]).index(best(data[:, col])) genome = list(map(int, data[row, :-2])) if include_metrics: genome += list(data[row, -2:]) return genome def decode_best(self, csv_path, metric="accuracy"): return self.decode(self.best_genome(csv_path, metric, False))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/devol/devol/devol.py
training/helpers/devol/devol/devol.py
""" Run a genetic algorithm to find an appropriate architecture for some image classification task with Keras+TF. To use, define a `GenomeHandler` defined in genomehandler.py. Then pass it, with training data, to a DEvol instance to run the genetic algorithm. See the readme for more detailed instructions. """ from __future__ import print_function import random as rand import csv import operator import gc import os from datetime import datetime from keras.callbacks import EarlyStopping from keras.models import load_model import keras.backend as K from sklearn.metrics import log_loss import numpy as np if K.backend() == 'tensorflow': import tensorflow as tf __all__ = ['DEvol'] METRIC_OPS = [operator.__lt__, operator.__gt__] METRIC_OBJECTIVES = [min, max] class DEvol: """ Object which carries out genetic search and returns top performing model upon completion. """ def __init__(self, genome_handler, data_path=""): """ Initialize a DEvol object which carries out the training and evaluation of a genetic search. Args: genome_handler (GenomeHandler): the genome handler object defining the restrictions for the architecture search space data_path (str): the file which the genome encodings and metric data will be stored in """ self.genome_handler = genome_handler self.datafile = data_path or (datetime.now().ctime() + '.csv') self._bssf = -1 if os.path.isfile(data_path) and os.stat(data_path).st_size > 1: raise ValueError(('Non-empty file %s already exists. Please change' 'file path to prevent overwritten genome data.' % data_path)) print("Genome encoding and metric data stored at", self.datafile, "\n") with open(self.datafile, 'a') as csvfile: writer = csv.writer(csvfile, delimiter=',', quotechar='"', quoting=csv.QUOTE_MINIMAL) metric_cols = ["Val Loss", "Val Accuracy"] genome = genome_handler.genome_representation() + metric_cols writer.writerow(genome) def set_objective(self, metric): """ Set the metric for optimization. Can also be done by passing to `run`. Args: metric (str): either 'acc' to maximize classification accuracy, or else 'loss' to minimize the loss function """ if metric == 'acc': metric = 'accuracy' if metric not in ['loss', 'accuracy']: raise ValueError(('Invalid metric name {} provided - should be' '"accuracy" or "loss"').format(metric)) self._metric = metric self._objective = "max" if self._metric == "accuracy" else "min" self._metric_index = 1 if self._metric == 'loss' else -1 self._metric_op = METRIC_OPS[self._objective == 'max'] self._metric_objective = METRIC_OBJECTIVES[self._objective == 'max'] def run(self, dataset, num_generations, pop_size, epochs, fitness=None, metric='accuracy'): """ Run genetic search on dataset given number of generations and population size Args: dataset : tuple or list of numpy arrays in form ((train_data, train_labels), (validation_data, validation_labels)) num_generations (int): number of generations to search pop_size (int): initial population size epochs (int): epochs for each model eval, passed to keras model.fit fitness (None, optional): scoring function to be applied to population scores, will be called on a numpy array which is a min/max scaled version of evaluated model metrics, so It should accept a real number including 0. If left as default just the min/max scaled values will be used. metric (str, optional): must be "accuracy" or "loss" , defines what to optimize during search Returns: keras model: best model found with weights """ self.set_objective(metric) # If no validation data is given set it to None if len(dataset) == 2: (self.x_train, self.y_train), (self.x_test, self.y_test) = dataset self.x_val = None self.y_val = None else: (self.x_train, self.y_train), (self.x_test, self.y_test), (self.x_val, self.y_val) = dataset # generate and evaluate initial population members = self._generate_random_population(pop_size) pop = self._evaluate_population(members, epochs, fitness, 0, num_generations) # evolve for gen in range(1, num_generations): members = self._reproduce(pop, gen) pop = self._evaluate_population(members, epochs, fitness, gen, num_generations) return load_model('best-model.h5') def _reproduce(self, pop, gen): members = [] # 95% of population from crossover for _ in range(int(len(pop) * 0.95)): members.append(self._crossover(pop.select(), pop.select())) # best models survive automatically members += pop.get_best(len(pop) - int(len(pop) * 0.95)) # randomly mutate for imem, mem in enumerate(members): members[imem] = self._mutate(mem, gen) return members def _evaluate(self, genome, epochs): model = self.genome_handler.decode(genome) loss, accuracy = None, None fit_params = { 'x': self.x_train, 'y': self.y_train, 'validation_split': 0.1, 'epochs': epochs, 'verbose': 1, 'callbacks': [ EarlyStopping(monitor='val_loss', patience=1, verbose=1) ] } if self.x_val is not None: fit_params['validation_data'] = (self.x_val, self.y_val) try: model.fit(**fit_params) loss, accuracy = model.evaluate(self.x_test, self.y_test, verbose=0) except Exception as e: loss, accuracy = self._handle_broken_model(model, e) self._record_stats(model, genome, loss, accuracy) return model, loss, accuracy def _record_stats(self, model, genome, loss, accuracy): with open(self.datafile, 'a') as csvfile: writer = csv.writer(csvfile, delimiter=',', quotechar='"', quoting=csv.QUOTE_MINIMAL) row = list(genome) + [loss, accuracy] writer.writerow(row) met = loss if self._metric == 'loss' else accuracy if (self._bssf is -1 or self._metric_op(met, self._bssf) and accuracy is not 0): try: os.remove('best-model.h5') except OSError: pass self._bssf = met model.save('best-model.h5') def _handle_broken_model(self, model, error): del model n = self.genome_handler.n_classes loss = log_loss(np.concatenate(([1], np.zeros(n - 1))), np.ones(n) / n) accuracy = 1 / n gc.collect() if K.backend() == 'tensorflow': K.clear_session() tf.reset_default_graph() print('An error occurred and the model could not train:') print(error) print(('Model assigned poor score. Please ensure that your model' 'constraints live within your computational resources.')) return loss, accuracy def _evaluate_population(self, members, epochs, fitness, igen, ngen): fit = [] for imem, mem in enumerate(members): self._print_evaluation(imem, len(members), igen, ngen) res = self._evaluate(mem, epochs) v = res[self._metric_index] del res fit.append(v) fit = np.array(fit) self._print_result(fit, igen) return _Population(members, fit, fitness, obj=self._objective) def _print_evaluation(self, imod, nmod, igen, ngen): fstr = '\nmodel {0}/{1} - generation {2}/{3}:\n' print(fstr.format(imod + 1, nmod, igen + 1, ngen)) def _generate_random_population(self, size): return [self.genome_handler.generate() for _ in range(size)] def _print_result(self, fitness, generation): result_str = ('Generation {3}:\t\tbest {4}: {0:0.4f}\t\taverage:' '{1:0.4f}\t\tstd: {2:0.4f}') print(result_str.format(self._metric_objective(fitness), np.mean(fitness), np.std(fitness), generation + 1, self._metric)) def _crossover(self, genome1, genome2): cross_ind = rand.randint(0, len(genome1)) child = genome1[:cross_ind] + genome2[cross_ind:] return child def _mutate(self, genome, generation): # increase mutations as program continues num_mutations = max(3, generation // 4) return self.genome_handler.mutate(genome, num_mutations) class _Population(object): def __len__(self): return len(self.members) def __init__(self, members, fitnesses, score, obj='max'): self.members = members scores = fitnesses - fitnesses.min() if scores.max() > 0: scores /= scores.max() if obj == 'min': scores = 1 - scores if score: self.scores = score(scores) else: self.scores = scores self.s_fit = sum(self.scores) def get_best(self, n): combined = [(self.members[i], self.scores[i]) for i in range(len(self.members))] sorted(combined, key=(lambda x: x[1]), reverse=True) return [x[0] for x in combined[:n]] def select(self): dart = rand.uniform(0, self.s_fit) sum_fits = 0 for i in range(len(self.members)): sum_fits += self.scores[i] if sum_fits >= dart: return self.members[i]
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/devol/example/demo.py
training/helpers/devol/example/demo.py
from __future__ import print_function from keras.datasets import mnist from keras.utils.np_utils import to_categorical import numpy as np from keras import backend as K from devol import DEvol, GenomeHandler # **Prepare dataset** # This problem uses mnist, a handwritten digit classification problem used # for many introductory deep learning examples. Here, we load the data and # prepare it for use by the GPU. We also do a one-hot encoding of the labels. K.set_image_data_format("channels_last") (x_train, y_train), (x_test, y_test) = mnist.load_data() x_train = x_train.reshape(x_train.shape[0], 28, 28, 1).astype('float32') / 255 x_test = x_test.reshape(x_test.shape[0], 28, 28, 1).astype('float32') / 255 y_train = to_categorical(y_train) y_test = to_categorical(y_test) dataset = ((x_train, y_train), (x_test, y_test)) # **Prepare the genome configuration** # The `GenomeHandler` class handles the constraints that are imposed upon # models in a particular genetic program. See `genome-handler.py` # for more information. genome_handler = GenomeHandler(max_conv_layers=6, max_dense_layers=2, # includes final dense layer max_filters=256, max_dense_nodes=1024, input_shape=x_train.shape[1:], n_classes=10) # **Create and run the genetic program** # The next, and final, step is create a `DEvol` and run it. Here we specify # a few settings pertaining to the genetic program. The program # will save each genome's encoding, as well as the model's loss and # accuracy, in a `.csv` file printed at the beginning of program. # The best model is returned decoded and with `epochs` training done. devol = DEvol(genome_handler) model = devol.run(dataset=dataset, num_generations=20, pop_size=20, epochs=5) print(model.summary())
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/setup.py
training/helpers/autobazaar/setup.py
#!/usr/bin/env python # -*- coding: utf-8 -*- from setuptools import find_packages, setup with open('README.md') as readme_file: readme = readme_file.read() with open('HISTORY.md') as history_file: history = history_file.read() install_requires = [ 'baytune>=0.2.1,<0.3', 'mlblocks>=0.3.2,<0.4', 'mlprimitives>=0.2.2,<0.3', 'scikit-learn>=0.20,<0.21', 'mit-d3m>=0.2.1,<0.3', 'numpy>=1.16.5,<1.17', 'gitpython>=3.1.1,<4' ] tests_require = [ 'pytest>=3.4.2', 'pytest-cov>=2.6.0', ] setup_requires = [ 'pytest-runner>=2.11.1', ] development_requires = [ # general 'pip>=9.0.1', 'bumpversion>=0.5.3,<0.6', 'watchdog>=0.8.3,<0.11', # docs 'm2r>=0.2.0,<0.3', 'autodocsumm>=0.1.10,<0.2', 'Sphinx>=1.7.1,<3', 'sphinx_rtd_theme>=0.2.4,<0.5', # style check 'flake8>=3.7.7,<4', 'isort>=4.3.4,<5', # fix style issues 'autoflake>=1.1,<2', 'autopep8>=1.4.3,<2', # distribute on PyPI 'twine>=1.10.0,<4', 'wheel>=0.30.0', # Advanced testing 'coverage>=4.5.1,<6', 'tox>=2.9.1,<4', ] setup( author='MIT Data To AI Lab', author_email='dailabmit@gmail.com', classifiers=[ 'Development Status :: 3 - Alpha', 'Intended Audience :: Developers', 'License :: OSI Approved :: MIT License', 'Natural Language :: English', 'Programming Language :: Python :: 3', 'Programming Language :: Python :: 3.5', 'Programming Language :: Python :: 3.6', 'Programming Language :: Python :: 3.7', 'Topic :: Scientific/Engineering :: Artificial Intelligence', ], description='The Machine Learning Bazaar', entry_points={ 'console_scripts': [ 'abz=autobazaar.__main__:main' ] }, extras_require={ 'dev': development_requires + tests_require, 'tests': tests_require, }, include_package_data=True, install_requires=install_requires, license='MIT license', long_description=readme + '\n\n' + history, long_description_content_type='text/markdown', keywords='automl machine learning hyperparameters tuning classification regression autobazaar', name='autobazaar', packages=find_packages(include=['autobazaar', 'autobazaar.*']), python_requires='>=3.5', setup_requires=setup_requires, test_suite='tests', tests_require=tests_require, url='https://github.com/HDI-project/AutoBazaar', version='0.2.1-dev', zip_safe=False, )
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/autobazaar/search.py
training/helpers/autobazaar/autobazaar/search.py
# -*- coding: utf-8 -*- """AutoBazaar Search Module. This module contains the PipelineSearcher, which is the class that contains the main logic of the Auto Machine Learning process. """ import gc import itertools import json import logging import os import signal import warnings from collections import defaultdict from datetime import datetime, timedelta import numpy as np import pandas as pd from btb import HyperParameter from btb.tuning import GP, GPEi, Uniform from mit_d3m.loaders import get_loader from mlblocks.mlpipeline import MLPipeline from sklearn.model_selection import KFold, StratifiedKFold from autobazaar.pipeline import ABPipeline from autobazaar.utils import ensure_dir, make_dumpable, remove_dots, restore_dots LOGGER = logging.getLogger(__name__) warnings.filterwarnings("ignore", category=DeprecationWarning) TEMPLATES_DIR = os.path.join(os.path.dirname(__file__), 'templates') TRIVIAL_PIPELINE_METHOD = { 'classification': 'mode', 'regression': 'median', 'collaborativeFiltering': 'median', 'graphMatching': 'mode', } TUNERS = { 'gp': GP, 'gpei': GPEi, 'uniform': Uniform } PARAM_TYPES = { 'str': 'string', } class StopSearch(KeyboardInterrupt): pass class UnsupportedProblem(Exception): pass def log_times(name, append=False): def decorator(wrapped): def wrapper(self, *args, **kwargs): start = datetime.utcnow() result = wrapped(self, *args, **kwargs) elapsed = (datetime.utcnow() - start).total_seconds() if append: attribute = getattr(self, name, None) if attribute is None: attribute = list() setattr(self, name, attribute) attribute.append(elapsed) else: setattr(self, name, elapsed) return result return wrapper return decorator class PipelineSearcher(object): """PipelineSearcher class. This class is responsible for searching the best pipeline to solve a given dataset and problem. """ def __init__(self, pipelines_dir, db=None, test_id=None, tuner_type='gp', cv_splits=5, random_state=0): self._db = db self._pipelines_dir = pipelines_dir ensure_dir(self._pipelines_dir) self._cv_splits = cv_splits self._random_state = random_state self._tuner_type = tuner_type self._tuner_class = TUNERS[tuner_type] self._test_id = test_id def _dump_pipelines(self): LOGGER.info('Dumping best pipelines') dumped = list() gc.collect() for details in self._to_dump: pipeline = details['pipeline'] if not pipeline.dumped: pipeline.fit(self.data_params) mlpipeline = pipeline.pipeline LOGGER.info("Dumping pipeline %s: %s", pipeline.id, pipeline.pipeline) LOGGER.info("Hyperparameters: %s", mlpipeline.get_hyperparameters()) pipeline.dump(self._pipelines_dir) details['pipeline'] = pipeline.id dumped.append(details) gc.collect() else: LOGGER.info("Skipping already dumped pipeline %s", pipeline.id) return dumped def _set_for_dump(self, pipeline): self._to_dump.append({ 'elapsed': (datetime.utcnow() - self.start_time).total_seconds(), 'iterations': len(self.pipelines) - 1, 'cv_score': self.best_pipeline.score, 'rank': self.best_pipeline.rank, 'pipeline': pipeline, 'load_time': self.load_time, 'trivial_time': self.trivial_time, 'cv_time': np.sum(self.cv_times), }) def _save_pipeline(self, pipeline): pipeline_dict = pipeline.to_dict(True) pipeline_dict['_id'] = pipeline.id pipeline_dict['ts'] = datetime.utcnow() self.pipelines.append(pipeline_dict) if self._db: insertable = remove_dots(pipeline_dict) insertable.pop('problem_doc') insertable['dataset'] = self.dataset_id insertable['tuner_type'] = self._tuner_type insertable['test_id'] = self._test_id self._db.pipelines.insert_one(insertable) @log_times('trivial_time') def _build_trivial_pipeline(self): LOGGER.info("Building the Trivial pipeline") try: method = TRIVIAL_PIPELINE_METHOD.get(self.task_type) pipeline_dict = { 'name': 'trivial.{}'.format(method), 'primitives': ['mlprimitives.custom.trivial.TrivialPredictor'], 'init_params': { 'mlprimitives.custom.trivial.TrivialPredictor': { 'method': method } } } pipeline = ABPipeline(pipeline_dict, self.loader, self.metric, self.problem_doc) pipeline.cv_score(self.data_params.X, self.data_params.y, self.data_params.context, cv=self.kf) self._save_pipeline(pipeline) return pipeline except Exception: # if the Trivial pipeline crashes we can do nothing, # so we just log the error and move on. LOGGER.exception("The Trivial pipeline crashed.") def _load_template_json(self, template_name): if template_name.endswith('.json'): template_filename = template_name name = template_name[:-5] else: name = template_name template_name = template_name.replace('/', '.') + '.json' template_filename = os.path.join(TEMPLATES_DIR, template_name) if os.path.exists(template_filename): with open(template_filename, 'r') as template_file: template_dict = json.load(template_file) template_dict['name'] = name return template_dict def _find_template(self, template_name): match = { 'metadata.name': template_name } cursor = self._db.pipelines.find(match) templates = list(cursor.sort('metadata.insert_ts', -1).limit(1)) if templates: template = templates[0] template['name'] = template.pop('metadata')['name'] template['template'] = str(template.pop('_id')) return restore_dots(template) def _load_template(self, template_name): if self._db: template = self._find_template(template_name) if template: return template return self._load_template_json(template_name) def _get_template(self, template_name=None): if template_name: template = self._load_template(template_name) if not template: raise ValueError("Template {} not found".format(template_name)) primitives = '\n'.join(template['primitives']) LOGGER.info('Using template %s:\n%s', template_name, primitives) return template else: problem_type = [ self.data_modality, self.task_type, self.task_subtype ] for levels in reversed(range(1, 4)): # Try the following options: # modality/task/subtask/default # modality/task/default # modality/default template_name = '/'.join(problem_type[:levels] + ['default']) template = self._load_template(template_name) if template: primitives = '\n'.join(template['primitives']) LOGGER.info('Using template %s:\n%s', template_name, primitives) return template # Nothing has been found for this modality/task/subtask combination problem_type = '/'.join(problem_type) LOGGER.error('Problem type not supported %s', problem_type) raise UnsupportedProblem(problem_type) @log_times('cv_times', append=True) def _cv_pipeline(self, params=None): pipeline_dict = self.template_dict.copy() if params: pipeline_dict['hyperparameters'] = params pipeline = ABPipeline(pipeline_dict, self.loader, self.metric, self.problem_doc) X = self.data_params.X y = self.data_params.y context = self.data_params.context try: pipeline.cv_score(X, y, context, cv=self.kf) except KeyboardInterrupt: raise except Exception: LOGGER.exception("Crash cross validating pipeline %s", pipeline.id) return None return pipeline def _create_tuner(self, pipeline): # Build an MLPipeline to get the tunables and the default params mlpipeline = MLPipeline.from_dict(self.template_dict) tunable_hyperparameters = mlpipeline.get_tunable_hyperparameters() tunables = [] tunable_keys = [] for block_name, params in tunable_hyperparameters.items(): for param_name, param_details in params.items(): key = (block_name, param_name) param_type = param_details['type'] param_type = PARAM_TYPES.get(param_type, param_type) if param_type == 'bool': param_range = [True, False] else: param_range = param_details.get('range') or param_details.get('values') value = HyperParameter(param_type, param_range) tunables.append((key, value)) tunable_keys.append(key) # Create the tuner LOGGER.info('Creating %s tuner', self._tuner_class.__name__) self.tuner = self._tuner_class(tunables) if pipeline: try: # Add the default params and the score obtained by them to the tuner. default_params = defaultdict(dict) for block_name, params in pipeline.pipeline.get_hyperparameters().items(): for param, value in params.items(): key = (block_name, param) if key in tunable_keys: if value is None: raise ValueError('None value is not supported') default_params[key] = value if pipeline.rank is not None: self.tuner.add(default_params, 1 - pipeline.rank) except ValueError: pass def _set_checkpoint(self): next_checkpoint = self.checkpoints.pop(0) interval = next_checkpoint - self.current_checkpoint self._stop_time = datetime.utcnow() + timedelta(seconds=interval) LOGGER.info("Setting %s seconds checkpoint in %s seconds: %s", next_checkpoint, interval, self._stop_time) signal.alarm(interval) self.current_checkpoint = next_checkpoint def _checkpoint(self, signum=None, frame=None, final=False): signal.alarm(0) checkpoint_name = 'Final' if final else str(self.current_checkpoint) + ' seconds' LOGGER.info("%s checkpoint reached", checkpoint_name) try: if self.best_pipeline: self._set_for_dump(self.best_pipeline) except KeyboardInterrupt: raise except Exception: LOGGER.exception("Checkpoint dump crashed") if final or not bool(self.checkpoints): self.current_checkpoint = None # LOGGER.warn("Stopping Search") # raise StopSearch() else: self._set_checkpoint() def _check_stop(self): if self._stop_time and self._stop_time < datetime.utcnow(): LOGGER.warn("Stop Time already passed. Stopping Search!") raise StopSearch() def _setup_search(self, d3mds, budget, checkpoints, template_name): self.start_time = datetime.utcnow() self.cv_times = list() # Problem variables self.problem_id = d3mds.get_problem_id() self.task_type = d3mds.get_task_type() self.task_subtype = d3mds.problem.get_task_subtype() # TODO: put this in mit-d3m loaders if self.task_type == 'vertex_classification': self.task_type = 'vertex_nomination' self.problem_doc = d3mds.problem_doc # Dataset variables self.dataset_id = d3mds.dataset_id self.data_modality = d3mds.get_data_modality() # TODO: put this in mit-d3m loaders if self.data_modality == 'edgeList': self.data_modality = 'graph' self.metric = d3mds.get_metric() self.loader = get_loader(self.data_modality, self.task_type) self.best_pipeline = None self.pipelines = [] self.checkpoints = sorted(checkpoints or []) self.current_checkpoint = 0 self._to_dump = [] if not self.checkpoints and budget is None: self.budget = 1 else: self.budget = budget self.template_dict = self._get_template(template_name) LOGGER.info("Running TA2 Search") LOGGER.info("Problem Id: %s", self.problem_id) LOGGER.info(" Data Modality: %s", self.data_modality) LOGGER.info(" Task type: %s", self.task_type) LOGGER.info(" Task subtype: %s", self.task_subtype) LOGGER.info(" Metric: %s", self.metric) LOGGER.info(" Checkpoints: %s", self.checkpoints) LOGGER.info(" Budget: %s", self.budget) @log_times('load_time') def _load_data(self, d3mds): self.data_params = self.loader.load(d3mds) def _setup_cv(self): if isinstance(self.data_params.y, pd.Series): min_samples = self.data_params.y.value_counts().min() else: y = self.data_params.y min_samples = y.groupby(list(y.columns)).size().min() if self.task_type == 'classification' and min_samples >= self._cv_splits: self.kf = StratifiedKFold( n_splits=self._cv_splits, shuffle=True, random_state=self._random_state ) else: self.kf = KFold( n_splits=self._cv_splits, shuffle=True, random_state=self._random_state ) def search(self, d3mds, template_name=None, budget=None, checkpoints=None): try: self._setup_search(d3mds, budget, checkpoints, template_name) self._load_data(d3mds) self._setup_cv() # Build the trivial pipeline self.best_pipeline = self._build_trivial_pipeline() # Do not continue if there is no budget or no fit data if budget == 0 or not len(self.data_params.X): raise StopSearch() # Build the default pipeline default_pipeline = self._cv_pipeline() if default_pipeline: self.best_pipeline = default_pipeline self._save_pipeline(default_pipeline) if budget == 1: raise StopSearch() elif budget is not None: iterator = range(budget - 1) else: iterator = itertools.count() # infinite range # Build the tuner self._create_tuner(default_pipeline) LOGGER.info("Starting the tuning loop") if self.checkpoints: signal.signal(signal.SIGALRM, self._checkpoint) self._set_checkpoint() else: self._stop_time = None for iteration in iterator: self._check_stop() proposed_params = self.tuner.propose() params = make_dumpable(proposed_params) LOGGER.info("Cross validating pipeline %s", iteration + 1) pipeline = self._cv_pipeline(params) if pipeline and (pipeline.rank is not None): self.tuner.add(proposed_params, 1 - pipeline.rank) LOGGER.info("Saving pipeline %s: %s", iteration + 1, pipeline.id) self._save_pipeline(pipeline) if not self.best_pipeline or (pipeline.rank < self.best_pipeline.rank): self.best_pipeline = pipeline LOGGER.info('Best pipeline so far: %s; rank: %s, score: %s', self.best_pipeline, self.best_pipeline.rank, self.best_pipeline.score) else: self.tuner.add(proposed_params, -1000000) except KeyboardInterrupt: pass finally: signal.alarm(0) if self.current_checkpoint: self._checkpoint(final=True) elif self.best_pipeline and not checkpoints: self._set_for_dump(self.best_pipeline) if self.best_pipeline: LOGGER.info('Best pipeline for problem %s found: %s; rank: %s, score: %s', self.problem_id, self.best_pipeline, self.best_pipeline.rank, self.best_pipeline.score) else: LOGGER.info('No pipeline could be found for problem %s', self.problem_id) return self._dump_pipelines()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/autobazaar/__main__.py
training/helpers/autobazaar/autobazaar/__main__.py
#!/usr/bin/env python # -*- coding: utf-8 -*- """AutoBazaar Command Line Module.""" import argparse import gc import json import os import shutil import socket import sys import traceback import warnings from datetime import datetime import cloudpickle import pandas as pd from mit_d3m import metrics from mit_d3m.dataset import D3MDS from mit_d3m.db import get_db from mit_d3m.stats import get_stats from mit_d3m.utils import logging_setup, make_abs import autobazaar from autobazaar.search import TUNERS, PipelineSearcher from autobazaar.utils import encode_score, make_keras_picklable warnings.filterwarnings(action='ignore') VERSION = autobazaar.get_version() def _load_targets(datasets_dir, dataset, problem): score_phase = 'SCORE' if problem: score_phase += '_' + problem score_dir = os.path.join(datasets_dir, dataset, score_phase) csv_path = os.path.join(score_dir, 'targets.csv') if not os.path.exists(csv_path): csv_path = os.path.join(score_dir, 'dataset_SCORE', 'tables', 'learningData.csv') return pd.read_csv(csv_path, index_col='d3mIndex') def _get_metric(problem_path): problem_schema = os.path.join(problem_path, 'problemDoc.json') with open(problem_schema, 'r') as f: problem_doc = json.load(f) problem_metrics = problem_doc['inputs']['performanceMetrics'] if len(problem_metrics) > 1: raise Exception("Wrong number of metrics") return metrics.METRICS_DICT[problem_metrics[0]['metric']] def _get_dataset_paths(datasets_dir, dataset, phase, problem): if problem: full_phase = phase + '_' + problem else: full_phase = phase root_dir = os.path.join(datasets_dir, dataset, full_phase) dataset_path = os.path.join(root_dir, 'dataset_' + phase) problem_path = os.path.join(root_dir, 'problem_' + phase) return dataset_path, problem_path def _search_pipeline(dataset, problem, template, input_dir, output_dir, budget, checkpoints, splits, db, tuner_type, test_id): dataset_path, problem_path = _get_dataset_paths(input_dir, dataset, 'TRAIN', problem) d3mds = D3MDS(dataset_path, problem_path) searcher = PipelineSearcher( output_dir, cv_splits=splits, db=db, tuner_type=tuner_type, test_id=test_id ) return searcher.search(d3mds, template, budget=budget, checkpoints=checkpoints) def _test_pipeline(dataset, problem, pipeline_id, input_dir, output_dir): dataset_path, problem_path = _get_dataset_paths(input_dir, dataset, 'TEST', problem) pipeline_path = os.path.join(output_dir, '{}.pkl'.format(pipeline_id)) with open(pipeline_path, 'rb') as pipeline_pkl: pipeline = cloudpickle.load(pipeline_pkl) print('Executing best pipeline {}'.format(pipeline)) d3mds = D3MDS(dataset_path, problem_path) with warnings.catch_warnings(): warnings.simplefilter('ignore') predictions = pipeline.predict(d3mds) return predictions def _score_predictions(dataset, problem, predictions, input_dir): dataset_path, problem_path = _get_dataset_paths(input_dir, dataset, 'TEST', problem) metric = _get_metric(problem_path) predictions = predictions.set_index('d3mIndex') targets = _load_targets(input_dir, dataset, problem)[predictions.columns] if len(targets.columns) > 1 or len(predictions.columns) > 1: raise Exception("I don't know how to handle these") if any(targets.index != predictions.index): raise Exception("Different indexes cannot be compared") targets = targets.iloc[:, 0] predictions = predictions.iloc[:, 0] score = encode_score(metric, targets, predictions) print("Score: {}".format(score)) summary = {'predictions': predictions, 'targets': targets} print(pd.DataFrame(summary).describe()) return score def _format_exception(e): error = '{}'.format(e.__class__.__name__) str_e = str(e) if str_e: error += ' - ' + str_e return error def _box_print(message): length = len(message) + 10 print(length * '#') print('#### {} ####'.format(message)) print(length * '#') def _insert_test(args, dataset): insert_ts = datetime.utcnow() document = { 'test_id': args.test_id, 'dataset': dataset, 'timeout': args.timeout, 'checkpoints': args.checkpoints, 'budget': args.budget, 'template': args.template, 'status': 'running', 'insert_ts': insert_ts, 'update_ts': insert_ts, 'version': VERSION, 'hostname': socket.gethostname(), 'tuner_type': args.tuner_type, 'splits': args.splits, } args.db.tests.insert_one(document) def _update_test(args, dataset, error, step): query = { 'test_id': args.test_id, 'dataset': dataset } update = { '$set': { 'status': 'error' if error else 'done', 'error': error, 'step': step, 'update_ts': datetime.utcnow() } } args.db.tests.update_one(query, update) def _insert_test_result(args, result): document = result.copy() document['test_id'] = args.test_id document['insert_ts'] = datetime.utcnow() args.db.test_results.insert_one(document) def _score_dataset(dataset, args): start_ts = datetime.utcnow() if args.db: _insert_test(args, dataset) result_base = { 'dataset': dataset, 'score': None, 'elapsed': None, 'iterations': None, 'error': None, 'step': None, 'load_time': None, 'trivial_time': None, 'cv_time': None, 'cv_score': None, 'rank': None } results = [] step = None error = None try: step = 'SEARCH' _box_print('Searching {}'.format(dataset)) # cleanup if not args.keep: shutil.rmtree(args.output, ignore_errors=True) search_results = _search_pipeline( dataset, args.problem, args.template, args.input, args.output, args.budget, args.checkpoints, args.splits, args.db, args.tuner_type, args.test_id ) gc.collect() for search_result in search_results or []: result = result_base.copy() result.update(search_result) results.append(result) pipeline = result['pipeline'] try: step = 'TEST' _box_print('Executing {}'.format(dataset)) predictions = _test_pipeline(dataset, args.problem, pipeline, args.input, args.output) step = 'SCORE' _box_print('Scoring {}'.format(dataset)) score = _score_predictions(dataset, args.problem, predictions, args.input) result['score'] = score gc.collect() except Exception as e: error = _format_exception(e) print("Scoring pipeline {} for dataset {} failed on step {} with error {}" .format(pipeline, dataset, step, error)) traceback.print_exc() result['error'] = error result['step'] = step if args.db: _insert_test_result(args, result) except Exception as e: error = _format_exception(e) print("Dataset {} failed on step {} with error {}".format(dataset, step, error)) traceback.print_exc() result_base['step'] = step result_base['error'] = error result_base['elapsed'] = (datetime.utcnow() - start_ts).total_seconds() results.append(result_base) if args.db: _update_test(args, dataset, error, step) return results def _prepare_search(args): make_keras_picklable() if not args.datasets and not args.all: print('Please provide at least one dataset name or add the --all option') sys.exit(1) args.datasets = _get_datasets(args) if args.db: args.db = get_db( config=args.db_config, database=args.db_name, host=args.db_host, port=args.db_port, user=args.db_user, password=args.db_password ) if args.checkpoints: args.checkpoints = [int(c) for c in args.checkpoints.split(',')] elif args.timeout: args.checkpoints = [args.timeout] if args.test_id is None: args.test_id = datetime.utcnow().strftime('%Y%m%d%H%M%S%f') def _score_datasets(args): if args.report and os.path.exists(args.report): report = pd.read_csv(args.report) else: report = pd.DataFrame(columns=['dataset']) for dataset, row in args.datasets.iterrows(): dataset_score = report[report.dataset == dataset] if dataset_score.empty or dataset_score.score.isnull().values[0]: if not dataset_score.empty: # clean-up report = report[report.dataset != dataset].copy() scores = _score_dataset(dataset, args) if scores: scores = pd.DataFrame(scores) scores = scores.merge(pd.DataFrame([row]), left_on='dataset', right_index=True) report = report.append(scores, ignore_index=True, sort=False) report = report.reindex(REPORT_COLUMNS, axis=1) if args.report: report.to_csv(args.report, index=False) return report def _search(args): _prepare_search(args) print("{} - Processing Datasets: {}".format(args.test_id, args.datasets.index.values)) report = _score_datasets(args) report = report.reindex(REPORT_COLUMNS, axis=1) columns = REPORT_COLUMNS[1:] print(report.set_index('dataset').to_string(columns=columns)) def _get_datasets(args): if args.all: datasets = [ d for d in os.listdir(args.input) if os.path.isdir(os.path.join(args.input, d)) ] else: datasets = args.datasets exclude = getattr(args, 'exclude', None) or [] datasets = [dataset for dataset in datasets if dataset not in exclude] try: summary = get_stats(datasets, args.input) except KeyError: print("No matching datasets found") sys.exit(1) summary = summary.set_index('dataset').reindex(datasets) summary = summary[~summary.data_modality.isnull()] for field in ['data_modality', 'task_type', 'task_subtype']: value = getattr(args, field) if value: summary = summary[summary[field] == value] if summary.empty: print("No matching datasets found") sys.exit(1) return summary REPORT_COLUMNS = [ 'dataset', 'pipeline', 'score', 'rank', 'cv_score', 'metric', 'data_modality', 'task_type', 'task_subtype', 'elapsed', 'iterations', 'load_time', 'trivial_time', 'cv_time', 'error', 'step' ] def _list(args): args.all = True datasets = _get_datasets(args) datasets = datasets.reset_index().sort_values('dataset').set_index('dataset') columns = [ 'data_modality', 'task_type', 'task_subtype', 'metric', 'size_human', 'train_samples' ] datasets = datasets.reindex(columns, axis=1) if args.report: print("Storing datasets as {}".format(args.report)) datasets[columns].to_csv(args.report, index=True) else: print(datasets.to_string(columns=columns, index=True)) class ArgumentParser(argparse.ArgumentParser): def convert_arg_line_to_args(self, arg_line): return arg_line.split() def _path_type(string): try: return make_abs(string) except ValueError: error = "Not a valid path: '{0}'.".format(string) raise argparse.ArgumentTypeError(error) def _get_parser(): # Logging logging_args = ArgumentParser(add_help=False) logging_args.add_argument('-v', '--verbose', action='count', default=0) logging_args.add_argument('-l', '--logfile') # Report report_args = ArgumentParser(add_help=False) report_args.add_argument('-r', '--report', type=_path_type, help='Store results in the given CVS file.') # Dataset Selection dataset_args = ArgumentParser(add_help=False) dataset_args.add_argument('-i', '--input', default='input', type=_path_type, help='Input datasets folder. Defaults to `input`.') dataset_args.add_argument('-o', '--output', type=_path_type, help='Output pipelines folder. Defaults to `output`.', default='output') dataset_args.add_argument('-p', '--problem', default='', help='Problem suffix. Only needed if the dataset has more than one.') dataset_args.add_argument('-M', '--data-modality', type=str, help='Only process datasets of the given Data Modality.') dataset_args.add_argument('-T', '--task-type', type=str, help='Only process datasets of the given Task type') dataset_args.add_argument('-S', '--task-subtype', type=str, help='Only process datasets of the given Task Subtype') # Search Configuration search_args = ArgumentParser(add_help=False) search_args.add_argument('-b', '--budget', type=int, help='If given, maximum number tuning iterations to perform.') search_args.add_argument('-s', '--splits', type=int, default=5, help='Number of Cross Validation Folds. Defaults to 5') search_args.add_argument('-c', '--checkpoints', help=('Comma separated list of time checkpoints in seconds where ' 'the best pipeline so far will be dumped and stored.')) search_args.add_argument('-t', '--timeout', type=int, help='Timeout in seconds. Ignored if checkpoints are given.') search_args.add_argument('-u', '--tuner-type', default='gp', choices=TUNERS.keys(), help='Type of tuner to use. Defaults to "gp"') search_args.add_argument('--template', help='Template to use. If not given, use the most appropriate one.') search_args.add_argument('-e', '--exclude', nargs='+', help='Exclude these datasets. Useful in combination with --all.') search_args.add_argument('-a', '--all', action='store_true', help='Process all the datasets found in the input folder.') search_args.add_argument('-k', '--keep', action='store_true', help='Keep previous results in the output folder.') search_args.add_argument('--test-id', help='test_id associated with this run.') search_args.add_argument('datasets', nargs='*', help='Datasets to process. Ignored if --all use used.') # Backend configuration db_args = ArgumentParser(add_help=False) db_args.add_argument('--db', action='store_true', help='Use a MongoDB backend to store the results.') db_args.add_argument('--db-config', help='MongoDB configuraiton JSON file.') db_args.add_argument('--db-host', default='localhost') db_args.add_argument('--db-port', default=27017, type=int) db_args.add_argument('--db-name', default='autobazaar') db_args.add_argument('--db-user') db_args.add_argument('--db-password') parser = ArgumentParser( prog='autobazaar', description='AutoBazaar Experiments Suite', fromfile_prefix_chars='@', parents=[logging_args] ) parser.add_argument('--version', action='version', version='%(prog)s {version}'.format(version=VERSION)) subparsers = parser.add_subparsers(title='command', help='Command to execute') parser.set_defaults(command=None) list_ = subparsers.add_parser('list', parents=[logging_args, dataset_args, report_args], help='List the available datasets that match the conditions.') list_.set_defaults(command=_list) search_parents = [ logging_args, dataset_args, search_args, report_args, db_args ] search_ = subparsers.add_parser('search', parents=search_parents, help='Search the best pipeline for the given datasets.') search_.set_defaults(command=_search) return parser def main(): parser = _get_parser() args = parser.parse_args() if not args.command: parser.print_help() parser.exit() logging_setup(args.verbose, args.logfile) gc.enable() args.command(args) if __name__ == '__main__': main()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/autobazaar/utils.py
training/helpers/autobazaar/autobazaar/utils.py
# -*- coding: utf-8 -*- import os import tempfile from collections import defaultdict from datetime import datetime import numpy as np from sklearn.preprocessing import LabelEncoder def encode_score(scorer, expected, observed): if expected.dtype == 'object': le = LabelEncoder() expected = le.fit_transform(expected) observed = le.transform(observed) return scorer(expected, observed) def ensure_dir(directory): """Create diretory if it does not exist yet.""" if not os.path.exists(directory): os.makedirs(directory) def make_dumpable(params, datetimes=False): """Get nested dicts of params to allow json dumping. Also work around this: https://github.com/HDI-Project/BTB/issues/79 And convert numpy types to primitive types. Optionally dump datetimes to ISO format. Args: params (dict): Params dictionary with tuples as keys. datetimes (bool): whether to convert datetimes to ISO strings or not. Returns: dict: Dumpable params as a tree of dicts and nested sub-dicts. """ nested_params = defaultdict(dict) for (block, param), value in params.items(): if isinstance(value, np.integer): value = int(value) elif isinstance(value, np.floating): value = float(value) elif isinstance(value, np.ndarray): value = value.tolist() elif isinstance(value, np.bool_): value = bool(value) elif value == 'None': value = None elif datetimes and isinstance(value, datetime): value = value.isoformat() nested_params[block][param] = value return nested_params def _walk(document, transform): if not isinstance(document, dict): return document new_doc = dict() for key, value in document.items(): if isinstance(value, dict): value = _walk(value, transform) elif isinstance(value, list): value = [_walk(v, transform) for v in value] new_key, new_value = transform(key, value) new_doc[new_key] = new_value return new_doc def remove_dots(document): """Replace dots with dashes in all the keys from the dictionary.""" return _walk(document, lambda key, value: (key.replace('.', '-'), value)) def restore_dots(document): """Replace dashes with dots in all the keys from the dictionary.""" return _walk(document, lambda key, value: (key.replace('-', '.'), value)) def make_keras_picklable(): """Make the keras models picklable.""" import keras.models # noqa: lazy import slow dependencies def __getstate__(self): model_str = "" with tempfile.NamedTemporaryFile(suffix='.hdf5', delete=True) as fd: keras.models.save_model(self, fd.name, overwrite=True) model_str = fd.read() return {'model_str': model_str} def __setstate__(self, state): with tempfile.NamedTemporaryFile(suffix='.hdf5', delete=True) as fd: fd.write(state['model_str']) fd.flush() try: model = keras.models.load_model(fd.name) except ValueError: from keras.applications import mobilenet from keras.utils.generic_utils import CustomObjectScope scope = { 'relu6': mobilenet.relu6, 'DepthwiseConv2D': mobilenet.DepthwiseConv2D } with CustomObjectScope(scope): model = keras.models.load_model(fd.name) self.__dict__ = model.__dict__ cls = keras.models.Model cls.__getstate__ = __getstate__ cls.__setstate__ = __setstate__
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/autobazaar/pipeline.py
training/helpers/autobazaar/autobazaar/pipeline.py
# -*- coding: utf-8 -*- """AutoBazaar Pipeline Module.""" import json import logging import os import random import uuid from collections import Counter import cloudpickle import numpy as np import pandas as pd from mit_d3m.loaders import get_loader from mit_d3m.metrics import METRICS_DICT from mlblocks import MLPipeline from autobazaar.utils import encode_score LOGGER = logging.getLogger(__name__) class ABPipeline(object): """AutoBazaar Pipeline Class.""" def _extract_hyperparameters(self, preprocessing_primitives): block_names_count = Counter() block_names = list() for primitive in preprocessing_primitives: block_names_count.update([primitive]) block_count = block_names_count[primitive] block_names.append('{}#{}'.format(primitive, block_count)) pre_params = dict() hyperparameters = self.pipeline_dict['hyperparameters'].copy() for block_name in block_names: block_params = hyperparameters.pop(block_name, None) if block_params: pre_params[block_name] = block_params return pre_params, hyperparameters def __init__(self, pipeline_dict, loader, metric, problem_doc): self.pipeline_dict = pipeline_dict self.name = pipeline_dict['name'] self.template = pipeline_dict.get('template') self.loader = loader self.metric = metric self.problem_doc = problem_doc preprocessing_blocks = self.pipeline_dict.get('preprocessing_blocks') if preprocessing_blocks: preprocessing = pipeline_dict.copy() preprocessing_primitives = preprocessing['primitives'][:preprocessing_blocks] preprocessing['primitives'] = preprocessing_primitives self._preprocessing = preprocessing tunable = pipeline_dict.copy() tunable_primitives = tunable['primitives'][preprocessing_blocks:] tunable['primitives'] = tunable_primitives self._tunable = tunable pre_params, tun_params = self._extract_hyperparameters(preprocessing_primitives) self._preprocessing['hyperparameters'] = pre_params self._tunable['hyperparameters'] = tun_params else: self._preprocessing = None self._tunable = pipeline_dict self.id = str(uuid.uuid4()) self.cv_scores = list() self.rank = None self.score = None self.dumped = False self.fitted = False self.pipeline = MLPipeline.from_dict(pipeline_dict) def fit(self, data_params): """Fit the pipeline on the given params.""" X, y = data_params.X, data_params.y self.pipeline = MLPipeline.from_dict(self.pipeline_dict) self.pipeline.fit(X, y, **data_params.context) self.fitted = True def predict(self, d3mds): """Get predictions for the given D3MDS.""" data_params = self.loader.load(d3mds) predictions = self.pipeline.predict(data_params.X, **data_params.context) out_df = pd.DataFrame() out_df['d3mIndex'] = data_params.y.index out_df[d3mds.target_column] = predictions return out_df def _get_split(self, X, y, indexes): if hasattr(X, 'iloc'): X = X.iloc[indexes] else: X = X[indexes] if y is not None: if hasattr(y, 'iloc'): y = y.iloc[indexes] else: y = y[indexes] return X, y def _get_score(self): score = np.mean(self.cv_scores) std = np.std(self.cv_scores) if 'Error' in self.metric: rank = score elif score <= 1: rank = 1 - score else: raise ValueError("Found a score > 1 in a maximization problem: {}".format(score)) return score, std, rank def preprocess(self, X, y, context): """Execute the preprocessing steps of the pipeline.""" if self._preprocessing: LOGGER.info("Executing preprocessing pipeline") pipeline = MLPipeline.from_dict(self._preprocessing) pipeline.fit(X, y, **context) return pipeline.predict(X, **context) else: LOGGER.info("No preprocessing steps found") return X def cv_score(self, X, y, context, metric=None, cv=None): """Cross Validate this pipeline.""" scorer = METRICS_DICT[metric or self.metric] LOGGER.debug('CV Scoring pipeline %s', self) self.cv_scores = list() for fold, (train_index, test_index) in enumerate(cv.split(X, y)): LOGGER.debug('Scoring fold: %s', fold) X_train, y_train = self._get_split(X, y, train_index) X_test, y_test = self._get_split(X, y, test_index) pipeline = MLPipeline.from_dict(self._tunable) pipeline.fit(X_train, y_train, **context) pred = pipeline.predict(X_test, **context) score = encode_score(scorer, y_test, pred) self.cv_scores.append(score) LOGGER.debug('Fold %s score: %s', fold, score) score, std, rank = self._get_score() LOGGER.debug('CV score: %s +/- %s; rank: %s', score, std, rank) self.score = score self.std = std self.rank = rank + random.random() * 1.e-12 # to avoid collisions def to_dict(self, problem_doc=False): """Return the details of this pipeline in a dict.""" pipeline_dict = self.pipeline.to_dict().copy() pipeline_dict.update({ 'id': self.id, 'name': self.name, 'template': self.template, 'loader': self.loader.to_dict(), 'score': self.score, 'rank': self.rank, 'metric': self.metric }) if problem_doc: pipeline_dict['problem_doc'] = self.problem_doc return pipeline_dict def __repr__(self): return 'ABPipeline({})'.format(json.dumps(self.to_dict(), indent=4)) @classmethod def from_dict(cls, pipeline_dict): """Load a pipeline from a dict.""" pipeline_dict = pipeline_dict.copy() loader = get_loader(**pipeline_dict.pop('loader')) metric = pipeline_dict['metric'] problem_doc = pipeline_dict.pop('problem_doc') return cls(pipeline_dict, loader, metric, problem_doc) def dump(self, output_dir, rank=None): """Dump this pipeline using pickle.""" if rank is None: rank = self.rank LOGGER.info('Dumping pipeline with rank %s: %s', rank, self.id) self.dumped = True pickle_path = os.path.join(output_dir, '{}.pkl'.format(self.id)) with open(pickle_path, "wb") as pickle_file: LOGGER.info("Outputting pipeline %s", pickle_file.name) cloudpickle.dump(self, pickle_file) json_path = os.path.join(output_dir, '{}.json'.format(self.id)) self.pipeline.save(json_path)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/autobazaar/__init__.py
training/helpers/autobazaar/autobazaar/__init__.py
# -*- coding: utf-8 -*- """ AutoBazaar top module. AutoBazaar is an AutoML system created to execute the experiments associated with the [The Machine Learning Bazaar Paper: Harnessing the ML Ecosystem for Effective System Development](https://arxiv.org/pdf/1905.08942.pdf) by the [Human-Data Interaction (HDI) Project](https://hdi-dai.lids.mit.edu/) at LIDS, MIT. * Free software: MIT license * Documentation: https://HDI-Project.github.io/AutoBazaar """ import os import git __author__ = 'MIT Data To AI Lab' __copyright__ = 'Copyright (c) 2019, MIT Data To AI Lab' __email__ = 'dailabmit@gmail.com' __license__ = 'MIT' __version__ = '0.2.1-dev' def _get_commit(): try: base_path = os.path.dirname(__file__) repo = git.Repo(base_path, search_parent_directories=True) commit = repo.commit().hexsha[0:7] if repo.is_dirty(untracked_files=False): commit += '*' return commit except git.InvalidGitRepositoryError: return None def get_version(): commit = _get_commit() if commit: return '{} - {}'.format(__version__, commit) return __version__
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autobazaar/docs/conf.py
training/helpers/autobazaar/docs/conf.py
# -*- coding: utf-8 -*- # # AutoBazaar documentation build configuration file, created by # sphinx-quickstart on Fri Jan 6 13:06:48 2017. # # This file is execfile()d with the current directory set to its # containing dir. # # Note that not all possible configuration values are present in this # autogenerated file. # # All configuration values have a default; values that are commented out # serve to show the default. # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. import sphinx_rtd_theme # For read the docs theme import autobazaar # -- General configuration --------------------------------------------- # If your documentation needs a minimal Sphinx version, state it here. # # needs_sphinx = '1.0' # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom ones. extensions = [ 'm2r', 'sphinx.ext.autodoc', 'sphinx.ext.githubpages', 'sphinx.ext.viewcode', 'sphinx.ext.napoleon', 'autodocsumm', ] autodoc_default_options = { 'autosummary': True, } # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix(es) of source filenames. # You can specify multiple suffix as a list of string: source_suffix = ['.rst', '.md'] # The master toctree document. master_doc = 'index' # General information about the project. project = 'AutoBazaar' slug = 'autobazaar' title = project + ' Documentation' copyright = '2019, MIT Data to AI Lab' author = 'Carles Sala, Micah Smith, Max Kanter, Kalyan Veeramachaneni' description = 'The Machine Learning Bazaar Experiments' user = 'HDI-project' # The version info for the project you're documenting, acts as replacement # for |version| and |release|, also used in various other places throughout # the built documents. # # The short X.Y version. version = autobazaar.__version__ # The full version, including alpha/beta/rc tags. release = autobazaar.__version__ # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. # # This is also used if you do content translation via gettext catalogs. # Usually you set "language" from the command line for these cases. language = None # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. # This patterns also effect to html_static_path and html_extra_path exclude_patterns = ['.py', '_build', 'Thumbs.db', '.DS_Store', '**.ipynb_checkpoints'] # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' # If true, `todo` and `todoList` produce output, else they produce nothing. todo_include_todos = False # -- Options for HTML output ------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # html_theme = 'sphinx_rtd_theme' html_theme_path = [sphinx_rtd_theme.get_html_theme_path()] # Readthedocs additions html_context = { 'display_github': True, 'github_user': user, 'github_repo': project, 'github_version': 'master', 'conf_py_path': '/docs/', } # Theme options are theme-specific and customize the look and feel of a # theme further. For a list of options available for each theme, see the # documentation. html_theme_options = { 'collapse_navigation': False, 'display_version': False, } # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". # The name of an image file (relative to this directory) to use as a favicon of # the docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32 # pixels large. html_favicon = 'images/dai-logo-white.ico' html_logo = 'images/dai-logo-white-200.png' # -- Options for HTMLHelp output --------------------------------------- # Output file base name for HTML help builder. htmlhelp_basename = slug + 'doc' # -- Options for LaTeX output ------------------------------------------ latex_elements = { } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, author, documentclass # [howto, manual, or own class]). latex_documents = [( master_doc, slug + '.tex', title, author, 'manual' )] # -- Options for manual page output ------------------------------------ # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [( master_doc, slug, title, [author], 1 )] # -- Options for Texinfo output ---------------------------------------- # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [( master_doc, slug, title, author, slug, description, 'Miscellaneous' )]
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/archived/train_autobazaar_old.py
training/helpers/archived/train_autobazaar_old.py
import warnings, datetime, uuid, os, json, shutil, pickle, random warnings.filterwarnings('ignore') from sklearn.model_selection import train_test_split from sklearn.metrics import accuracy_score from sklearn.metrics import r2_score, mean_squared_error from sklearn.ensemble import RandomForestRegressor from sklearn.ensemble import ExtraTreesRegressor from sklearn.model_selection import cross_val_score from sklearn.metrics import make_scorer import pandas as pd import csv, io ''' Taken from the example here: https://github.com/HDI-Project/BTB/blob/master/notebooks/BTBSession%20-%20Example.ipynb Note that autobazaar is used as the primary model trainer for BTB sessions. https://github.com/HDI-Project/AutoBazaar Tutorial: https://hdi-project.github.io/AutoBazaar/readme.html#install Data: Must be formatted (https://github.com/mitll/d3m-schema/blob/master/documentation/datasetSchema.md) Case 1: Single table In many openml and other tabular cases, all the learning data is contained in a single tabular file. In this case, an example dataset will look like the following. ─ 196_autoMpg ├── 196_autoMpg_dataset │   ├── datasetDoc.json │   └── tables │   └── learningData.csv ├── 196_autoMpg_problem │   ├── dataSplits.csv │   └── problemDoc.json ├── SCORE │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   ├── problem_TEST │   │   ├── dataSplits.csv │   │   └── problemDoc.json │   └── targets.csv ├── TEST │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   └── problem_TEST │   ├── dataSplits.csv │   └── problemDoc.json └── TRAIN ├── dataset_TRAIN │   ├── datasetDoc.json │   └── tables │   └── learningData.csv └── problem_TRAIN ├── dataSplits.csv └── problemDoc.json ''' def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ def convert_(X_train, y_train, labels): feature_list=labels data=dict() print(len(feature_list)) print(len(X_train[0])) # time.sleep(50) indices=list() for i in range(len(X_train)): indices.append(i) for j in range(len(feature_list)-1): if i > 0: # print(data[feature_list[j]]) try: # print(feature_list[j]) # print(data) # print(X_train[i][j]) # print(data[feature_list[j]]) # time.sleep(2) data[feature_list[j]]=data[feature_list[j]]+[X_train[i][j]] except: pass # print(data) # time.sleep(50) # print(str(i)+'-i') # print(j) else: data[feature_list[j]]=[X_train[i][j]] print(data) data['class_']=y_train data['d3mIndex']=indices data=pd.DataFrame(data, columns = list(data)) print(data) print(list(data)) # time.sleep(500) return data def split_data(data): # get training and testing numbers train_num=int(0.80*len(data)) test_num=len(data)-train_num print('TRAINING SAMPLES') print(train_num) print('TESTING SAMPLES') print(test_num) # now write the rows rows=list() train_count=0 test_count=0 train_rows=list() test_rows=list() i=0 totalcount=train_num+test_num # randomize the numbers of i i_list=list() for i in range(totalcount): i_list.append(i) random.shuffle(i_list) for i in range(len(i_list)): if train_num > train_count: rows.append([i_list[i], 'TRAIN', 0, 0]) train_rows.append(i_list[i]) i=i+1 train_count=train_count+1 print(train_count) # print(train_num) elif test_num > test_count: rows.append([i_list[i], 'TEST', 0, 0]) test_rows.append(i_list[i]) i=i+1 test_count=test_count+1 # print(len(rows)) print([test_num, test_count, train_num, train_count]) # field names fields = ['d3mIndex', 'type', 'repeat', 'fold'] # name of csv file filename = "dataSplits.csv" # writing to csv file with open(filename, 'w') as csvfile: # creating a csv writer object csvwriter = csv.writer(csvfile) # writing the fields csvwriter.writerow(fields) # writing the data rows csvwriter.writerows(rows) # now split this data into another csv print(train_rows) train_data=data.iloc[train_rows,:] train_data.to_csv('train.csv') print(test_rows) test_data=data.iloc[test_rows,:] test_data.to_csv('test.csv') return filename def create_dataset_json(foldername, trainingcsv): # create the template .JSON file necessary for the featurization dataset_name=foldername dataset_id="%s_dataset"%(foldername) columns=list() colnames=list(pd.read_csv(trainingcsv)) for i in range(len(colnames)): if colnames[i] != 'class_': columns.append({"colIndex": i, "colName": colnames[i], "colType": "real", "role": ["attribute"]}) else: columns.append({"colIndex": i, "colName": 'class_', "colType": "real", "role": ["suggestedTarget"]}) i1=i data={"about": { "datasetID": dataset_id, "datasetName":dataset_name, "humanSubjectsResearch": False, "license":"CC", "datasetSchemaVersion":"3.0", "redacted":False }, "dataResources": [ { "resID": "0", "resPath": 'tables/learningData.csv', "resType": "table", "resFormat": ["text/csv"], "isCollection": False, "columns":columns, } ] } filename='datasetDoc.json' jsonfile=open(filename,'w') json.dump(data,jsonfile) jsonfile.close() return dataset_id, filename, i1 def create_problem_json(mtype, folder,i1): if mtype == 'classification': data = { "about": { "problemID": "%s_problem"%(folder), "problemName": "%s_problem"%(folder), "problemDescription": "not applicable", "taskType": "classification", "taskSubType": "multiClass", "problemVersion": "1.0", "problemSchemaVersion": "3.0" }, "inputs": { "data": [ { "datasetID": "%s"%(folder), "targets": [ { "targetIndex": 0, "resID": "0", "colIndex": i1, "colName": 'class_', } ] } ], "dataSplits": { "method": "holdOut", "testSize": 0.2, "stratified": True, "numRepeats": 0, "randomSeed": 42, "splitsFile": "dataSplits.csv" }, "performanceMetrics": [ { "metric": "accuracy" } ] }, "expectedOutputs": { "predictionsFile": "predictions.csv" } } elif mtype == 'regression': data={"about": { "problemID": "%s_problem"%(folder), "problemName": "%s_problem"%(folder), "problemDescription": "not applicable", "taskType": "regression", "taskSubType": "univariate", "problemVersion": "1.0", "problemSchemaVersion": "3.0" }, "inputs": { "data": [ { "datasetID": "%s_dataset"%(folder), "targets": [ { "targetIndex": 0, "resID": "0", "colIndex": i1, "colName": "class_" } ] } ], "dataSplits": { "method": "holdOut", "testSize": 0.2, "stratified": True, "numRepeats": 0, "randomSeed": 42, "splitsFile": "dataSplits.csv" }, "performanceMetrics": [ { "metric": "meanSquaredError" } ] }, "expectedOutputs": { "predictionsFile": "predictions.csv" } } jsonfile=open('problemDoc.json','w') json.dump(data,jsonfile) jsonfile.close() def train_autobazaar(alldata, labels, mtype, jsonfile, problemtype, default_features, settings): print('installing package configuration') # curdir=os.getcwd() # os.chdir(prev_dir(curdir)+'/training/helpers/autobazaar') # os.system('make install-develop') # os.chdir(curdir) # create file names model_name=jsonfile[0:-5]+'_'+str(default_features).replace("'",'').replace('"','')+'_btb' if mtype == 'c': model_name=model_name+'_classification' mtype='classification' elif mtype == 'r': model_name=model_name+'_regression' mtype='regression' folder=model_name jsonfilename=model_name+'.json' csvfilename=model_name+'.csv' model_name=model_name+'.pickle' # this should be the model directory hostdir=os.getcwd() # open a sample featurization labels_dir=prev_dir(hostdir)+'/train_dir/'+jsonfilename.split('_')[0] os.chdir(labels_dir) listdir=os.listdir() features_file='' for i in range(len(listdir)): if listdir[i].endswith('.json'): features_file=listdir[i] # load features file and get labels labels_=json.load(open(features_file))['features'][problemtype][default_features]['labels'] os.chdir(hostdir) # make a temporary folder for the training session try: os.mkdir(folder) os.chdir(folder) except: shutil.rmtree(folder) os.mkdir(folder) os.chdir(folder) # make the data arrays print('creating training data...') all_data = convert_(alldata, labels, labels_) all_data.to_csv(csvfilename,index=False) data=pd.read_csv(csvfilename) # create required .JSON files dataset_id, dataset_filename, i1=create_dataset_json(folder, csvfilename) problem_filename=create_problem_json(mtype, folder, i1) split_data(data) # get the current directory abz_dir=os.getcwd() # make necessary directories # now create proper tree structure ''' ─ 196_autoMpg ├── 196_autoMpg_dataset │   ├── datasetDoc.json │   └── tables │   └── learningData.csv ├── 196_autoMpg_problem │   ├── dataSplits.csv │   └── problemDoc.json ├── SCORE │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   ├── problem_TEST │   │   ├── dataSplits.csv │   │   └── problemDoc.json │   └── targets.csv ├── TEST │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   └── problem_TEST │   ├── dataSplits.csv │   └── problemDoc.json └── TRAIN ├── dataset_TRAIN │   ├── datasetDoc.json │   └── tables │   └── learningData.csv └── problem_TRAIN ├── dataSplits.csv └── problemDoc.json ''' dataset_folder=folder+'_dataset' problem_folder=folder+'_problem' # make datasets folder os.mkdir(dataset_folder) os.chdir(dataset_folder) os.mkdir('tables') shutil.copy(abz_dir+'/datasetDoc.json', os.getcwd()+'/datasetDoc.json') shutil.copy(abz_dir+'/'+csvfilename, os.getcwd()+'/tables/'+csvfilename) os.chdir('tables') os.rename(csvfilename, 'learningData.csv') # make problem folder os.chdir(abz_dir) os.mkdir(problem_folder) os.chdir(problem_folder) shutil.copy(abz_dir+'/problemDoc.json', os.getcwd()+'/problemDoc.json') shutil.copy(abz_dir+'/dataSplits.csv', os.getcwd()+'/dataSplits.csv') os.chdir(abz_dir) os.mkdir('TEST') os.chdir('TEST') os.mkdir('dataset_TEST') shutil.copy(abz_dir+'/'+dataset_folder+'/datasetDoc.json', os.getcwd()+'/dataset_TEST/datasetDoc.json') os.mkdir('problem_TEST') shutil.copy(abz_dir+'/'+problem_folder+'/problemDoc.json',os.getcwd()+'/problem_TEST/problemDoc.json') shutil.copy(abz_dir+'/'+problem_folder+'/dataSplits.csv', os.getcwd()+'/problem_TEST/dataSplits.csv') os.chdir('dataset_TEST') os.mkdir('tables') shutil.copy(abz_dir+'/test.csv', os.getcwd()+'/tables/test.csv') os.chdir('tables') os.rename('test.csv', 'learningData.csv') os.chdir(abz_dir) os.mkdir('TRAIN') os.chdir('TRAIN') os.mkdir('dataset_TRAIN') os.chdir('dataset_TRAIN') os.mkdir('tables') shutil.copy(abz_dir+'/datasetDoc.json', os.getcwd()+'/datasetDoc.json') shutil.copy(abz_dir+'/train.csv', os.getcwd()+'/tables/train.csv') os.chdir('tables') os.rename('train.csv','learningData.csv') os.chdir(abz_dir+'/TRAIN') os.mkdir('problem_TRAIN') shutil.copy(abz_dir+'/'+problem_folder+'/problemDoc.json',os.getcwd()+'/problem_TRAIN/problemDoc.json') shutil.copy(abz_dir+'/'+problem_folder+'/dataSplits.csv', os.getcwd()+'/problem_TRAIN/dataSplits.csv') os.chdir(abz_dir) os.mkdir('SCORE') os.chdir('SCORE') shutil.copytree(abz_dir+'/TEST/dataset_TEST',os.getcwd()+'/dataset_SCORE') shutil.copytree(abz_dir+'/TEST/problem_TEST', os.getcwd()+'/problem_SCORE') os.chdir(hostdir) # this works for really any input configuration - regression or classificatoin (as this is covered in config files) try: os.mkdir('input') except: pass # remove if file exists try: shutil.copytree(folder, os.getcwd()+'/input/'+folder) except: shutil.rmtree(os.getcwd()+'/input/'+folder) shutil.copytree(folder, os.getcwd()+'/input/'+folder) os.system('abz search %s -c20,30,40 -b10'%(folder)) # now go to output folder os.chdir('output') listdir=os.listdir() for i in range(len(listdir)): if listdir[i].endswith('.json'): g=json.load(open(listdir[i])) # os.remove(listdir[i]) elif listdir[i].endswith('.pkl'): picklefile=folder+'.pickle' shutil.copy(os.getcwd()+'/'+listdir[i],os.getcwd()+'/'+folder+'.pickle') model=pickle.load(open(picklefile, 'rb')) # load some training data in X_train, X_test, y_train, y_test = train_test_split(alldata, labels, train_size=0.750, test_size=0.250) # make some predictions and get accuracy measure if mtype=='classification': y_pred=model.predict(X_test) accuracy=accuracy_score(y_test, y_pred) data={'sample type': problemtype, 'feature_set':default_features, 'model name':picklefile, 'training params': g, 'accuracy': float(accuracy), 'model_type': 'autobazaar_%s'%(mtype), 'settings': settings, 'training params': g} elif mtype=='regression': y_pred=model.predict(X_test) mse_error=mean_squared_error(y_true, y_pred) data={'sample type': problemtype, 'feature_set':default_features, 'model name':picklefile, 'training params': g, 'mse_error': float(mse_error), 'model_type': 'autobazaar_%s'%(mtype), 'settings': settings, 'training params': g} jsonfile=open(folder+'.json','w') json.dump(data,jsonfile) jsonfile.close() # now get all them transferred os.chdir(hostdir) try: os.chdir(problemtype+'_models') except: os.mkdir(problemtype+'_models') os.chdir(problemtype+'_models') # copy necessary files shutil.copy(hostdir+'/output/'+picklefile, os.getcwd()+'/'+picklefile) shutil.copy(hostdir+'/output/'+jsonfile, os.getcwd()+'/'+jsonfile) # delete inactive directories os.chdir(hostdir) shutil.rmtree('input') shutil.rmtree('output') shutil.rmtree(folder) # go back to model directory os.chdir(problemtype+'_models') # get variables model_dir=hostdir+'/%s_models/'%(problemtype) model_name=picklefile # finally done! Whew - what a lot of data transformations here to get this to work return model_name, model_dir
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/archived/train_autobazaar.py
training/helpers/archived/train_autobazaar.py
import warnings, datetime, uuid, os, json, shutil, pickle, random warnings.filterwarnings('ignore') from sklearn.model_selection import train_test_split from sklearn.metrics import accuracy_score from sklearn.metrics import r2_score, mean_squared_error from sklearn.ensemble import RandomForestRegressor from sklearn.ensemble import ExtraTreesRegressor from sklearn.model_selection import cross_val_score from sklearn.metrics import make_scorer import pandas as pd import csv, io, time def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ print('installing package configuration') curdir=os.getcwd() os.chdir(prev_dir(curdir)+'/training/helpers/autobazaar') os.system('make install-develop') os.chdir(curdir) ''' Taken from the example here: https://github.com/HDI-Project/BTB/blob/master/notebooks/BTBSession%20-%20Example.ipynb Note that autobazaar is used as the primary model trainer for BTB sessions. https://github.com/HDI-Project/AutoBazaar Tutorial: https://hdi-project.github.io/AutoBazaar/readme.html#install Data: Must be formatted (https://github.com/mitll/d3m-schema/blob/master/documentation/datasetSchema.md) Case 1: Single table In many openml and other tabular cases, all the learning data is contained in a single tabular file. In this case, an example dataset will look like the following. ─ 196_autoMpg ├── 196_autoMpg_dataset │   ├── datasetDoc.json │   └── tables │   └── learningData.csv ├── 196_autoMpg_problem │   ├── dataSplits.csv │   └── problemDoc.json ├── SCORE │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   ├── problem_TEST │   │   ├── dataSplits.csv │   │   └── problemDoc.json │   └── targets.csv ├── TEST │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   └── problem_TEST │   ├── dataSplits.csv │   └── problemDoc.json └── TRAIN ├── dataset_TRAIN │   ├── datasetDoc.json │   └── tables │   └── learningData.csv └── problem_TRAIN ├── dataSplits.csv └── problemDoc.json ''' def convert_(X_train, y_train, labels): feature_list=labels data=dict() print(len(feature_list)) print(len(X_train[0])) # time.sleep(50) indices=list() for i in range(len(X_train)): indices.append(i) for j in range(len(feature_list)-1): if i > 0: try: data[feature_list[j]]=data[feature_list[j]]+[X_train[i][j]] except: pass else: data[feature_list[j]]=[X_train[i][j]] print(data) data['class_']=y_train data['d3mIndex']=indices data=pd.DataFrame(data, columns = list(data)) print(data) print(list(data)) return data def split_data(data): # get training and testing numbers train_num=int(0.80*len(data)) test_num=len(data)-train_num print('TRAINING SAMPLES') print(train_num) print('TESTING SAMPLES') print(test_num) # now write the rows rows=list() train_count=0 test_count=0 train_rows=list() test_rows=list() i=0 totalcount=train_num+test_num # randomize the numbers of i i_list=list() for i in range(totalcount): i_list.append(i) random.shuffle(i_list) for i in range(len(i_list)): if train_num > train_count: rows.append([i_list[i], 'TRAIN', 0, 0]) train_rows.append(i_list[i]) i=i+1 train_count=train_count+1 print(train_count) # print(train_num) elif test_num > test_count: rows.append([i_list[i], 'TEST', 0, 0]) test_rows.append(i_list[i]) i=i+1 test_count=test_count+1 # print(len(rows)) print([test_num, test_count, train_num, train_count]) # field names fields = ['d3mIndex', 'type', 'repeat', 'fold'] # name of csv file filename = "dataSplits.csv" # writing to csv file with open(filename, 'w') as csvfile: # creating a csv writer object csvwriter = csv.writer(csvfile) # writing the fields csvwriter.writerow(fields) # writing the data rows csvwriter.writerows(rows) # now split this data into another csv print(train_rows) train_data=data.iloc[train_rows,:] train_data.to_csv('train.csv') print(test_rows) test_data=data.iloc[test_rows,:] test_data.to_csv('test.csv') return filename def create_dataset_json(foldername, trainingcsv): # create the template .JSON file necessary for the featurization dataset_name=foldername dataset_id="%s_dataset"%(foldername) columns=list() colnames=list(pd.read_csv(trainingcsv)) for i in range(len(colnames)): if colnames[i] != 'class_': columns.append({"colIndex": i, "colName": colnames[i], "colType": "real", "role": ["attribute"]}) else: columns.append({"colIndex": i, "colName": 'class_', "colType": "real", "role": ["suggestedTarget"]}) i1=i data={"about": { "datasetID": dataset_id, "datasetName":dataset_name, "humanSubjectsResearch": False, "license":"CC", "datasetSchemaVersion":"3.0", "redacted":False }, "dataResources": [ { "resID": "0", "resPath": 'tables/learningData.csv', "resType": "table", "resFormat": ["text/csv"], "isCollection": False, "columns":columns, } ] } filename='datasetDoc.json' jsonfile=open(filename,'w') json.dump(data,jsonfile) jsonfile.close() return dataset_id, filename, i1 def create_problem_json(mtype, folder,i1): if mtype == 'c': data = { "about": { "problemID": "%s_problem"%(folder), "problemName": "%s_problem"%(folder), "problemDescription": "not applicable", "taskType": "classification", "taskSubType": "multiClass", "problemVersion": "1.0", "problemSchemaVersion": "3.0" }, "inputs": { "data": [ { "datasetID": "%s"%(folder), "targets": [ { "targetIndex": 0, "resID": "0", "colIndex": i1, "colName": 'class_', } ] } ], "dataSplits": { "method": "holdOut", "testSize": 0.2, "stratified": True, "numRepeats": 0, "randomSeed": 42, "splitsFile": "dataSplits.csv" }, "performanceMetrics": [ { "metric": "accuracy" } ] }, "expectedOutputs": { "predictionsFile": "predictions.csv" } } elif mtype == 'r': data={"about": { "problemID": "%s_problem"%(folder), "problemName": "%s_problem"%(folder), "problemDescription": "not applicable", "taskType": "regression", "taskSubType": "univariate", "problemVersion": "1.0", "problemSchemaVersion": "3.0" }, "inputs": { "data": [ { "datasetID": "%s_dataset"%(folder), "targets": [ { "targetIndex": 0, "resID": "0", "colIndex": i1, "colName": "class_" } ] } ], "dataSplits": { "method": "holdOut", "testSize": 0.2, "stratified": True, "numRepeats": 0, "randomSeed": 42, "splitsFile": "dataSplits.csv" }, "performanceMetrics": [ { "metric": "meanSquaredError" } ] }, "expectedOutputs": { "predictionsFile": "predictions.csv" } } jsonfile=open('problemDoc.json','w') json.dump(data,jsonfile) jsonfile.close() def train_autobazaar(X_train,X_test,y_train,y_test,mtype,common_name_model,problemtype,classes,default_featurenames,transform_model,settings,model_session): # create file names model_name=common_name_model+'.pickle' folder=common_name_model # this should be the model directory hostdir=os.getcwd() # make a temporary folder for the training session try: os.mkdir(folder) os.chdir(folder) except: shutil.rmtree(folder) os.mkdir(folder) os.chdir(folder) # make the data arrays print('creating training data...') if transform_model != '': csvfilename=common_name_model+'_all.csv' shutil.copy(hostdir+'/'+model_session+'/data/'+common_name_model.split('_')[0]+'_all.csv', os.getcwd()+'/'+csvfilename) else: csvfilename=common_name_model+'_all_transformed.csv' shutil.copy(hostdir+'/'+model_session+'/data/'+common_name_model.split('_')[0]+'_all_transformed.csv', os.getcwd()+'/'+csvfilename) # create required .JSON files data=pd.read_csv(csvfilename) dataset_id, dataset_filename, i1=create_dataset_json(folder, csvfilename) problem_filename=create_problem_json(mtype, folder, i1) split_data(data) # get the current directory abz_dir=os.getcwd() # make necessary directories # now create proper tree structure ''' ─ 196_autoMpg ├── 196_autoMpg_dataset │   ├── datasetDoc.json │   └── tables │   └── learningData.csv ├── 196_autoMpg_problem │   ├── dataSplits.csv │   └── problemDoc.json ├── SCORE │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   ├── problem_TEST │   │   ├── dataSplits.csv │   │   └── problemDoc.json │   └── targets.csv ├── TEST │   ├── dataset_TEST │   │   ├── datasetDoc.json │   │   └── tables │   │   └── learningData.csv │   └── problem_TEST │   ├── dataSplits.csv │   └── problemDoc.json └── TRAIN ├── dataset_TRAIN │   ├── datasetDoc.json │   └── tables │   └── learningData.csv └── problem_TRAIN ├── dataSplits.csv └── problemDoc.json ''' dataset_folder=folder+'_dataset' problem_folder=folder+'_problem' # make datasets folder os.mkdir(dataset_folder) os.chdir(dataset_folder) os.mkdir('tables') shutil.copy(abz_dir+'/datasetDoc.json', os.getcwd()+'/datasetDoc.json') shutil.copy(abz_dir+'/'+csvfilename, os.getcwd()+'/tables/'+csvfilename) os.chdir('tables') os.rename(csvfilename, 'learningData.csv') # make problem folder os.chdir(abz_dir) os.mkdir(problem_folder) os.chdir(problem_folder) shutil.copy(abz_dir+'/problemDoc.json', os.getcwd()+'/problemDoc.json') shutil.copy(abz_dir+'/dataSplits.csv', os.getcwd()+'/dataSplits.csv') os.chdir(abz_dir) os.mkdir('TEST') os.chdir('TEST') os.mkdir('dataset_TEST') shutil.copy(abz_dir+'/'+dataset_folder+'/datasetDoc.json', os.getcwd()+'/dataset_TEST/datasetDoc.json') os.mkdir('problem_TEST') shutil.copy(abz_dir+'/'+problem_folder+'/problemDoc.json',os.getcwd()+'/problem_TEST/problemDoc.json') shutil.copy(abz_dir+'/'+problem_folder+'/dataSplits.csv', os.getcwd()+'/problem_TEST/dataSplits.csv') os.chdir('dataset_TEST') os.mkdir('tables') shutil.copy(abz_dir+'/test.csv', os.getcwd()+'/tables/test.csv') os.chdir('tables') os.rename('test.csv', 'learningData.csv') os.chdir(abz_dir) os.mkdir('TRAIN') os.chdir('TRAIN') os.mkdir('dataset_TRAIN') os.chdir('dataset_TRAIN') os.mkdir('tables') shutil.copy(abz_dir+'/datasetDoc.json', os.getcwd()+'/datasetDoc.json') shutil.copy(abz_dir+'/train.csv', os.getcwd()+'/tables/train.csv') os.chdir('tables') os.rename('train.csv','learningData.csv') os.chdir(abz_dir+'/TRAIN') os.mkdir('problem_TRAIN') shutil.copy(abz_dir+'/'+problem_folder+'/problemDoc.json',os.getcwd()+'/problem_TRAIN/problemDoc.json') shutil.copy(abz_dir+'/'+problem_folder+'/dataSplits.csv', os.getcwd()+'/problem_TRAIN/dataSplits.csv') os.chdir(abz_dir) os.mkdir('SCORE') os.chdir('SCORE') shutil.copytree(abz_dir+'/TEST/dataset_TEST',os.getcwd()+'/dataset_SCORE') shutil.copytree(abz_dir+'/TEST/problem_TEST', os.getcwd()+'/problem_SCORE') os.chdir(hostdir) # this works for really any input configuration - regression or classificatoin (as this is covered in config files) try: os.mkdir('input') except: pass # remove if file exists try: shutil.copytree(folder, os.getcwd()+'/input/'+folder) except: shutil.rmtree(os.getcwd()+'/input/'+folder) shutil.copytree(folder, os.getcwd()+'/input/'+folder) print(os.getcwd()) time.sleep(30) os.system('abz search %s -c20,30,40 -b10'%(folder)) # now go to output folder os.chdir('output') listdir=os.listdir() for i in range(len(listdir)): if listdir[i].endswith('.json'): g=json.load(open(listdir[i])) # os.remove(listdir[i]) elif listdir[i].endswith('.pkl'): picklefile=folder+'.pickle' shutil.copy(os.getcwd()+'/'+listdir[i],hostdir+'/'+model_name) files.append(model_name) model_dir=hostdir return model_name, model_dir, files
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/archived/WIP_train_mlbox-WIP.py
training/helpers/archived/WIP_train_mlbox-WIP.py
import os, json, shutil, pickle from sklearn.model_selection import train_test_split from sklearn.metrics import accuracy_score, mean_squared_log_error import pandas as pd print('installing library') os.system('pip3 install mlbox==0.8.4') from mlbox.preprocessing import * from mlbox.optimisation import * from mlbox.prediction import * ''' From the documentation: https://mlbox.readthedocs.io/en/latest/ ''' # install mlblocks def train_mlbox(alldata, labels, mtype, jsonfile, problemtype, default_features, settings): # name model modelname=jsonfile[0:-5]+'_mlbox_'+str(default_features).replace("'",'').replace('"','') # training and testing sets X_train, X_test, y_train, y_test = train_test_split(alldata, labels, train_size=0.750, test_size=0.250) df = {"train" : pd.DataFrame(X_train), "target" : pd.DataFrame(y_train), "test" : pd.DataFrame(X_test)} print(df) if mtype=='c': # rename files with classification modelname=modelname+'_classification' model_name=modelname+'.pickle' jsonfilename=modelname+'.json' # from sklearn.datasets import load_boston # dataset = load_boston() # df = {"train" : pd.DataFrame(dataset.data), "target" : pd.Series(dataset.target)} # print(df['train'][0]) # print(type(df['train'][0])) # data = Drift_thresholder().fit_transform(df) #deleting non-stable variables space = { 'ne__numerical_strategy' : {"space" : [0, 'mean']}, 'ce__strategy' : {"space" : ["label_encoding", "random_projection", "entity_embedding"]}, 'fs__strategy' : {"space" : ["variance", "rf_feature_importance"]}, 'fs__threshold': {"search" : "choice", "space" : [0.1, 0.2, 0.3]}, 'est__strategy' : {"space" : ["LightGBM"]}, 'est__max_depth' : {"search" : "choice", "space" : [5,6]}, 'est__subsample' : {"search" : "uniform", "space" : [0.6,0.9]} } best = Optimiser().optimise(space, df, max_evals = 5) mse_ =Optimiser().evaluate(best, df) pipeline = Predictor().fit_predict(best, df) print(best) print(mse_) # saving model print('saving model') modelfile=open(model_name,'wb') pickle.dump(pipeline, modelfile) modelfile.close() # SAVE JSON FILE print('saving .JSON file (%s)'%(jsonfilename)) jsonfile=open(jsonfilename,'w') data={'sample type': problemtype, 'feature_set':default_features, 'model name':jsonfilename[0:-5]+'.pickle', 'accuracy':accuracy, 'model type':'mlblocks_regression', 'settings': settings, } json.dump(data,jsonfile) jsonfile.close() if mtype=='r': # rename files with regression modelname=modelname+'_regression' model_name=modelname+'.pickle' jsonfilename=modelname+'.json' params = {"ne__numerical_strategy" : 0, "ce__strategy" : "label_encoding", "fs__threshold" : 0.1, "stck__base_estimators" : [Regressor(strategy="RandomForest"), Regressor(strategy="ExtraTrees")], "est__strategy" : "Linear"} best = Optimiser().optimise(params, df, max_evals = 5) mse_error =Optimiser().evaluate(best, df) # saving model print('saving model') modelfile=open(model_name,'wb') pickle.dump(pipeline, modelfile) modelfile.close() # save JSON print('saving .JSON file (%s)'%(jsonfilename)) jsonfile=open(jsonfilename,'w') data={'sample type': problemtype, 'feature_set':default_features, 'model name':jsonfilename[0:-5]+'.pickle', 'mse_error':mse_error, 'model type':'mlblocks_regression', 'settings': settings, } json.dump(data,jsonfile) jsonfile.close() cur_dir2=os.getcwd() try: os.chdir(problemtype+'_models') except: os.mkdir(problemtype+'_models') os.chdir(problemtype+'_models') # now move all the files over to proper model directory shutil.copy(cur_dir2+'/'+model_name, os.getcwd()+'/'+model_name) shutil.copy(cur_dir2+'/'+jsonfilename, os.getcwd()+'/'+jsonfilename) os.remove(cur_dir2+'/'+model_name) os.remove(cur_dir2+'/'+jsonfilename) # get model directory model_dir=os.getcwd() return model_name, model_dir
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/archived/train_pLDA.py
training/helpers/archived/train_pLDA.py
''' PLDA implementation from https://github.com/RaviSoji/plda/blob/master/mnist_demo/mnist_demo.ipynb ''' import os, sys, pickle import helpers.plda.plda as plda import numpy as np import matplotlib.pyplot as plt from sklearn.model_selection import train_test_split from sklearn.model_selection import cross_val_score def train_pLDA(alldata, labels): # get train and test data training_data, testing_data, training_labels, testing_labels = train_test_split(alldata, labels, train_size=0.750, test_size=0.250) training_data = training_data.reshape(training_data.shape) testing_data = testing_data.reshape(testing_data.shape) # optimize number of principal components in terms of accuracy acclist=list() compnum=list() for i in range(2, len(training_data[0]),1): #try: classifier = plda.Classifier() numcomponents=i classifier.fit_model(training_data, training_labels, n_principal_components=numcomponents) predictions, log_p_predictions = classifier.predict(testing_data) accuracy=(testing_labels == predictions).mean() print(accuracy) if accuracy > 1: pass else: acclist.append(accuracy) print(accuracy) print(i) #except: # if dimension too high, break it #print('error') maxacc=max(acclist) numcomponents=acclist.index(maxacc)+1 # now retrain with proper parameters classifier = plda.Classifier() classifier.fit_model(training_data, training_labels, n_principal_components=numcomponents) predictions, log_p_predictions = classifier.predict(testing_data) accuracy=(testing_labels == predictions).mean() print('max acc %s with %s components'%(maxacc, numcomponents)) Psi = classifier.model.Psi A = classifier.model.A inv_A = classifier.model.inv_A m = classifier.model.m # Indices of the subspace used for classification. relevant_U_dims = classifier.model.relevant_U_dims # # Prior Gaussian Parameters # classifier.model.prior_params.keys() # # Posterior Gaussian Parameters # classifier.model.posterior_params.keys() # classifier.model.posterior_params[0].keys() # # Posterior Predictive Gaussian Parameters # classifier.model.posterior_predictive_params.keys() # classifier.model.posterior_predictive_params[0].keys() ''' Transforming Data to PLDA Space There are 4 "spaces" that result from the transformations the model performs: Data space ('D'), Preprocessed data space ('X'), Latent space ('U'), and The "effective" subspace of the latent space ('U_model'), which is essentially the set of dimensions the model actually uses for prediction. ''' # U_model = classifier.model.transform(training_data, from_space='D', to_space='U_model') # print(training_data.shape) # print(U_model.shape) # D = classifier.model.transform(U_model, from_space='U_model', to_space='D') # print(U_model.shape) # print(D.shape) # create dump of classifier print('saving classifier to disk') f=open('plda_classifier.pickle','wb') pickle.dump(classifier,f) f.close()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/archived/train_autosklearn.py
training/helpers/archived/train_autosklearn.py
import numpy as np import matplotlib.pyplot as plt import pandas as pd from sklearn.model_selection import GridSearchCV, train_test_split import autosklearn.classification as asklc import sklearn.metrics import os, shutil def train_autosklearn(alldata, labels, mtype, jsonfile, problemtype, default_features): foldername=jsonfile[0:-5]+'_autosklearn_%s'%(default_features) X_train, X_test, y_train, y_test = train_test_split(alldata, labels, train_size=0.750, test_size=0.250, random_state=42, shuffle=True) feature_types = (['numerical'] * len(X_train[0])) automl = asklc.AutoSklearnClassifier( time_left_for_this_task=60, per_run_time_limit=300, ml_memory_limit=10240, tmp_folder=os.getcwd()+'/'+foldername+'_tmp', output_folder=os.getcwd()+'/'+foldername, delete_tmp_folder_after_terminate=False, delete_output_folder_after_terminate=False) automl.fit(X_train, y_train, dataset_name=jsonfile[0:-5], feat_type=feature_types) y_predictions = automl.predict(X_test) acc= sklearn.metrics.accuracy_score(y_true=y_test, y_pred=y_predictions) print("Accuracy:", acc) print('saving classifier to disk') f=open(modelname+'.pickle','wb') pickle.dump(automl,f) f.close() data={'sample type': problemtype, 'training script': 'autosklearn', 'feature_set':default_features, 'model name':modelname+'.pickle', 'accuracy':acc, 'model type':'sc_'+classifiername, } g2=open(modelname+'.json','w') json.dump(data,g2) g2.close() cur_dir2=os.getcwd() try: os.chdir(problemtype+'_models') except: os.mkdir(problemtype+'_models') os.chdir(problemtype+'_models') # now move all the files over to proper model directory shutil.move(cur_dir2+'/'+modelname+'.json', os.getcwd()+'/'+modelname+'.json') shutil.move(cur_dir2+'/'+modelname+'.pickle', os.getcwd()+'/'+modelname+'.pickle')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/archived/train_autokeras.py
training/helpers/archived/train_autokeras.py
''' @Train_autokeras script. Take in a dataset, convert it to pytorch dataloader format, ingest it in autokeras, output model in './models directory' This will make it easier to deploy automated machine learning models into the future. Note that grid search can be expensive + take up to 24 hours on most GPUs / CPUs to optimize a model. ''' from autokeras import MlpModule, CnnModule from autokeras.backend.torch.loss_function import classification_loss from autokeras.backend.torch.loss_function import regression_loss from autokeras.nn.metric import Accuracy from autokeras.utils import pickle_from_file from sklearn.model_selection import train_test_split # pre processing from sklearn.model_selection import cross_val_score from sklearn.model_selection import KFold from sklearn.preprocessing import LabelEncoder from sklearn.pipeline import Pipeline from pandas import Series from sklearn.model_selection import train_test_split from keras.preprocessing.sequence import pad_sequences from sklearn.preprocessing import MinMaxScaler from sklearn.preprocessing import minmax_scale from sklearn.preprocessing import MaxAbsScaler from sklearn.preprocessing import StandardScaler from sklearn.preprocessing import RobustScaler from sklearn.preprocessing import Normalizer from sklearn.preprocessing.data import QuantileTransformer import numpy as np from keras.utils import to_categorical import keras.models from keras import layers from keras.models import Sequential,model_from_json from keras.layers import Dense, Dropout import torch, time, shutil, os import torch.utils.data as utils # skip the CNN for neural architecture search because it doesn't work unless an image type really. def train_autokeras(classes, alldata, labels, mtype, jsonfile, problemtype, default_features): ## this is a CNN architecture modelname=jsonfile[0:-5]+'_autokeras_%s'%(default_features) TEST_FOLDER = modelname x_train, x_test, y_train, y_test = train_test_split(alldata, labels, train_size=0.750, test_size=0.250) # we have to do some odd re-shapes to get the data loader to work for the autokeras module (keep this in mind when loading new data in) x_train=x_train.reshape(x_train.shape+(1,)) y_train=y_train.reshape(y_train.shape+(1,)+(1,)) x_test=x_test.reshape(x_test.shape+(1,)) y_test=y_test.reshape(y_test.shape+(1,)+(1,)) print(x_train.shape) print(y_train.shape) tensor_x = torch.stack([torch.Tensor(i) for i in x_train]) # transform to torch tensors tensor_y = torch.stack([torch.Tensor(i) for i in y_train]) my_dataset = utils.TensorDataset(tensor_x, tensor_y) # create your datset training_data = utils.DataLoader(my_dataset) # create your dataloader tensor_x = torch.stack([torch.Tensor(i) for i in x_test]) # transform to torch tensors tensor_y = torch.stack([torch.Tensor(i) for i in y_test]) my_dataset = utils.TensorDataset(tensor_x,tensor_y) # create your datset test_data = utils.DataLoader(my_dataset) # create your dataloader print(test_data) input_shape=x_train[0].shape n_output_node=1 # cnnModule = CnnModule(loss=classification_loss, metric=Accuracy, searcher_args={}, path=TEST_FOLDER, verbose=False) if mtype == 'c': # metric = Accuracy is for classification # loss = classiciation_loss for classification mlpModule = MlpModule(loss=classification_loss, metric=Accuracy, searcher_args={}, path=TEST_FOLDER, verbose=True) elif mtype == 'r': # metric = MSE for regression # loss = regression_loss for regression mlpModule = MlpModule(loss=regression_loss, metric=MSE, searcher_args={}, path=TEST_FOLDER, verbose=True) timelimit=60 print('training MLP model for %s hours'%(timelimit/(60*60))) mlpModule.fit(n_output_node, input_shape, training_data, test_data, time_limit=timelimit) mlpModule.final_fit(training_data, test_data, trainer_args=None, retrain=False) # # serialize model to JSON # mlpModule.export_autokeras_model(modelname+'.pickle') # print("\n Saved %s.pickle model to disk"%(modelname)) # # test opening model and making predictions # model=pickle_from_file(modelname+'.pickle') # results=model.evaluate(x_test, y_test) # print(results) cur_dir2=os.getcwd() try: os.chdir(problemtype+'_models') except: os.mkdir(problemtype+'_models') os.chdir(problemtype+'_models') # now move all the files over to proper model directory shutil.copytree(cur_dir2+'/'+TEST_FOLDER, os.getcwd() + '/'+TEST_FOLDER) shutil.rmtree(cur_dir2+'/'+TEST_FOLDER)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/setup.py
training/helpers/hyperband/setup.py
from __future__ import print_function import sys from setuptools import setup, find_packages with open('requirements.txt') as f: INSTALL_REQUIRES = [l.strip() for l in f.readlines() if l] try: import numpy except ImportError: print('numpy is required during installation') sys.exit(1) try: import scipy except ImportError: print('scipy is required during installation') sys.exit(1) setup(name='scikit-hyperband', version='0.0.1', description='A scikit-learn compatible implementation of hyperband for model selection', author='Thomas Huijskens', packages=find_packages(), install_requires=INSTALL_REQUIRES, author_email='thomas_huijskens@hotmail.com', )
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/hyperband/search.py
training/helpers/hyperband/hyperband/search.py
""" ========= Hyperband ========= This module contains a scikit-learn compatible implementation of the hyperband algorithm[^1]. Compared to the civismlext implementation, this supports multimetric scoring, and the option to turn the last round of hyperband (the randomized search round) off. References ---------- .. [1] Li, L., Jamieson, K., DeSalvo, G., Rostamizadeh, A. and Talwalkar, A., 2017. Hyperband: A novel bandit-based approach to hyperparameter optimization. The Journal of Machine Learning Research, 18(1), pp.6765-6816. """ import copy import numpy as np from scipy.stats import rankdata from sklearn.utils import check_random_state from sklearn.model_selection._search import BaseSearchCV, ParameterSampler __all__ = ['HyperbandSearchCV'] class HyperbandSearchCV(BaseSearchCV): """Hyperband search on hyper parameters. HyperbandSearchCV implements a ``fit`` and a ``score`` method. It also implements ``predict``, ``predict_proba``, ``decision_function``, ``transform`` and ``inverse_transform`` if they are implemented in the estimator used. The parameters of the estimator used to apply these methods are optimized by cross-validated search over parameter settings using the hyperband algorithm [1]_ . If all parameters are presented as a list, sampling without replacement is performed. If at least one parameter is given as a distribution, sampling with replacement is used. It is highly recommended to use continuous distributions for continuous parameters. Read more in the scikit-learn `User Guide <http://scikit-learn.org/stable/modules/grid_search.html#randomized-parameter-search>`_. Parameters ---------- estimator : estimator object. A object of that type is instantiated for each grid point. This is assumed to implement the scikit-learn estimator interface. Either estimator needs to provide a ``score`` function, or ``scoring`` must be passed. param_distributions : dict Dictionary with parameters names (string) as keys and distributions or lists of parameters to try. Distributions must provide a ``rvs`` method for sampling (such as those from scipy.stats.distributions). If a list is given, it is sampled uniformly. resource_param : str, default='n_estimators' The name of the cost parameter for the estimator ``estimator`` to be fitted. Typically, this is the number of decision trees ``n_estimators`` in an ensemble or the number of iterations for estimators trained with stochastic gradient descent. eta : float, default=3 The inverse of the proportion of configurations that are discarded in each round of hyperband. min_iter : int, default=1 The minimum amount of resource that should be allocated to the cost parameter ``resource_param`` for a single configuration of the hyperparameters. max_iter : int, default=81 The maximum amount of resource that can be allocated to the cost parameter ``resource_param`` for a single configuration of the hyperparameters. skip_last : int, default=0 The number of last rounds to skip. For example, this can be used to skip the last round of hyperband, which is standard randomized search. It can also be used to inspect intermediate results, although warm-starting HyperbandSearchCV is not supported. scoring : string, callable, list/tuple, dict or None, default: None A single string (see :ref:`scoring_parameter`) or a callable (see :ref:`scoring`) to evaluate the predictions on the test set. For evaluating multiple metrics, either give a list of (unique) strings or a dict with names as keys and callables as values. NOTE that when using custom scorers, each scorer should return a single value. Metric functions returning a list/array of values can be wrapped into multiple scorers that return one value each. See :ref:`multimetric_grid_search` for an example. If None, the estimator's default scorer (if available) is used. n_jobs : int, default=1 Number of jobs to run in parallel. pre_dispatch : int, or string, optional Controls the number of jobs that get dispatched during parallel execution. Reducing this number can be useful to avoid an explosion of memory consumption when more jobs get dispatched than CPUs can process. This parameter can be: - None, in which case all the jobs are immediately created and spawned. Use this for lightweight and fast-running jobs, to avoid delays due to on-demand spawning of the jobs - An int, giving the exact number of total jobs that are spawned - A string, giving an expression as a function of n_jobs, as in '2*n_jobs' iid : boolean, default=True If True, the data is assumed to be identically distributed across the folds, and the loss minimized is the total loss per sample, and not the mean loss across the folds. cv : int, cross-validation generator or an iterable, optional Determines the cross-validation splitting strategy. Possible inputs for cv are: - None, to use the default 3-fold cross validation, - integer, to specify the number of folds in a `(Stratified)KFold`, - An object to be used as a cross-validation generator. - An iterable yielding train, test splits. For integer/None inputs, if the estimator is a classifier and ``y`` is either binary or multiclass, :class:`sklearn.model_selection.StratifiedKFold` is used. In all other cases, :class:`sklearn.model_selection.KFold` is used. Refer `User Guide <http://scikit-learn.org/stable/modules/cross_validation.html>`_ for the various cross-validation strategies that can be used here. refit : boolean, or string default=True Refit an estimator using the best found parameters on the whole dataset. For multiple metric evaluation, this needs to be a string denoting the scorer that would be used to find the best parameters for refitting the estimator at the end. The refitted estimator is made available at the ``best_estimator_`` attribute and permits using ``predict`` directly on this ``HyperbandSearchCV`` instance. Also for multiple metric evaluation, the attributes ``best_index_``, ``best_score_`` and ``best_parameters_`` will only be available if ``refit`` is set and all of them will be determined w.r.t this specific scorer. See ``scoring`` parameter to know more about multiple metric evaluation. verbose : integer Controls the verbosity: the higher, the more messages. random_state : int, RandomState instance or None, optional, default=None Pseudo random number generator state used for random uniform sampling from lists of possible values instead of scipy.stats distributions. If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by `np.random`. error_score : 'raise' (default) or numeric Value to assign to the score if an error occurs in estimator fitting. If set to 'raise', the error is raised. If a numeric value is given, FitFailedWarning is raised. This parameter does not affect the refit step, which will always raise the error. return_train_score : boolean, optional, default=False If ``False``, the ``cv_results_`` attribute will not include training scores. Attributes ---------- cv_results_ : dict of numpy (masked) ndarrays A dict with keys as column headers and values as columns, that can be imported into a pandas ``DataFrame``. For instance the below given table +------------+-----------+------------+-----------------+---+---------+ |param_kernel|param_gamma|param_degree|split0_test_score|...|rank_t...| +============+===========+============+=================+===+=========+ | 'poly' | -- | 2 | 0.8 |...| 2 | +------------+-----------+------------+-----------------+---+---------+ | 'poly' | -- | 3 | 0.7 |...| 4 | +------------+-----------+------------+-----------------+---+---------+ | 'rbf' | 0.1 | -- | 0.8 |...| 3 | +------------+-----------+------------+-----------------+---+---------+ | 'rbf' | 0.2 | -- | 0.9 |...| 1 | +------------+-----------+------------+-----------------+---+---------+ will be represented by a ``cv_results_`` dict of:: { 'param_kernel': masked_array(data = ['poly', 'poly', 'rbf', 'rbf'], mask = [False False False False]...) 'param_gamma': masked_array(data = [-- -- 0.1 0.2], mask = [ True True False False]...), 'param_degree': masked_array(data = [2.0 3.0 -- --], mask = [False False True True]...), 'split0_test_score' : [0.8, 0.7, 0.8, 0.9], 'split1_test_score' : [0.82, 0.5, 0.7, 0.78], 'mean_test_score' : [0.81, 0.60, 0.75, 0.82], 'std_test_score' : [0.02, 0.01, 0.03, 0.03], 'rank_test_score' : [2, 4, 3, 1], 'split0_train_score' : [0.8, 0.9, 0.7], 'split1_train_score' : [0.82, 0.5, 0.7], 'mean_train_score' : [0.81, 0.7, 0.7], 'std_train_score' : [0.03, 0.03, 0.04], 'mean_fit_time' : [0.73, 0.63, 0.43, 0.49], 'std_fit_time' : [0.01, 0.02, 0.01, 0.01], 'mean_score_time' : [0.007, 0.06, 0.04, 0.04], 'std_score_time' : [0.001, 0.002, 0.003, 0.005], 'params' : [{'kernel': 'poly', 'degree': 2}, ...], } NOTE The key ``'params'`` is used to store a list of parameter settings dicts for all the parameter candidates. The ``mean_fit_time``, ``std_fit_time``, ``mean_score_time`` and ``std_score_time`` are all in seconds. For multi-metric evaluation, the scores for all the scorers are available in the ``cv_results_`` dict at the keys ending with that scorer's name (``'_<scorer_name>'``) instead of ``'_score'`` shown above. ('split0_test_precision', 'mean_train_precision' etc.) best_estimator_ : estimator or dict Estimator that was chosen by the search, i.e. estimator which gave highest score (or smallest loss if specified) on the left out data. Not available if ``refit=False``. For multi-metric evaluation, this attribute is present only if ``refit`` is specified. See ``refit`` parameter for more information on allowed values. best_score_ : float Mean cross-validated score of the best_estimator. For multi-metric evaluation, this is not available if ``refit`` is ``False``. See ``refit`` parameter for more information. best_params_ : dict Parameter setting that gave the best results on the hold out data. For multi-metric evaluation, this is not available if ``refit`` is ``False``. See ``refit`` parameter for more information. best_index_ : int The index (of the ``cv_results_`` arrays) which corresponds to the best candidate parameter setting. The dict at ``search.cv_results_['params'][search.best_index_]`` gives the parameter setting for the best model, that gives the highest mean score (``search.best_score_``). For multi-metric evaluation, this is not available if ``refit`` is ``False``. See ``refit`` parameter for more information. scorer_ : function or a dict Scorer function used on the held out data to choose the best parameters for the model. For multi-metric evaluation, this attribute holds the validated ``scoring`` dict which maps the scorer key to the scorer callable. n_splits_ : int The number of cross-validation splits (folds/iterations). References ---------- .. [1] Li, L., Jamieson, K., DeSalvo, G., Rostamizadeh, A. and Talwalkar, A., 2017. Hyperband: A novel bandit-based approach to hyperparameter optimization. The Journal of Machine Learning Research, 18(1), pp.6765-6816. Notes ----- The parameters selected are those that maximize the score of the held-out data, according to the scoring parameter. If `n_jobs` was set to a value higher than one, the data is copied for each parameter setting (and not `n_jobs` times). This is done for efficiency reasons if individual jobs take very little time, but may raise errors if the dataset is large and not enough memory is available. A workaround in this case is to set `pre_dispatch`. Then, the memory is copied only `pre_dispatch` many times. A reasonable value for `pre_dispatch` is `2 * n_jobs`. See Also -------- :class:`sklearn.model_selection.GridSearchCV`: Does exhaustive search over a grid of parameters. :class:`sklearn.model_selection.ParameterSampler`: A generator over parameter settings, constructed from param_distributions. """ def __init__(self, estimator, param_distributions, resource_param='n_estimators', eta=3, min_iter=1, max_iter=81, skip_last=0, scoring=None, n_jobs=1, iid=True, refit=True, cv=None, verbose=0, pre_dispatch='2*n_jobs', random_state=None, error_score='raise', return_train_score=False): self.param_distributions = param_distributions self.resource_param = resource_param self.eta = eta self.min_iter = min_iter self.max_iter = max_iter self.skip_last = skip_last self.random_state = random_state super(HyperbandSearchCV, self).__init__( estimator=estimator, scoring=scoring, n_jobs=n_jobs, iid=iid, refit=refit, cv=cv, verbose=verbose, pre_dispatch=pre_dispatch, error_score=error_score, return_train_score=return_train_score) def _run_search(self, evaluate_candidates): self._validate_input() s_max = int(np.floor(np.log(self.max_iter / self.min_iter) / np.log(self.eta))) B = (s_max + 1) * self.max_iter refit_metric = self.refit if self.multimetric_ else 'score' random_state = check_random_state(self.random_state) if self.skip_last > s_max: raise ValueError('skip_last is higher than the total number of rounds') for round_index, s in enumerate(reversed(range(s_max + 1))): n = int(np.ceil(int(B / self.max_iter / (s + 1)) * np.power(self.eta, s))) # initial number of iterations per config r = self.max_iter / np.power(self.eta, s) configurations = list(ParameterSampler(param_distributions=self.param_distributions, n_iter=n, random_state=random_state)) if self.verbose > 0: print('Starting bracket {0} (out of {1}) of hyperband' .format(round_index + 1, s_max + 1)) for i in range((s + 1) - self.skip_last): n_configs = np.floor(n / np.power(self.eta, i)) # n_i n_iterations = int(r * np.power(self.eta, i)) # r_i n_to_keep = int(np.floor(n_configs / self.eta)) if self.verbose > 0: msg = ('Starting successive halving iteration {0} out of' ' {1}. Fitting {2} configurations, with' ' resource_param {3} set to {4}') if n_to_keep > 0: msg += ', and keeping the best {5} configurations.' msg = msg.format(i + 1, s + 1, len(configurations), self.resource_param, n_iterations, n_to_keep) print(msg) # Set the cost parameter for every configuration parameters = copy.deepcopy(configurations) for configuration in parameters: configuration[self.resource_param] = n_iterations results = evaluate_candidates(parameters) if n_to_keep > 0: top_configurations = [x for _, x in sorted(zip(results['rank_test_%s' % refit_metric], results['params']), key=lambda x: x[0])] configurations = top_configurations[:n_to_keep] if self.skip_last > 0: print('Skipping the last {0} successive halving iterations' .format(self.skip_last)) def fit(self, X, y=None, groups=None, **fit_params): """Run fit with all sets of parameters. Parameters ---------- X : array-like, shape = [n_samples, n_features] Training vector, where n_samples is the number of samples and n_features is the number of features. y : array-like, shape = [n_samples] or [n_samples, n_output], optional Target relative to X for classification or regression; None for unsupervised learning. groups : array-like, with shape (n_samples,), optional Group labels for the samples used while splitting the dataset into train/test set. **fit_params : dict of string -> object Parameters passed to the ``fit`` method of the estimator """ super().fit(X, y, groups, **fit_params) s_max = int(np.floor(np.log(self.max_iter / self.min_iter) / np.log(self.eta))) B = (s_max + 1) * self.max_iter brackets = [] for round_index, s in enumerate(reversed(range(s_max + 1))): n = int(np.ceil(int(B / self.max_iter / (s + 1)) * np.power(self.eta, s))) n_configs = int(sum([np.floor(n / np.power(self.eta, i)) for i in range((s + 1) - self.skip_last)])) bracket = (round_index + 1) * np.ones(n_configs) brackets.append(bracket) self.cv_results_['hyperband_bracket'] = np.hstack(brackets) return self def _validate_input(self): if not isinstance(self.min_iter, int) or self.min_iter <= 0: raise ValueError('min_iter should be a positive integer, got %s' % self.min_iter) if not isinstance(self.max_iter, int) or self.max_iter <= 0: raise ValueError('max_iter should be a positive integer, got %s' % self.max_iter) if self.max_iter < self.min_iter: raise ValueError('max_iter should be bigger than min_iter, got' 'max_iter=%d and min_iter=%d' % (self.max_iter, self.min_iter)) if not isinstance(self.skip_last, int) or self.skip_last < 0: raise ValueError('skip_last should be an integer, got %s' % self.skip_last) if not isinstance(self.eta, int) or not self.eta > 1: raise ValueError('eta should be a positive integer, got %s' % self.eta) if self.resource_param not in self.estimator.get_params().keys(): raise ValueError('resource_param is set to %s, but base_estimator %s ' 'does not have a parameter with that name' % (self.resource_param, self.estimator.__class__.__name__))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/hyperband/__init__.py
training/helpers/hyperband/hyperband/__init__.py
""" """ from .search import HyperbandSearchCV __all__ = ['HyperbandSearchCV']
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/hyperband/tests/__init__.py
training/helpers/hyperband/hyperband/tests/__init__.py
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/hyperband/tests/test_validation.py
training/helpers/hyperband/hyperband/tests/test_validation.py
from nose.tools import raises from hyperband import HyperbandSearchCV from sklearn.ensemble import RandomForestClassifier from scipy.stats import randint as sp_randint def setup(): model = RandomForestClassifier() param_dist = {"max_depth": [3, None], "max_features": sp_randint(1, 11), "min_samples_split": sp_randint(2, 11), "min_samples_leaf": sp_randint(1, 11), "bootstrap": [True, False], "criterion": ["gini", "entropy"]} return model, param_dist @raises(ValueError) def test_check_min_iter(): model, param_dist = setup() HyperbandSearchCV(model, param_dist, min_iter=-1)._validate_input() @raises(ValueError) def test_check_max_iter(): model, param_dist = setup() HyperbandSearchCV(model, param_dist, max_iter=-1)._validate_input() @raises(ValueError) def test_check_min_iter_smaller_max_iter(): model, param_dist = setup() HyperbandSearchCV(model, param_dist, min_iter=30, max_iter=15)._validate_input() @raises(ValueError) def test_check_skip_last(): model, param_dist = setup() HyperbandSearchCV(model, param_dist, skip_last=-1)._validate_input() @raises(ValueError) def test_check_eta(): model, param_dist = setup() HyperbandSearchCV(model, param_dist, eta=0)._validate_input() @raises(ValueError) def test_check_resource_param(): model, param_dist = setup() HyperbandSearchCV(model, param_dist, resource_param='wrong_name')._validate_input()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/hyperband/tests/test_hyperband.py
training/helpers/hyperband/hyperband/tests/test_hyperband.py
from nose.tools import raises from hyperband import HyperbandSearchCV from sklearn.ensemble import RandomForestClassifier from scipy.stats import randint as sp_randint from sklearn.datasets import load_digits from sklearn.utils import check_random_state def setup(): model = RandomForestClassifier() rng = check_random_state(42) param_dist = {'max_depth': [3, None], 'max_features': sp_randint(1, 11), 'min_samples_split': sp_randint(2, 11), 'bootstrap': [True, False], 'criterion': ['gini', 'entropy']} digits = load_digits() X, y = digits.data, digits.target return model, param_dist, X, y, rng def test_multimetric_hyperband(): model, param_dist, X, y, rng = setup() # multimetric scoring is only supported for 1-D classification first_label = (y == 1) y[first_label] = 1 y[~first_label] = 0 multimetric = [ 'roc_auc', 'accuracy' ] search = HyperbandSearchCV(model, param_dist, refit='roc_auc', scoring=multimetric, random_state=rng) search.fit(X, y) assert('mean_test_roc_auc' in search.cv_results_.keys()) assert('mean_test_accuracy' in search.cv_results_.keys()) assert (len(search.cv_results_['hyperband_bracket']) == 187) def test_min_resource_param(): model, param_dist, X, y, rng = setup() search = HyperbandSearchCV(model, param_dist, min_iter=3, random_state=rng, verbose=1) search.fit(X, y) assert(search.cv_results_['param_n_estimators'].data.min() == 3) @raises(ValueError) def test_skip_last_raise(): model, param_dist, X, y, rng = setup() search = HyperbandSearchCV(model, param_dist, skip_last=10, random_state=rng) search.fit(X, y) def test_skip_last(): model, param_dist, X, y, rng = setup() search = HyperbandSearchCV(model, param_dist, skip_last=1, random_state=rng) search.fit(X, y) # 177 Because in every round the last search is dropped # 187 - (1 + 1 + 1 + 2 + 5) assert (len(search.cv_results_['hyperband_bracket']) == 177)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/examples/random_forest_example.py
training/helpers/hyperband/examples/random_forest_example.py
""" ================================================================== Tuning the hyperparameters of a random forest model with hyperband ================================================================== """ from hyperband import HyperbandSearchCV from scipy.stats import randint as sp_randint from sklearn.datasets import load_digits from sklearn.ensemble import RandomForestClassifier from sklearn.preprocessing import LabelBinarizer if __name__ == '__main__': model = RandomForestClassifier() param_dist = { 'max_depth': [3, None], 'max_features': sp_randint(1, 11), 'min_samples_split': sp_randint(2, 11), 'min_samples_leaf': sp_randint(1, 11), 'bootstrap': [True, False], 'criterion': ['gini', 'entropy'] } digits = load_digits() X, y = digits.data, digits.target y = LabelBinarizer().fit_transform(y) search = HyperbandSearchCV(model, param_dist, resource_param='n_estimators', scoring='roc_auc', n_jobs=1, verbose=1) search.fit(X, y) print(search.best_params_) print(search.best_score_)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/hyperband/doc/conf.py
training/helpers/hyperband/doc/conf.py
# -*- coding: utf-8 -*- # # project-template documentation build configuration file, created by # sphinx-quickstart on Mon Jan 18 14:44:12 2016. # # This file is execfile()d with the current directory set to its # containing dir. # # Note that not all possible configuration values are present in this # autogenerated file. # # All configuration values have a default; values that are commented out # serve to show the default. import sys import os import sphinx_rtd_theme # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. sys.path.insert(0, os.path.abspath('..')) # -- General configuration --------------------------------------------------- # Try to override the matplotlib configuration as early as possible try: import gen_rst except: pass # -- General configuration ------------------------------------------------ # If your documentation needs a minimal Sphinx version, state it here. #needs_sphinx = '1.0' # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom # ones. extensions = [ 'sphinx.ext.autodoc', 'sphinx.ext.autosummary', 'sphinx.ext.doctest', 'sphinx.ext.intersphinx', 'sphinx.ext.todo', 'numpydoc', 'sphinx.ext.ifconfig', 'sphinx.ext.viewcode', 'sphinx_gallery.gen_gallery' ] numpydoc_show_class_members = False # pngmath / imgmath compatibility layer for different sphinx versions import sphinx from distutils.version import LooseVersion if LooseVersion(sphinx.__version__) < LooseVersion('1.4'): extensions.append('sphinx.ext.pngmath') else: extensions.append('sphinx.ext.imgmath') sphinx_gallery_conf = { # path to your examples scripts 'examples_dirs' : '../examples', # path where to save gallery generated examples 'gallery_dirs' : 'auto_examples', 'backreferences_dir': os.path.join('generated'), } # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix of source filenames. source_suffix = '.rst' # The encoding of source files. #source_encoding = 'utf-8-sig' # Generate the plots for the gallery plot_gallery = True # The master toctree document. master_doc = 'index' # General information about the project. project = u'scikit-hyperband' copyright = u'2018, Thomas Huijskens' # The version info for the project you're documenting, acts as replacement for # |version| and |release|, also used in various other places throughout the # built documents. # # The short X.Y version. version = '0.1' # The full version, including alpha/beta/rc tags. release = '0.1.0' # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. #language = None # There are two options for replacing |today|: either, you set today to some # non-false value, then it is used: #today = '' # Else, today_fmt is used as the format for a strftime call. #today_fmt = '%B %d, %Y' # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. exclude_patterns = ['_build'] # The reST default role (used for this markup: `text`) to use for all # documents. #default_role = None # If true, '()' will be appended to :func: etc. cross-reference text. #add_function_parentheses = True # If true, the current module name will be prepended to all description # unit titles (such as .. function::). #add_module_names = True # If true, sectionauthor and moduleauthor directives will be shown in the # output. They are ignored by default. #show_authors = False # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' # A list of ignored prefixes for module index sorting. #modindex_common_prefix = [] # If true, keep warnings as "system message" paragraphs in the built documents. #keep_warnings = False # -- Options for HTML output ---------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. html_theme = 'sphinx_rtd_theme' # Theme options are theme-specific and customize the look and feel of a theme # further. For a list of options available for each theme, see the # documentation. #html_theme_options = {} # Add any paths that contain custom themes here, relative to this directory. html_theme_path = [sphinx_rtd_theme.get_html_theme_path()] # The name for this set of Sphinx documents. If None, it defaults to # "<project> v<release> documentation". #html_title = None # A shorter title for the navigation bar. Default is the same as html_title. #html_short_title = None # The name of an image file (relative to this directory) to place at the top # of the sidebar. #html_logo = None # The name of an image file (within the static path) to use as favicon of the # docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32 # pixels large. #html_favicon = None # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = ['_static'] # Add any extra paths that contain custom files (such as robots.txt or # .htaccess) here, relative to this directory. These files are copied # directly to the root of the documentation. #html_extra_path = [] # If not '', a 'Last updated on:' timestamp is inserted at every page bottom, # using the given strftime format. #html_last_updated_fmt = '%b %d, %Y' # If true, SmartyPants will be used to convert quotes and dashes to # typographically correct entities. #html_use_smartypants = True # Custom sidebar templates, maps document names to template names. #html_sidebars = {} # Additional templates that should be rendered to pages, maps page names to # template names. #html_additional_pages = {} # If false, no module index is generated. #html_domain_indices = True # If false, no index is generated. #html_use_index = True # If true, the index is split into individual pages for each letter. #html_split_index = False # If true, links to the reST sources are added to the pages. #html_show_sourcelink = True # If true, "Created using Sphinx" is shown in the HTML footer. Default is True. #html_show_sphinx = True # If true, "(C) Copyright ..." is shown in the HTML footer. Default is True. #html_show_copyright = True # If true, an OpenSearch description file will be output, and all pages will # contain a <link> tag referring to it. The value of this option must be the # base URL from which the finished HTML is served. #html_use_opensearch = '' # This is the file name suffix for HTML files (e.g. ".xhtml"). #html_file_suffix = None # Output file base name for HTML help builder. htmlhelp_basename = 'project-templatedoc' # -- Options for LaTeX output --------------------------------------------- latex_elements = { # The paper size ('letterpaper' or 'a4paper'). #'papersize': 'letterpaper', # The font size ('10pt', '11pt' or '12pt'). #'pointsize': '10pt', # Additional stuff for the LaTeX preamble. #'preamble': '', } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, # author, documentclass [howto, manual, or own class]). latex_documents = [ ('index', 'scikit-hyperband.tex', u'scikit-hyperband Documentation', u'Thomas Huijskens', 'manual'), ] # The name of an image file (relative to this directory) to place at the top of # the title page. #latex_logo = None # For "manual" documents, if this is true, then toplevel headings are parts, # not chapters. #latex_use_parts = False # If true, show page references after internal links. #latex_show_pagerefs = False # If true, show URL addresses after external links. #latex_show_urls = False # Documents to append as an appendix to all manuals. #latex_appendices = [] # If false, no module index is generated. #latex_domain_indices = True # -- Options for manual page output --------------------------------------- # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [ ('index', 'scikit-hyperband', u'scikit-hyperband Documentation', [u'Thomas Huijskens'], 1) ] # If true, show URL addresses after external links. #man_show_urls = False # -- Options for Texinfo output ------------------------------------------- # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [ ('index', 'scikit-hyperband', u'scikit-hyperband Documentation', u'Thomas Huijskens', 'scikit-hyperband', 'scikit-learn compatible implementation of hyperband.', 'Miscellaneous'), ] # def generate_example_rst(app, what, name, obj, options, lines): # # generate empty examples files, so that we don't get # # inclusion errors if there are no examples for a class / module # examples_path = os.path.join(app.srcdir, "modules", "generated", # "%s.examples" % name) # if not os.path.exists(examples_path): # # touch file # open(examples_path, 'w').close() # # # def setup(app): # app.connect('autodoc-process-docstring', generate_example_rst) # Documents to append as an appendix to all manuals. #texinfo_appendices = [] # If false, no module index is generated. #texinfo_domain_indices = True # How to display URL addresses: 'footnote', 'no', or 'inline'. #texinfo_show_urls = 'footnote' # If true, do not generate a @detailmenu in the "Top" node's menu. #texinfo_no_detailmenu = False # Example configuration for intersphinx: refer to the Python standard library. intersphinx_mapping = {'http://docs.python.org/': None}
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/gender_tpot_regression/model/gender_tpot_regression.py
training/helpers/gender_tpot_regression/model/gender_tpot_regression.py
import numpy as np import json, pickle import pandas as pd from sklearn.ensemble import ExtraTreesRegressor from sklearn.feature_selection import SelectPercentile, f_regression from sklearn.linear_model import ElasticNetCV from sklearn.model_selection import train_test_split from sklearn.pipeline import make_pipeline, make_union from tpot.builtins import StackingEstimator # NOTE: Make sure that the outcome column is labeled 'target' in the data file g=json.load(open('gender_tpot_regression.json')) tpot_data=np.array(g['labels']) features=np.array(g['data']) training_features, testing_features, training_target, testing_target = \ train_test_split(features, tpot_data, random_state=None) # Average CV score on the training set was: -0.13558964188025885 exported_pipeline = make_pipeline( SelectPercentile(score_func=f_regression, percentile=19), StackingEstimator(estimator=ElasticNetCV(l1_ratio=0.2, tol=0.0001)), StackingEstimator(estimator=ExtraTreesRegressor(bootstrap=True, max_features=0.7000000000000001, min_samples_leaf=14, min_samples_split=15, n_estimators=100)), ElasticNetCV(l1_ratio=0.9, tol=0.001) ) exported_pipeline.fit(training_features, training_target) print('saving classifier to disk') f=open('gender_tpot_regression.pickle','wb') pickle.dump(exported_pipeline,f) f.close()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/gender_tpot_classifier/model/gender_tpot_classifier.py
training/helpers/gender_tpot_classifier/model/gender_tpot_classifier.py
import numpy as np import json, pickle import pandas as pd from sklearn.model_selection import train_test_split from sklearn.pipeline import make_pipeline from sklearn.preprocessing import Normalizer from sklearn.svm import LinearSVC # NOTE: Make sure that the outcome column is labeled 'target' in the data file g=json.load(open('gender_tpot_classifier.json')) tpot_data=np.array(g['labels']) features=np.array(g['data']) training_features, testing_features, training_target, testing_target = \ train_test_split(features, tpot_data, random_state=None) # Average CV score on the training set was: 0.8276292335115866 exported_pipeline = make_pipeline( Normalizer(norm="max"), LinearSVC(C=20.0, dual=True, loss="hinge", penalty="l2", tol=0.0001) ) exported_pipeline.fit(training_features, training_target) print('saving classifier to disk') f=open('gender_tpot_classifier.pickle','wb') pickle.dump(exported_pipeline,f) f.close()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autokaggle/tabular_supervised.py
training/helpers/autokaggle/tabular_supervised.py
from abc import abstractmethod import os from lightgbm import LGBMClassifier, LGBMRegressor from sklearn.model_selection import RandomizedSearchCV from sklearn.model_selection import StratifiedKFold, KFold from sklearn.metrics import roc_auc_score, f1_score, mean_squared_error import numpy as np import random from autokaggle.tabular_preprocessor import TabularPreprocessor from autokaggle.utils import rand_temp_folder_generator, ensure_dir class TabularSupervised: def __init__(self, path=None, verbose=True): """ Initialization function for tabular supervised learner. """ self.verbose = verbose self.is_trained = False self.clf = None self.objective = None self.tabular_preprocessor = None self.path = path if path is not None else rand_temp_folder_generator() ensure_dir(self.path) if self.verbose: print('Path:', path) self.save_filename = os.path.join(self.path, 'lgbm.txt') self.time_limit = None self.lgbm = None def search(self, search_space, search_iter, n_estimators, x, y): if 'n_estimators' in search_space: del search_space['n_estimators'] params = { 'boosting_type': ['gbdt'], 'min_child_weight': [5], 'min_split_gain': [1.0], 'subsample': [0.8], 'colsample_bytree': [0.6], 'max_depth': [10], 'n_estimators': n_estimators, 'num_leaves': [70], 'learning_rate': [0.04], } params.update(search_space) if self.verbose: print(params) folds = 3 score_metric, skf = self.get_skf(folds) random_search = RandomizedSearchCV(self.lgbm, param_distributions=params, n_iter=search_iter, scoring=score_metric, n_jobs=1, cv=skf, verbose=0, random_state=1001) random_search.fit(x, y) self.clf = random_search.best_estimator_ return random_search.best_params_ @abstractmethod def get_skf(self, folds): pass def fit(self, x, y, time_limit=None, data_info=None): """ This function should train the model parameters. Args: x: A numpy.ndarray instance containing the training data. y: training label vector. time_limit: remaining time budget. data_info: meta-features of the dataset, which is an numpy.ndarray describing the feature type of each column in raw_x. The feature type include: 'TIME' for temporal feature, 'NUM' for other numerical feature, and 'CAT' for categorical feature. Both inputs X and y are numpy arrays. If fit is called multiple times on incremental data (train, test1, test2, etc.) you should warm-start your training from the pre-trained model. Past data will NOT be available for re-training. """ if time_limit is None: time_limit = 24 * 60 * 60 self.time_limit = time_limit self.init_lgbm(y) self.tabular_preprocessor = TabularPreprocessor() if x.shape[1] == 0: raise ValueError("No feature exist!") x = self.tabular_preprocessor.fit(x, y, self.time_limit, data_info) if x.shape[0] > 600: grid_train_percentage = max(600.0 / x.shape[0], 0.1) else: grid_train_percentage = 1 grid_n = int(x.shape[0] * grid_train_percentage) idx = random.sample(list(range(x.shape[0])), grid_n) grid_train_x = x[idx, :] grid_train_y = y[idx] while x.shape[0] < 60: x = np.concatenate([x, x], axis=0) y = np.concatenate([y, y], axis=0) response_rate = sum(y) / len(y) if not self.is_trained: # Two-step cross-validation for hyperparameter selection if self.verbose: print('-----------------Search Regularization Params---------------------') if response_rate < 0.005: depth_choice = [5] else: depth_choice = [8, 10] params = { 'min_split_gain': [0.1], 'max_depth': depth_choice, 'min_child_weight': [5, 10, 30, 50, 60, 80, 100], 'colsample_bytree': [0.6, 0.7], 'learning_rate': [0.3], 'subsample': [0.8], 'num_leaves': [80], } search_iter = 14 n_estimators_choice = [50] best_param = self.search( params, search_iter, n_estimators_choice, grid_train_x, grid_train_y) if self.verbose: print('-----------------Search Learning Rate---------------------') for key, value in best_param.items(): best_param[key] = [value] best_param['learning_rate'] = [0.03, 0.045, 0.06, 0.075, 0.85, 0.95, 0.105, 0.12] n_estimators_choice = [100, 150, 200] search_iter = 16 self.search( best_param, search_iter, n_estimators_choice, grid_train_x, grid_train_y) if self.verbose: print('self.clf', self.clf) self.is_trained = True # Fit Model self.clf.fit(x, y) self.clf.booster_.save_model(self.save_filename) if self.verbose: print("The whole available data is: ") print("Real-FIT: dim(X)= [{:d}, {:d}]".format(x.shape[0], x.shape[1])) print('Feature Importance:') print(self.clf.feature_importances_) @abstractmethod def init_lgbm(self, y): pass def predict(self, x_test): """ This function should provide predictions of labels on (test) data. The function predict eventually casdn return probabilities or continuous values. """ x_test = self.tabular_preprocessor.encode(x_test) y = self.clf.predict(x_test, ) if y is None: raise ValueError("Tabular predictor does not exist") return y @abstractmethod def evaluate(self, x_test, y_test): pass def final_fit(self, x_train, y_train): x_train = self.tabular_preprocessor.encode(x_train) self.clf.fit(x_train, y_train) class TabularRegressor(TabularSupervised): """TabularRegressor class. It is used for tabular data regression with lightgbm regressor. """ def __init__(self, path=None): super().__init__(path) self.objective = 'regression' def evaluate(self, x_test, y_test): y_pred = self.predict(x_test) return mean_squared_error(y_test, y_pred) def init_lgbm(self, y): self.lgbm = LGBMRegressor(silent=False, verbose=-1, n_jobs=1, objective=self.objective) def get_skf(self, folds): return 'neg_mean_squared_error', KFold(n_splits=folds, shuffle=True, random_state=1001) class TabularClassifier(TabularSupervised): """TabularClassifier class. It is used for tabular data classification with lightgbm classifier. """ def init_lgbm(self, y): n_classes = len(set(y)) if n_classes == 2: self.objective = 'binary' self.lgbm = LGBMClassifier(silent=False, verbose=-1, n_jobs=1, objective=self.objective) else: self.objective = 'multiclass' self.lgbm = LGBMClassifier(silent=False, verbose=-1, n_jobs=1, num_class=n_classes, objective=self.objective) def evaluate(self, x_test, y_test): if self.verbose: print('objective:', self.objective) y_pred = self.predict(x_test) results = None if self.objective == 'binary': results = roc_auc_score(y_test, y_pred) elif self.objective == 'multiclass': results = f1_score(y_test, y_pred, average='weighted') return results def get_skf(self, folds): if self.lgbm.objective == 'binary': score_metric = 'roc_auc' skf = StratifiedKFold(n_splits=folds, shuffle=True, random_state=1001) else: score_metric = 'f1_weighted' skf = StratifiedKFold(n_splits=folds, shuffle=True, random_state=1001) return score_metric, skf
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autokaggle/utils.py
training/helpers/autokaggle/utils.py
import os import tempfile import string import random def ensure_dir(directory): """Create directory if it does not exist.""" if not os.path.exists(directory): os.makedirs(directory) def temp_path_generator(): sys_temp = tempfile.gettempdir() path = os.path.join(sys_temp, 'autokaggle') return path def rand_temp_folder_generator(): """Create and return a temporary directory with the path name '/temp_dir_name/autokeras' (E:g:- /tmp/autokeras).""" chars = string.ascii_uppercase + string.digits size = 6 random_suffix = ''.join(random.choice(chars) for _ in range(size)) sys_temp = temp_path_generator() path = sys_temp + '_' + random_suffix ensure_dir(path) return path
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/autokaggle/tabular_preprocessor.py
training/helpers/autokaggle/tabular_preprocessor.py
import numpy as np from pandas import DataFrame from scipy.stats import pearsonr LEVEL_HIGH = 32 def parallel_function(labels, first_batch_keys, task): if task == 'label': if min(labels) > first_batch_keys: labels = labels - np.min(labels) return labels.reshape(labels.shape[0], 1) elif task == 'frequency': cat_dict = {} n_rows = labels.shape[0] labels = np.expand_dims(labels, axis=1) if min(labels) > first_batch_keys: labels = labels - np.min(labels) frequencies = np.zeros((n_rows, 1)) for row_index in range(n_rows): key = labels[row_index, 0] if key in cat_dict: cat_dict[key] += 1 else: cat_dict[key] = 1 n_level = len(cat_dict) key_to_frequency = {} for key in cat_dict.keys(): key_to_frequency[key] = cat_dict[key] / n_rows * n_level for row_index in range(n_rows): key = labels[row_index, 0] frequencies[row_index][0] = key_to_frequency[key] return frequencies elif task == 'num_cat': df = DataFrame(data=labels) return df.join(df.groupby(1)[0].mean(), rsuffix='r', on=1).values[:, -1:] elif task == 'cat_cat': df = DataFrame(data=labels) df[3] = list(range(len(labels))) return df.join(df.groupby([0, 1]).count(), rsuffix='r', on=(0, 1)).values[:, -1:] elif task == 'train_num_cat': y = first_batch_keys[0] df = DataFrame(data=labels) fe = df.join(df.groupby(1)[0].mean(), rsuffix='r', on=1).values[:, -1:] mu = abs(pearsonr(np.squeeze(np.array(fe)), y)[0]) if np.isnan(mu): mu = 0 return [[first_batch_keys[1], first_batch_keys[2], mu, mu], first_batch_keys[3]] elif task == 'train_cat_cat': y = first_batch_keys[0] df = DataFrame(data=labels) df[3] = list(range(len(labels))) fe = df.join(df.groupby([0, 1]).count(), rsuffix='r', on=(0, 1)).values[:, -1:] mu = abs(pearsonr(np.squeeze(np.array(fe)), y)[0]) if np.isnan(mu): mu = 0 return [[first_batch_keys[1], first_batch_keys[2], mu], first_batch_keys[3]] return None def call_parallel(tasks): results = [] for t in tasks: results.append(parallel_function(t[0], t[1], t[2])) return results class TabularPreprocessor: def __init__(self): """ Initialization function for tabular preprocessor. """ self.num_cat_pair = {} self.total_samples = 0 self.cat_to_int_label = {} self.n_first_batch_keys = {} self.high_level_cat_keys = [] self.feature_add_high_cat = 0 self.feature_add_cat_num = 0 self.feature_add_cat_cat = 0 self.order_num_cat_pair = {} self.rest = None self.budget = None self.data_info = None self.n_time = None self.n_num = None self.n_cat = None def remove_useless(self, x): self.rest = np.where(np.max(x, 0) - np.min(x, 0) != 0)[0] return x[:, self.rest] def process_time(self, x): cols = range(self.n_time) if len(cols) > 10: cols = cols[:10] x_time = x[:, cols] for i in cols: for j in range(i + 1, len(cols)): x = np.append(x, np.expand_dims(x_time[:, i] - x_time[:, j], 1), 1) return x def extract_data(self, raw_x): # only get numerical variables ret = np.concatenate([raw_x['TIME'], raw_x['NUM'], raw_x['CAT']], axis=1) n_rows = ret.shape[0] n_num_col = ret.shape[1] - self.n_cat n_cat_col = self.n_cat if n_cat_col <= 0: return ret.astype(np.float64) # preprocess (multi-value) categorical data for col_index in range(n_num_col, n_num_col + n_cat_col): for row_index in range(n_rows): key = str(ret[row_index, col_index]) if key in self.cat_to_int_label[col_index]: ret[row_index, col_index] = self.cat_to_int_label[col_index][key] continue new_value = len(self.cat_to_int_label[col_index]) self.cat_to_int_label[col_index][key] = new_value ret[row_index, col_index] = new_value return ret.astype(np.float64) def cat_to_num(self, x, y=None): if y is not None: mark = self.n_time + self.n_num for col_index in range(self.n_time + self.n_num, self.n_time + self.n_num + self.n_cat): if self.n_first_batch_keys[col_index] <= LEVEL_HIGH: self.num_cat_pair[mark] = (col_index,) mark += 1 else: self.num_cat_pair[mark] = (col_index, col_index) mark += 1 mark_1 = 0 tasks = [] for i, cat_col_index1 in enumerate(self.high_level_cat_keys): for cat_col_index2 in self.high_level_cat_keys[i + 1:]: tasks.append((x[:, (cat_col_index1, cat_col_index2)], [y, cat_col_index1, cat_col_index2, mark_1], 'train_cat_cat')) mark_1 += 1 all_results = call_parallel(tasks) num_cat_pair_1 = {} pearsonr_dict_1 = {} for result in all_results: if result[0][-1] > 0.001: pearsonr_dict_1[result[1]] = result[0][-1] num_cat_pair_1[result[1]] = result[0] pearsonr_high_1 = sorted(pearsonr_dict_1, key=pearsonr_dict_1.get, reverse=True)[:self.feature_add_cat_cat] num_cat_pair_1 = {key: num_cat_pair_1[key] for key in pearsonr_high_1} num_cat_pair_1 = {i + mark: num_cat_pair_1[key] for i, key in enumerate(num_cat_pair_1)} self.num_cat_pair.update(num_cat_pair_1) mark += len(pearsonr_high_1) mark_2 = 0 tasks_2 = [] for cat_col_index in self.high_level_cat_keys: for num_col_index in range(self.n_time, self.n_time + self.n_num): tasks_2.append((x[:, (num_col_index, cat_col_index)], [y, num_col_index, cat_col_index, mark_2], 'train_num_cat')) mark_2 += 1 all_results = call_parallel(tasks_2) num_cat_pair_2 = {} pearsonr_dict_2 = {} for result in all_results: if result[0][-1] > 0.001: pearsonr_dict_2[result[1]] = result[0][-1] num_cat_pair_2[result[1]] = result[0] pearsonr_high_2 = sorted(pearsonr_dict_2, key=pearsonr_dict_2.get, reverse=True)[:self.feature_add_cat_num] num_cat_pair_2 = {key: num_cat_pair_2[key] for key in pearsonr_high_2} num_cat_pair_2 = {i + mark: num_cat_pair_2[key] for i, key in enumerate(num_cat_pair_2)} self.num_cat_pair.update(num_cat_pair_2) self.order_num_cat_pair = sorted(list(self.num_cat_pair.keys())) print('num_cat_pair_2:', num_cat_pair_2) tasks = [] for key in self.order_num_cat_pair: if len(self.num_cat_pair[key]) == 1: (col_index,) = self.num_cat_pair[key] tasks.append((x[:, col_index], self.n_first_batch_keys[col_index], 'label')) if len(self.num_cat_pair[key]) == 2: (col_index, col_index) = self.num_cat_pair[key] tasks.append((x[:, col_index], self.n_first_batch_keys[col_index], 'frequency')) if len(self.num_cat_pair[key]) == 3: (cat_col_index1, cat_col_index2, mu) = self.num_cat_pair[key] tasks.append((x[:, (cat_col_index1, cat_col_index2)], self.n_first_batch_keys[cat_col_index1], 'cat_cat')) elif len(self.num_cat_pair[key]) == 4: (num_col_index, cat_col_index, mu, a) = self.num_cat_pair[key] tasks.append((x[:, (num_col_index, cat_col_index)], self.n_first_batch_keys[cat_col_index], 'num_cat')) results = call_parallel(tasks) all_num = x.shape[1] - self.n_cat results = [x[:, :all_num]] + results ret = np.concatenate(results, axis=1) return ret def fit(self, raw_x, y, time_limit, data_info): """ This function should train the model parameters. Args: raw_x: a numpy.ndarray instance containing the training data. y: training label vector. time_limit: remaining time budget. data_info: meta-features of the dataset, which is an numpy.ndarray describing the feature type of each column in raw_x. The feature type include: 'TIME' for temporal feature, 'NUM' for other numerical feature, and 'CAT' for categorical feature. """ # Get Meta-Feature self.budget = time_limit self.data_info = data_info if data_info is not None else self.extract_data_info(raw_x) print('QQ: {}'.format(self.data_info)) self.n_time = sum(self.data_info == 'TIME') self.n_num = sum(self.data_info == 'NUM') self.n_cat = sum(self.data_info == 'CAT') self.total_samples = raw_x.shape[0] print('QQ1: {}'.format(self.n_time)) print('QQ2: {}'.format(self.n_num)) print('QQ3: {}'.format(self.n_cat)) raw_x = {'TIME': raw_x[:, self.data_info == 'TIME'], 'NUM': raw_x[:, self.data_info == 'NUM'], 'CAT': raw_x[:, self.data_info == 'CAT']} for col_index in range(self.n_num + self.n_time, self.n_num + self.n_time + self.n_cat): self.cat_to_int_label[col_index] = {} x = self.extract_data(raw_x) d_size = x.shape[0] * x.shape[1] / self.budget if d_size > 35000: self.feature_add_high_cat = 0 else: self.feature_add_high_cat = 10 # Iterate cat features for col_index in range(self.n_num + self.n_time, self.n_num + self.n_time + self.n_cat): self.n_first_batch_keys[col_index] = len(self.cat_to_int_label[col_index]) high_level_cat_keys_tmp = sorted(self.n_first_batch_keys, key=self.n_first_batch_keys.get, reverse=True)[ :self.feature_add_high_cat] for i in high_level_cat_keys_tmp: if self.n_first_batch_keys[i] > 1e2: self.high_level_cat_keys.append(i) # Convert NaN to zeros x = np.nan_to_num(x) # Encode high-order categorical data to numerical with frequency x = self.cat_to_num(x, y) x = self.process_time(x) x = self.remove_useless(x) return x def encode(self, raw_x, time_limit=None): """ This function should train the model parameters. Args: raw_x: a numpy.ndarray instance containing the training/testing data. time_limit: remaining time budget. Both inputs X and y are numpy arrays. If fit is called multiple times on incremental data (train, test1, test2, etc.) you should warm-start your training from the pre-trained model. Past data will NOT be available for re-training. """ # Get Meta-Feature if time_limit is None: if self.budget is None: time_limit = 24 * 60 * 60 self.budget = time_limit else: self.budget = time_limit raw_x = {'TIME': raw_x[:, self.data_info == 'TIME'], 'NUM': raw_x[:, self.data_info == 'NUM'], 'CAT': raw_x[:, self.data_info == 'CAT']} x = self.extract_data(raw_x) # Convert NaN to zeros x = np.nan_to_num(x) # Encode high-order categorical data to numerical with frequency x = self.cat_to_num(x) x = self.process_time(x) if self.rest is not None: x = x[:, self.rest] return x @staticmethod def extract_data_info(raw_x): """ This function extracts the data info automatically based on the type of each feature in raw_x. Args: raw_x: a numpy.ndarray instance containing the training data. """ data_info = [] row_num, col_num = raw_x.shape for col_idx in range(col_num): try: raw_x[:, col_idx].astype(np.float) data_info.append('NUM') except: data_info.append('CAT') return np.array(data_info)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/plda/__init__.py
training/helpers/plda/__init__.py
# Copyright 2017 Ravi Sojitra. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== from .plda import Classifier from .plda import Model
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/plda/plda/model.py
training/helpers/plda/plda/model.py
# Copyright 2017 Ravi Sojitra. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== import numpy as np from sklearn.decomposition import PCA from scipy.stats import multivariate_normal as gaussian from .optimizer import get_prior_params from .optimizer import get_posterior_params from .optimizer import get_posterior_predictive_params from .optimizer import optimize_maximum_likelihood from .optimizer import calc_scatter_matrices def get_space_walk(from_space, to_space): U_model_to_D = ['U_model', 'U', 'X', 'D'] D_to_U_model = U_model_to_D[::-1] assert from_space in U_model_to_D and to_space in U_model_to_D from_idx = U_model_to_D.index(from_space) to_idx = U_model_to_D.index(to_space) if to_idx < from_idx: spaces = D_to_U_model from_idx = D_to_U_model.index(from_space) to_idx = D_to_U_model.index(to_space) else: spaces = U_model_to_D from_spaces = [x for x in spaces[from_idx: to_idx]] to_spaces = [x for x in spaces[from_idx + 1: to_idx + 1]] return zip(from_spaces, to_spaces) def transform_D_to_X(data, pca): return data if pca is None else pca.transform(data) def transform_X_to_U(data, inv_A, m): return np.matmul(data - m, inv_A.T) def transform_U_to_U_model(data, relevant_U_dims): return data[..., relevant_U_dims] def transform_U_model_to_U(data, relevant_U_dims, U_dimensionality): shape = (*data.shape[:-1], U_dimensionality) U = np.zeros(shape) U[..., relevant_U_dims] = data return U def transform_U_to_X(data, A, m): return m + np.matmul(data, A.T) def transform_X_to_D(data, pca): return data if pca is None else pca.inverse_transform(data) class Model: def __init__(self, row_wise_data, labels, n_principal_components=None): assert len(row_wise_data.shape) == 2 assert len(labels) == row_wise_data.shape[0] self.pca = None self.m = None self.A = None self.Psi = None self.relevant_U_dims = None self.inv_A = None self.prior_params = None self.posterior_params = None self.posterior_predictive_params = None self.fit(row_wise_data, labels, n_principal_components) def calc_logp_posterior(self, v_model, category): assert v_model.shape[-1] == self.get_dimensionality('U_model') mean = self.posterior_params[category]['mean'] cov_diag = self.posterior_params[category]['cov_diag'] return gaussian(mean, np.diag(cov_diag)).logpdf(v_model) def calc_logp_posterior_predictive(self, U_model, category): assert U_model.shape[-1] == self.get_dimensionality('U_model') mean = self.posterior_predictive_params[category]['mean'] cov_diag = self.posterior_predictive_params[category]['cov_diag'] return gaussian(mean, np.diag(cov_diag)).logpdf(U_model) def calc_logp_marginal_likelihood(self, U_model): """ Computes the log marginal likelihood on axis=-2. """ assert U_model.shape[-1] == self.get_dimensionality('U_model') if len(U_model.shape) == 1: U_model = U_model[None, :] n = U_model.shape[-2] psi_diag = self.prior_params['cov_diag'] n_psi_plus_eye = n * psi_diag + 1 log_constant = -.5 * n * np.log(2 * np.pi) log_constant += -.5 * np.log(n_psi_plus_eye) sum_of_squares = np.sum(U_model ** 2, axis=-2) log_exponent_1 = -.5 * sum_of_squares mean = U_model.mean(axis=-2) log_exponent_2 = .5 * (n ** 2 * psi_diag * mean ** 2) log_exponent_2 /= n_psi_plus_eye logp_ml = log_constant + log_exponent_1 + log_exponent_2 logp_ml = np.sum(logp_ml, axis=-1) return logp_ml def calc_logp_prior(self, v_model): assert v_model.shape[-1] == self.get_dimensionality('U_model') mean = self.prior_params['mean'] cov_diag = self.prior_params['cov_diag'] return gaussian(mean, np.diag(cov_diag)).logpdf(v_model) def fit(self, data, labels, n_principal_components=None): if n_principal_components is None: S_b, S_w = calc_scatter_matrices(data, labels) matrix_rank = np.linalg.matrix_rank(S_w) else: matrix_rank = n_principal_components if matrix_rank != data.shape[-1]: self.pca = PCA(n_components=matrix_rank) self.pca.fit(data) X = self.transform(data, from_space='D', to_space='X') self.m, self.A, self.Psi, self.relevant_U_dims, self.inv_A = \ optimize_maximum_likelihood(X, labels) U_model = self.transform(X, from_space='X', to_space='U_model') self.prior_params = \ get_prior_params(self.Psi, self.relevant_U_dims) self.posterior_params = \ get_posterior_params(U_model, labels, self.prior_params) self.posterior_predictive_params = \ get_posterior_predictive_params(self.posterior_params) def get_dimensionality(self, space): if space == 'U_model': return self.relevant_U_dims.shape[0] elif space == 'U': return self.A.shape[0] elif space == 'X': return self.A.shape[0] elif space == 'D': if self.pca is None: return self.m.shape[0] else: return self.pca.n_features_ else: raise ValueError def transform(self, data, from_space, to_space): """ Potential_spaces: 'D' <---> 'X' <---> 'U' <---> 'U_model'. DESCRIPTION There are 6 basic transformations to move back and forth between the data space, 'D', and the model's space, 'U_model': 1. From D to X. (i.e. from data space to preprocessed space) Uses the minimum number of components from Principal Components Analysis that captures 100% of the variance in the data. 2. From X to U. (i.e. from preprocessed space to latent space) See the bottom of p.533 of Ioffe 2006. 3. From U to U_model. (i.e. from latent space to the model space) See Fig 2 on p.537 of Ioffe 2006. 4. From U_model to U. (i.e. from the model space to latent space) 5. From U to X. (i.e. from the latent space to the preprocessed space) 6. From X to D. (i.e. from the preprocessed space to the data space) """ if len(data.shape) == 1: data = data[None, :] if from_space == 'D' and to_space == 'X': return transform_D_to_X(data, self.pca) elif from_space == 'X' and to_space == 'U': return transform_X_to_U(data, self.inv_A, self.m) elif from_space == 'U' and to_space == 'U_model': return transform_U_to_U_model(data, self.relevant_U_dims) elif from_space == 'U_model' and to_space == 'U': dim = self.get_dimensionality('U') return transform_U_model_to_U(data, self.relevant_U_dims, dim) elif from_space == 'U' and to_space == 'X': return transform_U_to_X(data, self.A, self.m) elif from_space == 'X' and to_space == 'D': return transform_X_to_D(data, self.pca) else: transformed = data for space_1, space_2 in get_space_walk(from_space, to_space): transformed = self.transform(transformed, space_1, space_2) return transformed
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/plda/plda/optimizer.py
training/helpers/plda/plda/optimizer.py
# Copyright 2017 Ravi Sojitra. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== import numpy as np from scipy.linalg import eigh def optimize_maximum_likelihood(X, labels): """ Performs the optimization in Fig. 2 of p.537 of Ioffe 2006. DESCRIPTION - The main model parameters are `m`, `A`, and `Psi`. - However, to improve the performance (speed and numerical stability) of the plda.Model object, inv_A and relevant_U_dims are also returned here. ADDITIONAL NOTES Be sure to test that np.cov(X.T) is full rank before running this. Recall that there are 4 \"spaces\": 'D' (data) <---> 'X' (preprocessed) <---> 'U' (latent) <---> 'U_model' ARGUMENTS X (numpy.ndarray), shape=(n_data, n_dimensions) - Data in statistics format, i.e. row-wise. labels (list or numpy.ndarray), length=X.shape[0] - Labels for the data in `X`. - Must be sorted in the same order as `X`. RETURNS m (numpy.ndarray), shape=X.shape[-1] - The mean of the row vectors in X. - This is the prior mean fitted via maximum likelihood. A (numpy.ndarray), shape=(X.shape[-1], X.shape[-1]) - Transformation from X space to the latent U space. Psi (numpy.ndarray), shape=(X.shape[-1], X.shape[-1]) - The covariance matrix of the prior distribution on the category means in U space. relevant_U_dims (numpy.ndarray), shape=(len(np.unique(labels)) - 1,) - The \"effective\" latent dimensions, i.e. the ones that are actually used by the model. inv_A (numpy.ndarray), shape=A.shape - The inverse of the matrix A. - Transformation from the latent U space to the X space. """ assert len(X.shape) == 2 assert X.shape[0] == len(labels) m = calc_m(X) S_b, S_w = calc_scatter_matrices(X, labels) W = calc_W(S_b, S_w) Lambda_b = calc_Lambda_b(S_b, W) Lambda_w = calc_Lambda_w(S_w, W) n_avg = calc_n_avg(labels) A = calc_A(n_avg, Lambda_w, W) inv_A = np.linalg.inv(A) Psi = calc_Psi(Lambda_w, Lambda_b, n_avg) relevant_U_dims = get_relevant_U_dims(Psi) return m, A, Psi, relevant_U_dims, inv_A def as_dictionary_of_dictionaries(labels, means, cov_diags): """ Dictionary storing one dictionary of parameters per category. """ assert len(labels) == len(means) == len(cov_diags) all_params = dict() for label, mean, cov_diag in zip(labels, means, cov_diags): category_params = dict() category_params['mean'] = mean category_params['cov_diag'] = cov_diag all_params[label] = category_params return all_params def calc_A(n_avg, Lambda_w, W): """ See Fig. 2 on p.537 of Ioffe 2006. """ Lambda_w_diagonal = Lambda_w.diagonal() # Should be diagonal matrix. inv_W_T = np.linalg.inv(W.T) return inv_W_T * (n_avg / (n_avg - 1) * Lambda_w_diagonal) ** .5 def calc_Lambda_b(S_b, W): """ See Fig. 2 on p.537 of Ioffe 2006. """ return np.matmul(np.matmul(W.T, S_b), W) def calc_Lambda_w(S_w, W): """ See Fig. 2 on p.537 of Ioffe 2006. """ return np.matmul(np.matmul(W.T, S_w), W) def calc_m(X): """ See Fig. 2 on p.537 of Ioffe 2006. """ assert len(X.shape) == 2 return X.mean(axis=0) def calc_n_avg(Y): """ This is the \"hack\" suggested in Fig 2 on p.537 of Ioffe 2006. """ unique = np.unique(Y) return len(Y) / unique.shape[0] def calc_Psi(Lambda_w, Lambda_b, n_avg): """ See Fig. 2 on p.537 of Ioffe 2006. """ Lambda_w_diagonal = Lambda_w.diagonal() # Should be diagonal matrix. Lambda_b_diagonal = Lambda_b.diagonal() # Should be diagonal matrix. Psi = (n_avg - 1) / n_avg * Lambda_b_diagonal / Lambda_w_diagonal Psi -= 1 / n_avg Psi[Psi <= 0] = 0 return np.diag(Psi) def calc_scatter_matrices(X, Y): """ See Equations (1) on p.532 of Ioffe 2006. """ assert len(X.shape) == 2 assert X.shape[0] == len(Y) unique_labels = np.unique(Y) labels = np.asarray(Y) m = calc_m(X) N = X.shape[0] cov_ks = [] m_ks = [] n_ks = [] for k in unique_labels: bool_idxs = labels == k X_k = X[bool_idxs] m_ks.append(X_k.mean(axis=0)) n_ks.append(bool_idxs.sum()) cov_ks.append(np.cov(X_k.T)) n_ks = np.asarray(n_ks) m_ks = np.asarray(m_ks) m_ks_minus_m = m_ks - m S_b = np.matmul(m_ks_minus_m.T * (n_ks / N), m_ks_minus_m) S_w = np.asarray(cov_ks) * ((n_ks - 1) / N)[:, None, None] S_w = np.sum(S_w, axis=0) return S_b, S_w def calc_W(S_b, S_w): """ See Fig. 2 on p.537 of Ioffe 2006. """ eigenvalues, eigenvectors = eigh(S_b, S_w) return eigenvectors def get_posterior_params(U_model, Y, prior_params): labels = np.asarray(Y) prior_cov_diagonal = prior_params['cov_diag'] cov_diags = [] means = [] categories = [] for k in np.unique(labels): bool_idxs = labels == k U_model_k = U_model[bool_idxs] n_k = bool_idxs.sum() cov_diag = prior_cov_diagonal / (1 + n_k * prior_cov_diagonal) mean = U_model_k.sum(axis=0) * cov_diag cov_diags.append(cov_diag) means.append(mean) categories.append(k) return as_dictionary_of_dictionaries(categories, means, cov_diags) def get_posterior_predictive_params(posterior_params): """ Likelihood covariance matrix is an Identity matrix. """ pp_params = posterior_params.copy() for k, k_params in pp_params.items(): k_params['cov_diag'] += 1 return pp_params def get_prior_params(Psi, dims): """ See Equation (2) on p.533 of Ioffe 2006. """ cov_diag = Psi.diagonal()[dims] mean = np.zeros(dims.shape) return {'mean': mean, 'cov_diag': cov_diag} def get_relevant_U_dims(Psi): """ See Fig. 2 on p.537 of Ioffe 2006. """ relevant_dims = np.squeeze(np.argwhere(Psi.diagonal() != 0)) if relevant_dims.shape == (): relevant_dims = relevant_dims.reshape(1,) return relevant_dims
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/plda/plda/__init__.py
training/helpers/plda/plda/__init__.py
# Copyright 2017 Ravi Sojitra. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== from .classifier import Classifier from .model import Model
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/plda/plda/classifier.py
training/helpers/plda/plda/classifier.py
# Copyright 2017 Ravi Sojitra. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== import numpy as np from scipy.special import logsumexp from .model import Model class Classifier: def __init__(self): self.model = None def fit_model(self, X, Y, n_principal_components=None): self.model = Model(X, Y, n_principal_components) def predict(self, data, space='D', normalize_logps=False): """ Classifies data into categories present in the training data. DESCRIPTION Predictions are the MAP estimates, i.e. categories with the highest probabilities, following the procedure described in the first sentence on p.535 of Ioffe 2006. See the `calc_logp_pp_categories()` method for documentation on the actual equations. ARGUMENT data (numpy.ndarray), shape=(..., data_dimensionality) - Data must follow statistics convention (row-wise). - The last dimension of the array corresponds to data dimension. - The dimensionality of the data depends on the space (see below). PARAMETERS space (str) - Must be either 'D', 'X', 'U', or 'U_model', where 'D' is the data space, 'X' is the preprocessed space, 'U' is the latent space, and 'U_model' is the subspace of 'U' the model works in: 'D' <---> 'X' <---> 'U' <---> 'U_model'. - See `transform()` method in model.py for details on the relationship between spaces. normalize_logps (bool) - Whether or not to normalize the posterior predictive probabilities before returning them. RETURNS predictions (numpy.ndarray), shape=data.shape[:-1] logps (numpy.ndarray), shape=(*data.shape[:-1], n_categories) - Log posterior predictive probabilities for each category, if normalize_logps = False. - Log classification probabilities for each category, if normalize_logps = True. These are just the normalized posterior predictive probabilities, aka model certainties. """ if space != 'U_model': data = self.model.transform(data, from_space=space, to_space='U_model') logpps_k, K = self.calc_logp_pp_categories(data, normalize_logps) predictions = K[np.argmax(logpps_k, axis=-1)] return predictions, logpps_k def calc_logp_pp_categories(self, data, normalize_logps): """ Computes log posterior predictive probabilities for each category. DESCRIPTION The posterior predictive comes from p.535 of Ioffe 2006. The classification procedure is described in the first sentence of p.535, which clearly implies the prior on categories to be uniform. LATEX EQUATIONS Normalized posterior predictive (classification certainty): ``` \begin{align} p(y_* = T \mid \mathbf{u}_*) &= \frac{ p(\mathbf{u}_* \vert \mathbf{u}_1^T, \mathbf{u}_2^T, \dots, \mathbf{u}_n^T) }{\sum\limits_{k \in K} p(\mathbf{u}_* \vert \mathbf{u}_1^k, \mathbf{u}_2^k, \dots, \mathbf{u}_n^k) }, \end{align} ``` Posterior predictive ``` \begin{align} p(\mathbf{u}_* \vert \mathbf{u}_1^k, \mathbf{u}_2^k, \dots, \mathbf{u}_n^k) &= \int \dots \int p(\mathbf{u}_* \vert \mathbf{v}) p(\mathbf{v} \vert \mathbf{u}_1^k, \mathbf{u}_2^k, \dots, \mathbf{u}_n^k) d\mathbf{v} \\ &= \mathcal{N} \left( \mathbf{u}_* \mid \frac{ n \mathbf{\Psi} }{n \mathbf{\Psi} + \mathbf{I} } \mathbf{\bar{u}}^k, \mathbf{I} + \frac{\mathbf{\Psi} }{n \mathbf{\Psi} + \mathbf{I} } \right) \end{align} ``` ARGUMENT See documentation for the `predict()` method. PARAMTER See documentation for the `predict()` method. """ assert type(normalize_logps) == bool logpps_by_category = [] K = self.get_categories() for k in K: logpps_k = self.model.calc_logp_posterior_predictive(data, k) logpps_by_category.append(logpps_k) logpps_by_category = np.stack(logpps_by_category, axis=-1) if normalize_logps: norms = logsumexp(logpps_by_category, axis=-1) logps = logpps_by_category - norms[..., None] else: logps = logpps_by_category return logps, np.asarray(K) def get_categories(self): return [k for k in self.model.posterior_params.keys()]
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/setup.py
training/helpers/keras_compressor/setup.py
from setuptools import find_packages, setup setup( name='keras_compressor', version='0.0.1', packages=find_packages( exclude=['example'], ), url='', license='Apache License v2', author='Kosuke Kusano', author_email='kosuke_kusano@dwango.co.jp', description='', install_requires=[ 'numpy', 'h5py', 'keras>=2.0.0', 'scipy', 'scikit-learn', ], scripts=['bin/keras-compressor.py'], )
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/bin/keras-compressor.py
training/helpers/keras_compressor/bin/keras-compressor.py
#!/usr/bin/env python import argparse import logging import keras import keras.backend as K import numpy from keras.models import load_model from keras_compressor.compressor import compress def count_total_params(model): """Counts the number of parameters in a model See: https://github.com/fchollet/keras/blob/172397ebf45d58ba256c10004c6fce8b40df286b/keras/utils/layer_utils.py#L114-L117 :param model: Keras model instance :return: trainable_count, non_trainable_count :rtype: tuple of int """ trainable_count = int( numpy.sum([K.count_params(p) for p in set(model.trainable_weights)])) non_trainable_count = int( numpy.sum([K.count_params(p) for p in set(model.non_trainable_weights)])) return trainable_count, non_trainable_count def gen_argparser(): parser = argparse.ArgumentParser(description='compress keras model') parser.add_argument('model', type=str, metavar='model.h5', help='target model, whose loss is specified by `model.compile()`.') parser.add_argument('compressed', type=str, metavar='compressed.h5', help='compressed model path') parser.add_argument('--error', type=float, default=0.1, metavar='0.1', help='layer-wise acceptable error. ' 'If this value is larger, compressed model will be ' 'less accurate and achieve better compression rate. ' 'Default: 0.1') parser.add_argument('--log-level', type=str, default='INFO', choices=['CRITICAL', 'ERROR', 'WARNING', 'INFO', 'DEBUG'], help='log level. Default: INFO') return parser def main(): parser = gen_argparser() args = parser.parse_args() logging.basicConfig(level=getattr(logging, args.log_level)) model = load_model(args.model) # type: keras.models.Model total_params_before = sum(count_total_params(model)) model = compress(model, acceptable_error=args.error) total_params_after = sum(count_total_params(model)) model.save(args.compressed) print('\n'.join(( 'Compressed model', ' before #params {:>20,d}', ' after #params {:>20,d} ({:.2%})', )).format( total_params_before, total_params_after, 1 - float(total_params_after) / total_params_before, )) if __name__ == '__main__': main()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/cifar10/train.py
training/helpers/keras_compressor/example/cifar10/train.py
from keras import backend as K from keras.callbacks import EarlyStopping from keras.datasets import cifar10 from keras.layers import BatchNormalization, Conv2D, Dense, Dropout, Flatten, Input, MaxPool2D from keras.models import Model from keras.preprocessing.image import ImageDataGenerator from keras.utils.np_utils import to_categorical def preprocess(X): return X.astype('float32') / 255 * 2 - 1 class_num = 10 batch_size = 128 epochs = 300 (X_train, y_train), (X_test, y_test) = cifar10.load_data() X_train, X_test = preprocess(X_train), preprocess(X_test) y_train, y_test = to_categorical(y_train), to_categorical(y_test) _, img_rows, img_cols, channels = X_train.shape if K.image_data_format() == 'channels_first': X_train = X_train.reshape(X_train.shape[0], channels, img_rows, img_cols) X_test = X_test.reshape(X_test.shape[0], channels, img_rows, img_cols) else: X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, channels) X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, channels) def gen_model(): # refer to http://torch.ch/blog/2015/07/30/cifar.html img_input = Input(shape=(img_rows, img_cols, channels)) h = img_input h = Conv2D(64, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.3)(h) h = Conv2D(64, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = MaxPool2D((2, 2), strides=(2, 2))(h) h = Conv2D(128, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(128, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = MaxPool2D((2, 2), strides=(2, 2))(h) h = Conv2D(256, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(256, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(256, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = MaxPool2D((2, 2), strides=(2, 2))(h) h = Conv2D(512, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(512, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(512, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = MaxPool2D((2, 2), strides=(2, 2))(h) h = Conv2D(512, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(512, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = Dropout(0.4)(h) h = Conv2D(512, (3, 3), activation='relu', padding='same')(h) h = BatchNormalization()(h) h = MaxPool2D((2, 2), strides=(2, 2))(h) h = Flatten()(h) h = Dropout(0.5)(h) h = Dense(512, activation='relu')(h) h = BatchNormalization()(h) h = Dropout(0.5)(h) h = Dense(class_num, activation='softmax')(h) model = Model(img_input, h) return model model = gen_model() model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) datagen = ImageDataGenerator( zoom_range=0.05, width_shift_range=0.1, height_shift_range=0.1, horizontal_flip=True, ) datagen.fit(X_train) model.fit_generator( datagen.flow(X_train, y_train, batch_size=batch_size), steps_per_epoch=X_train.shape[0] // batch_size, epochs=epochs, validation_data=(X_test, y_test), callbacks=[ EarlyStopping(patience=20), ], ) score = model.evaluate(X_test, y_test) print('test accuracy: ', score[1]) # re-compile model # not to save optimizer variables in model data model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.save('model_raw.h5')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/cifar10/compress.py
training/helpers/keras_compressor/example/cifar10/compress.py
import logging from keras.models import load_model from keras_compressor.compressor import compress logging.basicConfig( level=logging.INFO, ) model = load_model('./model_raw.h5') model = compress(model, 3e-1) model.save('./model_compressed.h5')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/cifar10/finetune.py
training/helpers/keras_compressor/example/cifar10/finetune.py
import keras.backend as K import keras.callbacks as C from keras.datasets import cifar10 from keras.models import load_model from keras.utils.np_utils import to_categorical from keras_compressor import custom_objects def preprocess(X): return X.astype('float32') / 255 * 2 - 1 class_num = 10 batch_size = 128 epochs = 12 (X_train, y_train), (X_test, y_test) = cifar10.load_data() X_train, X_test = preprocess(X_train), preprocess(X_test) y_train, y_test = to_categorical(y_train), to_categorical(y_test) _, img_rows, img_cols, channel = X_train.shape if K.image_data_format() == 'channels_first': X_train = X_train.reshape(X_train.shape[0], channel, img_rows, img_cols) X_test = X_test.reshape(X_test.shape[0], channel, img_rows, img_cols) else: X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, channel) X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, channel) model = load_model('model_compressed.h5', custom_objects) model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.fit( X_train, y_train, batch_size=128, epochs=epochs, validation_data=(X_test, y_test), callbacks=[ C.EarlyStopping(patience=20), ], ) score = model.evaluate(X_test, y_test) print('test accuracy: ', score[1]) # re-compile model # not to save optimizer variables in model data model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.save('model_finetuned.h5')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/cifar10/evaluate.py
training/helpers/keras_compressor/example/cifar10/evaluate.py
import sys from keras import backend as K from keras.datasets import cifar10 from keras.models import load_model from keras.utils import to_categorical from keras_compressor import custom_objects def preprocess(X): return X.astype('float32') / 255 * 2 - 1 def usage(): print('{} model.h5'.format(sys.argv[0])) def load_cifar10(): (X_train, y_train), (X_test, y_test) = cifar10.load_data() X_train, X_test = preprocess(X_train), preprocess(X_test) y_train, y_test = to_categorical(y_train), to_categorical(y_test) _, img_rows, img_cols, channel = X_train.shape if K.image_data_format() == 'channels_first': X_train = X_train.reshape(X_train.shape[0], channel, img_rows, img_cols) X_test = X_test.reshape(X_test.shape[0], channel, img_rows, img_cols) else: X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, channel) X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, channel) return (X_train, y_train), (X_test, y_test) def main(): if len(sys.argv) != 2: usage() sys.exit(1) model_path = sys.argv[1] _, (X_test, y_test) = load_cifar10() model = load_model(model_path, custom_objects) result = model.evaluate(X_test, y_test, verbose=0) model.summary() print(result) if __name__ == '__main__': main()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/mnist/train.py
training/helpers/keras_compressor/example/mnist/train.py
from keras import backend as K from keras.callbacks import EarlyStopping from keras.datasets import mnist from keras.layers import Conv2D, Dense, Dropout, Flatten, Input, MaxPool2D from keras.models import Model from keras.utils.np_utils import to_categorical def preprocess(X): return X.astype('float32') / 255 class_num = 10 batch_size = 128 epochs = 100 (X_train, y_train), (X_test, y_test) = mnist.load_data() X_train, X_test = preprocess(X_train), preprocess(X_test) y_train, y_test = to_categorical(y_train), to_categorical(y_test) _, img_rows, img_cols = X_train.shape if K.image_data_format() == 'channels_first': X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols) X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols) else: X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1) def gen_model(): # from keras mnist tutorial img_input = Input(shape=(img_rows, img_cols, 1)) h = img_input h = Conv2D(32, (3, 3), activation='relu')(h) h = Dropout(0.25)(h) h = Conv2D(64, (3, 3), activation='relu')(h) h = MaxPool2D((2, 2))(h) h = Dropout(0.25)(h) h = Flatten()(h) h = Dense(128, activation='relu')(h) h = Dropout(0.5)(h) h = Dense(class_num, activation='softmax')(h) model = Model(img_input, h) return model model = gen_model() model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.fit( X_train, y_train, batch_size=128, epochs=epochs, validation_data=(X_test, y_test), callbacks=[ EarlyStopping(patience=3), ], ) score = model.evaluate(X_test, y_test) print('test accuracy: ', score[1]) # re-compile model # not to save optimizer variables in model data model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.save('model_raw.h5')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/mnist/compress.py
training/helpers/keras_compressor/example/mnist/compress.py
import logging from keras.models import load_model from keras_compressor.compressor import compress logging.basicConfig( level=logging.INFO, ) model = load_model('./model_raw.h5') model = compress(model, 7e-1) model.save('./model_compressed.h5')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/mnist/finetune.py
training/helpers/keras_compressor/example/mnist/finetune.py
from keras import backend as K from keras.callbacks import EarlyStopping from keras.datasets import mnist from keras.models import load_model from keras.utils.np_utils import to_categorical from keras_compressor.layers import custom_layers def preprocess(X): return X.astype('float32') / 255 class_num = 10 batch_size = 128 epochs = 100 (X_train, y_train), (X_test, y_test) = mnist.load_data() X_train, X_test = preprocess(X_train), preprocess(X_test) y_train, y_test = to_categorical(y_train), to_categorical(y_test) _, img_rows, img_cols = X_train.shape if K.image_data_format() == 'channels_first': X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols) X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols) else: X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1) model = load_model('model_compressed.h5', custom_layers) model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.fit( X_train, y_train, batch_size=128, epochs=epochs, validation_data=(X_test, y_test), callbacks=[ EarlyStopping(patience=3), ], ) score = model.evaluate(X_test, y_test) print('test accuracy: ', score[1]) # re-compile model # not to save optimizer variables in model data model.compile( loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'], ) model.save('model_finetuned.h5')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/example/mnist/evaluate.py
training/helpers/keras_compressor/example/mnist/evaluate.py
import sys from keras import backend as K from keras.datasets import mnist from keras.models import load_model from keras.utils import to_categorical from keras_compressor import custom_objects def preprocess(X): return X.astype('float32') / 255 def usage(): print('{} model.h5'.format(sys.argv[0])) def load_mnist(): (X_train, y_train), (X_test, y_test) = mnist.load_data() X_train, X_test = preprocess(X_train), preprocess(X_test) y_train, y_test = to_categorical(y_train), to_categorical(y_test) _, img_rows, img_cols = X_train.shape if K.image_data_format() == 'channels_first': X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols) X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols) else: X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1) return (X_train, y_train), (X_test, y_test) def main(): if len(sys.argv) != 2: usage() sys.exit(1) model_path = sys.argv[1] _, (X_test, y_test) = load_mnist() model = load_model(model_path, custom_objects) result = model.evaluate(X_test, y_test, verbose=0) model.summary() print(result) if __name__ == '__main__': main()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/utils.py
training/helpers/keras_compressor/keras_compressor/utils.py
from typing import Any, Dict, List from keras.engine.topology import Layer, Node def swap_layer_connection(old_layer: Layer, new_layer: Layer) -> None: '''connect nodes of calc graph for new_layer and disconnect ones for old_layers Keras manages calculation graph by nodes which hold connection between layres. To swap old layer and new layer, it is required to delete nodes of old layer and to create new nodes of new layer. :arg old_layer: Old layer. The connection to/from this layer will be removed. :arg new_layer: New layer. The connection to/from old layer will be connected to/from this layer. :return: None ''' # the set of inbound layer which have old outbound_node inbound_layers = set() # create new inbound nodes for node in old_layer._inbound_nodes: # type: Node Node( new_layer, node.inbound_layers, node.node_indices, node.tensor_indices, node.input_tensors, node.output_tensors, node.input_masks, node.output_masks, node.input_shapes, node.output_shapes, ) inbound_layers.union(set(node.inbound_layers)) # remove old outbound node of inbound layers for layer in inbound_layers: # type: Layer old_nodes = filter( lambda n: n.outbound_layer == old_layer, layer._outbound_nodes, ) for n in old_nodes: # type: Node layer._outbound_nodes.remove(n) # the set of outbound layer which have old inbound_nodes outbound_layers = set() # create new outbound nodes for node in old_layer._outbound_nodes: # type: Node layers = list(node.inbound_layers) while old_layer in layers: idx = layers.index(old_layer) layers[idx] = new_layer Node( node.outbound_layer, layers, node.node_indices, node.tensor_indices, node.input_tensors, node.output_tensors, node.input_masks, node.output_masks, node.input_shapes, node.output_shapes, ) outbound_layers.add(node.outbound_layer) # remove old inbound_node of outbound layers for layer in outbound_layers: # type: Layer old_nodes = filter( lambda n: old_layer in n.inbound_layers, layer._inbound_nodes, ) for n in old_nodes: layer._inbound_nodes.remove(n) def convert_config( base_config: Dict[str, Any], ignore_args: List[str], converts: Dict[str, List[str]], new_kwargs: Dict[str, Any] ) -> Dict[str, Any]: '''convert old layer's config to new layer's config. :param base_config: Base config. Generally a config of old layer. :param ignore_args: Ignore arg names. Not required arg names in new layer, though them is required in old layer. :param converts: Ignore name conversion dictionary, whose key is old layer's arg name in base_config, and whose value is new layer's arg names(list). :param new_kwargs: The new kwargs. :return: Converted config. ''' kwargs = {} for k, v in base_config.items(): if k in ignore_args: continue elif k in converts: for new_k in converts[k]: kwargs[new_k] = v else: kwargs[k] = v kwargs.update(new_kwargs) return kwargs
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/__init__.py
training/helpers/keras_compressor/keras_compressor/__init__.py
from .layers import custom_layers __all__ = ['custom_objects'] custom_objects = dict(custom_layers.items()) # shallow copy
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/compressor.py
training/helpers/keras_compressor/keras_compressor/compressor.py
import logging from collections import defaultdict from typing import Dict, List, Type from keras.engine import Layer, Model from .factorizer import Factorizer from .factorizers.svd import SVDFactorizer from .factorizers.tucker import TuckerFactorizer from .utils import swap_layer_connection logger = logging.getLogger(__name__) def compress(model: Model, acceptable_error: float, factorizers=None) -> Model: """compress model under acceptable error compress each model's layer by using given factorizers. If the factorizer compress the layer, swap the layer and compressed layer by re-creating node on computational graph. :param model: Target model :param acceptable_error: Layer-wize acceptable output error. If this value is smaller the compressed model will be more accurate. The calculation process of this error is depend on each factorizer. So see the implementation. :param factorizers: Applicable factorizers. Factorizer factorize each layer if factorizer can factorize the layer. :return: Compressed model """ if factorizers is None: factorizers = [SVDFactorizer, TuckerFactorizer] layer2factorizers = defaultdict(list) # type: Dict[Type[Layer], List[Type[Factorizer]]] for fact in factorizers: for layer in fact.factorize_target_layers: layer2factorizers[layer].append(fact) for layer_idx, layer in enumerate(model.layers): layer_class = type(layer) if layer_class not in layer2factorizers: logger.info( 'factorizer not found layer:{!r}'.format(layer) ) continue new_layer = None for factorizer in layer2factorizers[layer_class]: # type: Factorizer logger.info( 'factorizer found layer:{!r} factorizer:{!r}'.format( layer, factorizer, ) ) new_layer = factorizer.compress(layer, acceptable_error) if new_layer is None: # failed factorization logger.info( 'factorization failed layer:{!r} factorizer:{!r}'.format( layer, factorizer, ) ) continue else: # succeeded factorization break if new_layer is not None: logger.info( 'swap old/new layer old_layer:{!r} new_layer{!r}'.format( layer, new_layer, ) ) swap_layer_connection(layer, new_layer) model.layers[layer_idx] = new_layer new_model = Model(model.inputs, model.outputs) new_model.compile( optimizer=model.optimizer.__class__.__name__, # TODO: improve here # model.optimizer is instance of Optimizer and hold some variables for target model. # Optimizer must be re-initialized, because compress function changes model structure. loss=model.loss, metrics=model.metrics, loss_weights=model.loss_weights, sample_weight_mode=model.sample_weight_mode, ) return new_model
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/factorizer.py
training/helpers/keras_compressor/keras_compressor/factorizer.py
from typing import List, Optional, Type from keras.layers import Layer class Factorizer: factorize_target_layers = [] # type: List[Type[Layer]] @classmethod def compress(cls, layer: Layer, acceptable_error: float) -> Optional[Layer]: """try to compress the layer under acceptable_error. Outputs compressed layer if compression succeeded. If not, return None. :param layer: :param acceptable_error: :return: """ raise NotImplementedError
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/layers.py
training/helpers/keras_compressor/keras_compressor/layers.py
from keras import backend as K from keras import activations, constraints, initializers, regularizers from keras.engine import InputSpec, Layer from keras.layers import Dense from keras.utils import conv_utils class FactorizedDense(Layer): """Just your regular densely-connected NN layer. This layer based on `keras.layers.core.Dense` and behave like it. `FactorizedDense` implements the operation: `output = activation(dot(dot(input, pre_kernel), post_kernel) + bias)` where `activation` is the element-wise activation function passed as the `activation` argument, `pre_kernel` and `post_kernel` is a weights matrix created by the layer, and `bias` is a bias vector created by the layer (only applicable if `use_bias` is `True`). Note: if the input to the layer has a rank greater than 2, then it is flattened prior to the initial dot product with `pre_kernel`. # Arguments units: Positive integer, dimensionality of the output space. components: Positive integer or None, the size of internal components. If given None, the output is calculated as the same manner as `Dense` layer. activation: Activation function to use (see [activations](../activations.md)). If you don't specify anything, no activation is applied (ie. "linear" activation: `a(x) = x`). use_bias: Boolean, whether the layer uses a bias vector. pre_kernel_initializer: Initializer for the `kernel` weights matrix (see [initializers](../initializers.md)). post_kernel_initializer: Initializer for the `kernel` weights matrix (see [initializers](../initializers.md)). bias_initializer: Initializer for the bias vector (see [initializers](../initializers.md)). pre_kernel_regularizer: Regularizer function applied to the `kernel` weights matrix (see [regularizer](../regularizers.md)). post_kernel_regularizer: Regularizer function applied to the `kernel` weights matrix (see [regularizer](../regularizers.md)). bias_regularizer: Regularizer function applied to the bias vector (see [regularizer](../regularizers.md)). activity_regularizer: Regularizer function applied to the output of the layer (its "activation"). (see [regularizer](../regularizers.md)). kernel_constraint: Constraint function applied to the `kernel` weights matrix (see [constraints](../constraints.md)). bias_constraint: Constraint function applied to the bias vector (see [constraints](../constraints.md)). # Input shape nD tensor with shape: `(batch_size, ..., input_dim)`. The most common situation would be a 2D input with shape `(batch_size, input_dim)`. # Output shape nD tensor with shape: `(batch_size, ..., units)`. For instance, for a 2D input with shape `(batch_size, input_dim)`, the output would have shape `(batch_size, units)`. """ target_layer_types = [Dense] def __init__(self, units, components, activation=None, use_bias=True, pre_kernel_initializer='glorot_uniform', post_kernel_initializer='glorot_uniform', bias_initializer='zeros', pre_kernel_regularizer=None, post_kernel_regularizer=None, bias_regularizer=None, activity_regularizer=None, pre_kernel_constraint=None, post_kernel_constraint=None, bias_constraint=None, **kwargs): if 'input_shape' not in kwargs and 'input_dim' in kwargs: kwargs['input_shape'] = (kwargs.pop('input_dim'),) super(FactorizedDense, self).__init__(**kwargs) self.units = units self.components = components self.activation = activations.get(activation) self.use_bias = use_bias self.pre_kernel_initializer = initializers.get(pre_kernel_initializer) self.post_kernel_initializer = initializers.get(post_kernel_initializer) self.bias_initializer = initializers.get(bias_initializer) self.pre_kernel_regularizer = regularizers.get(pre_kernel_regularizer) self.post_kernel_regularizer = regularizers.get(post_kernel_regularizer) self.bias_regularizer = regularizers.get(bias_regularizer) self.activity_regularizer = regularizers.get(activity_regularizer) self.pre_kernel_constraint = constraints.get(pre_kernel_constraint) self.post_kernel_constraint = constraints.get(post_kernel_constraint) self.bias_constraint = constraints.get(bias_constraint) self.input_spec = InputSpec(min_ndim=2) self.supports_masking = True def build(self, input_shape): assert len(input_shape) >= 2 input_dim = input_shape[-1] is_factorized = self.components is not None if is_factorized: shape = (input_dim, self.components) else: shape = (input_dim, self.units) self.pre_kernel = self.add_weight(shape, initializer=self.pre_kernel_initializer, name='pre_kernel', regularizer=self.pre_kernel_regularizer, constraint=self.pre_kernel_constraint) if not is_factorized: self.post_kernel = None else: self.post_kernel = self.add_weight((self.components, self.units), initializer=self.post_kernel_initializer, name='kernel', regularizer=self.post_kernel_regularizer, constraint=self.post_kernel_constraint) if self.use_bias: self.bias = self.add_weight((self.units,), initializer=self.bias_initializer, name='bias', regularizer=self.bias_regularizer, constraint=self.bias_constraint) else: self.bias = None self.input_spec = InputSpec(min_ndim=2, axes={-1: input_dim}) self.built = True def call(self, inputs): h = K.dot(inputs, self.pre_kernel) if self.post_kernel is not None: h = K.dot(h, self.post_kernel) if self.use_bias: h = K.bias_add(h, self.bias) if self.activation is not None: h = self.activation(h) return h def compute_output_shape(self, input_shape): assert input_shape and len(input_shape) >= 2 assert input_shape[-1] output_shape = list(input_shape) output_shape[-1] = self.units return tuple(output_shape) def get_config(self): config = { 'units': self.units, 'activation': activations.serialize(self.activation), 'components': self.components, 'use_bias': self.use_bias, 'pre_kernel_initializer': initializers.serialize(self.pre_kernel_initializer), 'post_kernel_initializer': initializers.serialize(self.post_kernel_initializer), 'bias_initializer': initializers.serialize(self.bias_initializer), 'pre_kernel_regularizer': regularizers.serialize(self.pre_kernel_regularizer), 'post_kernel_regularizer': regularizers.serialize(self.post_kernel_regularizer), 'bias_regularizer': regularizers.serialize(self.bias_regularizer), 'activity_regularizer': regularizers.serialize(self.activity_regularizer), 'pre_kernel_constraint': constraints.serialize(self.pre_kernel_constraint), 'post_kernel_constraint': constraints.serialize(self.post_kernel_constraint), 'bias_constraint': constraints.serialize(self.bias_constraint) } base_config = super(FactorizedDense, self).get_config() return dict(list(base_config.items()) + list(config.items())) class FactorizedConv2DTucker(Layer): """2D convolution layer with tucker decomposition. This layer is based on `keras.layers.convolution.Conv2D` and behave like it. The difference is the kernel is factorized by tucker decomposition. If `use_bias` is True, a bias vector is created and added to the outputs. Finally, if `activation` is not `None`, it is applied to the outputs as well. When using this layer as the first layer in a model, provide the keyword argument `input_shape` (tuple of integers, does not include the sample axis), e.g. `input_shape=(128, 128, 3)` for 128x128 RGB pictures in `data_format="channels_last"`. # Arguments filters: Integer, the dimensionality of the output space (i.e. the number output of filters in the convolution). kernel_size: An integer or tuple/list of 2 integers, specifying the width and height of the 2D convolution window. Can be a single integer to specify the same value for all spatial dimensions. input_components: Integer or None, the number of components of kernel for the input channel axis. If given None, the factorization of input side is skipped. output_components: Integer or None, the number of components of kernel for the output channel axis. If given None, the factorization of output side is skipped. strides: An integer or tuple/list of 2 integers, specifying the strides of the convolution along the width and height. Can be a single integer to specify the same value for all spatial dimensions. Specifying any stride value != 1 is incompatible with specifying any `dilation_rate` value != 1. padding: one of `"valid"` or `"same"` (case-insensitive). data_format: A string, one of `channels_last` (default) or `channels_first`. The ordering of the dimensions in the inputs. `channels_last` corresponds to inputs with shape `(batch, width, height, channels)` while `channels_first` corresponds to inputs with shape `(batch, channels, width, height)`. It defaults to the `image_data_format` value found in your Keras config file at `~/.keras/keras.json`. If you never set it, then it will be "channels_last". dilation_rate: an integer or tuple/list of 2 integers, specifying the dilation rate to use for dilated convolution. Can be a single integer to specify the same value for all spatial dimensions. Currently, specifying any `dilation_rate` value != 1 is incompatible with specifying any stride value != 1. activation: Activation function to use (see [activations](../activations.md)). If you don't specify anything, no activation is applied (ie. "linear" activation: `a(x) = x`). use_bias: Boolean, whether the layer uses a bias vector. pre_kernel_initializer: Initializer for the `kernel` weights matrix (see [initializers](../initializers.md)). kernel_initializer: Initializer for the `kernel` weights matrix (see [initializers](../initializers.md)). post_kernel_initializer: Initializer for the `kernel` weights matrix (see [initializers](../initializers.md)). bias_initializer: Initializer for the bias vector (see [initializers](../initializers.md)). pre_kernel_regularizer: Regularizer function applied to the `kernel` weights matrix (see [regularizer](../regularizers.md)). kernel_regularizer: Regularizer function applied to the `kernel` weights matrix (see [regularizer](../regularizers.md)). post_kernel_regularizer: Regularizer function applied to the `kernel` weights matrix (see [regularizer](../regularizers.md)). bias_regularizer: Regularizer function applied to the bias vector (see [regularizer](../regularizers.md)). activity_regularizer: Regularizer function applied to the output of the layer (its "activation"). (see [regularizer](../regularizers.md)). pre_kernel_constraint: Constraint function applied to the kernel matrix (see [constraints](../constraints.md)). kernel_constraint: Constraint function applied to the kernel matrix (see [constraints](../constraints.md)). post_kernel_constraint: Constraint function applied to the kernel matrix (see [constraints](../constraints.md)). bias_constraint: Constraint function applied to the bias vector (see [constraints](../constraints.md)). # Input shape 4D tensor with shape: `(samples, channels, rows, cols)` if data_format='channels_first' or 4D tensor with shape: `(samples, rows, cols, channels)` if data_format='channels_last'. # Output shape 4D tensor with shape: `(samples, filters, new_rows, new_cols)` if data_format='channels_first' or 4D tensor with shape: `(samples, new_rows, new_cols, filters)` if data_format='channels_last'. `rows` and `cols` values might have changed due to padding. """ def __init__(self, filters, kernel_size, input_components=None, output_components=None, strides=(1, 1), padding='valid', data_format=None, dilation_rate=(1, 1), activation=None, use_bias=True, pre_kernel_initializer='glorot_uniform', kernel_initializer='glorot_uniform', post_kernel_initializer='glorot_uniform', bias_initializer='zeros', pre_kernel_regularizer=None, kernel_regularizer=None, post_kernel_regularizer=None, bias_regularizer=None, activity_regularizer=None, pre_kernel_constraint=None, kernel_constraint=None, post_kernel_constraint=None, bias_constraint=None, **kwargs): super(FactorizedConv2DTucker, self).__init__(**kwargs) rank = 2 self.rank = rank self.input_components = input_components self.output_components = output_components self.filters = filters self.output_components = output_components self.kernel_size = conv_utils.normalize_tuple(kernel_size, rank, 'kernel_size') self.strides = conv_utils.normalize_tuple(strides, rank, 'strides') self.padding = conv_utils.normalize_padding(padding) self.data_format = K.normalize_data_format(data_format) self.dilation_rate = conv_utils.normalize_tuple(dilation_rate, rank, 'dilation_rate') self.activation = activations.get(activation) self.use_bias = use_bias self.pre_kernel_initializer = initializers.get(pre_kernel_initializer) self.kernel_initializer = initializers.get(kernel_initializer) self.post_kernel_initializer = initializers.get(post_kernel_initializer) self.bias_initializer = initializers.get(bias_initializer) self.pre_kernel_regularizer = regularizers.get(pre_kernel_regularizer) self.kernel_regularizer = regularizers.get(kernel_regularizer) self.post_kernel_regularizer = regularizers.get(post_kernel_regularizer) self.bias_regularizer = regularizers.get(bias_regularizer) self.activity_regularizer = regularizers.get(activity_regularizer) self.pre_kernel_constraint = constraints.get(pre_kernel_constraint) self.kernel_constraint = constraints.get(kernel_constraint) self.post_kernel_constraint = constraints.get(post_kernel_constraint) self.bias_constraint = constraints.get(bias_constraint) self.input_spec = InputSpec(ndim=rank + 2) # batch, H, W, C def build(self, input_shape): if self.data_format == 'channels_first': channel_axis = 1 else: channel_axis = -1 if input_shape[channel_axis] is None: raise ValueError('The channel dimension of the inputs ' 'should be defined. Found `None`.') input_dim = input_shape[channel_axis] if self.input_components is None: input_components = input_dim else: input_components = self.input_components if self.output_components is None: output_components = self.filters else: output_components = self.output_components kernel_shape = self.kernel_size + (input_components, output_components) if self.input_components is None: self.pre_kernel = None else: pre_kernel_shape = (1, 1) + (input_dim, self.input_components) self.pre_kernel = self.add_weight(pre_kernel_shape, initializer=self.pre_kernel_initializer, name='pre_kernel', regularizer=self.pre_kernel_regularizer, constraint=self.pre_kernel_constraint) self.kernel = self.add_weight(kernel_shape, initializer=self.kernel_initializer, name='kernel', regularizer=self.kernel_regularizer, constraint=self.kernel_constraint) if self.output_components is None: self.post_kernel = None else: post_kernel_shape = (1, 1) + (self.output_components, self.filters) self.post_kernel = self.add_weight(post_kernel_shape, initializer=self.post_kernel_initializer, name='post_kernel', regularizer=self.post_kernel_regularizer, constraint=self.post_kernel_constraint) if self.use_bias: self.bias = self.add_weight((self.filters,), initializer=self.bias_initializer, name='bias', regularizer=self.bias_regularizer, constraint=self.bias_constraint) else: self.bias = None # Set input spec. self.input_spec = InputSpec(ndim=self.rank + 2, axes={channel_axis: input_dim}) self.built = True def call(self, inputs): h = inputs if self.pre_kernel is not None: h = K.conv2d( h, self.pre_kernel, strides=(1, 1), padding='valid', data_format=self.data_format, dilation_rate=(1, 1), ) h = K.conv2d( h, self.kernel, strides=self.strides, padding=self.padding, data_format=self.data_format, dilation_rate=self.dilation_rate, ) if self.post_kernel is not None: h = K.conv2d( h, self.post_kernel, strides=(1, 1), padding='valid', data_format=self.data_format, dilation_rate=(1, 1), ) outputs = h if self.use_bias: outputs = K.bias_add( outputs, self.bias, data_format=self.data_format) if self.activation is not None: return self.activation(outputs) return outputs def compute_output_shape(self, input_shape): if self.data_format == 'channels_last': space = input_shape[1:-1] new_space = [] for i in range(len(space)): new_dim = conv_utils.conv_output_length( space[i], self.kernel_size[i], padding=self.padding, stride=self.strides[i], dilation=self.dilation_rate[i]) new_space.append(new_dim) return (input_shape[0],) + tuple(new_space) + (self.filters,) if self.data_format == 'channels_first': space = input_shape[2:] new_space = [] for i in range(len(space)): new_dim = conv_utils.conv_output_length( space[i], self.kernel_size[i], padding=self.padding, stride=self.strides[i], dilation=self.dilation_rate[i]) new_space.append(new_dim) return (input_shape[0], self.filters) + tuple(new_space) def get_config(self): config = { 'input_components': self.input_components, 'output_components': self.output_components, 'filters': self.filters, 'kernel_size': self.kernel_size, 'strides': self.strides, 'padding': self.padding, 'data_format': self.data_format, 'dilation_rate': self.dilation_rate, 'activation': activations.serialize(self.activation), 'use_bias': self.use_bias, 'pre_kernel_initializer': initializers.serialize(self.pre_kernel_initializer), 'kernel_initializer': initializers.serialize(self.kernel_initializer), 'post_kernel_initializer': initializers.serialize(self.post_kernel_initializer), 'bias_initializer': initializers.serialize(self.kernel_initializer), 'pre_kernel_regularizer': regularizers.serialize(self.pre_kernel_regularizer), 'kernel_regularizer': regularizers.serialize(self.kernel_regularizer), 'post_kernel_regularizer': regularizers.serialize(self.post_kernel_regularizer), 'bias_regularizer': regularizers.serialize(self.bias_regularizer), 'activity_regularizer': regularizers.serialize(self.activity_regularizer), 'pre_kernel_constraint': constraints.serialize(self.pre_kernel_constraint), 'kernel_constraint': constraints.serialize(self.kernel_constraint), 'post_kernel_constraint': constraints.serialize(self.post_kernel_constraint), 'bias_constraint': constraints.serialize(self.bias_constraint) } base_config = super(FactorizedConv2DTucker, self).get_config() return dict(list(base_config.items()) + list(config.items())) custom_layers = { 'FactorizedConv2DTucker': FactorizedConv2DTucker, 'FactorizedDense': FactorizedDense, }
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/factorizers/tucker.py
training/helpers/keras_compressor/keras_compressor/factorizers/tucker.py
import itertools import logging import math from queue import PriorityQueue from typing import Optional, Tuple import numpy as np from keras import backend as K from keras.layers import Conv2D from keras_compressor.factorizer import Factorizer from keras_compressor.layers import FactorizedConv2DTucker from keras_compressor.utils import convert_config from sklearn.utils.extmath import randomized_svd logger = logging.getLogger(__name__) __all__ = ['TuckerFactorizer'] class ProblemData: '''Parameter search problem data structure ''' def __init__(self, x_range: range, y_range: range): self.x_range = x_range self.y_range = y_range def __str__(self): return '<Problem x_range={} y_range={}>'.format( self.x_range, self.y_range, ) def __lt__(self, other: 'ProblemData'): return self.diag_length < other.diag_length def __eq__(self, other: 'ProblemData'): return self.x_range == other.x_range and self.y_range == other.y_range @property def diag_length(self): return math.sqrt(len(self.x_range) ** 2 + len(self.y_range) ** 2) class Tucker: '''Pure tucker decomposition functions ''' @classmethod def factorize(cls, W: np.ndarray, in_comps: Optional[int], out_comps: Optional[int]) \ -> Tuple[np.ndarray, Optional[np.ndarray], Optional[np.ndarray]]: """pure tucker decomposition :param W: W x H x I x O :param in_comps: N :param out_comps: M :return: C: W x H x N x M, U_in: I x N U_out: O x M """ if in_comps is None: U_in = None else: U_in, _, _ = randomized_svd(cls._flatten(W, 2), in_comps) U_in = U_in.astype(W.dtype) if out_comps is None: U_out = None else: U_out, _, _ = randomized_svd(cls._flatten(W, 3), out_comps) U_out = U_out.astype(W.dtype) C = W.copy() if U_in is not None: C = np.einsum('whio,in->whno', C, U_in) if U_out is not None: C = np.einsum('whno,om->whnm', C, U_out) C = C.astype(W.dtype) return C, U_in, U_out @staticmethod def _get_matrix(W: np.ndarray, i: int, axis: int) -> np.ndarray: '''util function for tucker decomposition :param W: :param i: :param axis: :return: ''' sli = [slice(None) for _ in range(W.ndim)] sli[axis] = i return W[sli] @classmethod def _flatten(cls, W: np.ndarray, axis: int) -> np.ndarray: '''util function for tucker decomposition :param W: :param axis: :return: ''' dim = 1 dims = [] for i, v in enumerate(W.shape): if i != axis: dim *= v dims.append(v) res = np.zeros((W.shape[axis], dim)) for i in range(W.shape[axis]): res[i] = cls._get_matrix(W, i, axis).ravel() return res class TuckerParamSearcher: '''tucker decomposition parameter searcher ''' def __init__(self, W: np.ndarray): width, height, in_dim, out_dim = W.shape self.width = width self.height = height self.in_dim = in_dim self.out_dim = out_dim self.W = W self.best_point = None self.best_param_num = width * height * in_dim * out_dim self.prob_queue = PriorityQueue() def add_problem(self, prob: ProblemData): param_num = self.calc_min_param_num_by(prob) self.prob_queue.put((param_num, prob)) def calc_min_param_num_by(self, prob: ProblemData): res = [] for in_comp, out_comp in itertools.product( [prob.x_range.start, prob.x_range.stop], [prob.y_range.start, prob.y_range.stop], ): res.append(self.calc_param_num(in_comp, out_comp)) return min(res) def calc_param_num(self, in_comp: int, out_comp: int): params = self.width * self.height * in_comp * out_comp if in_comp != self.in_dim: # compression in input channel params += self.in_dim * in_comp if out_comp != self.out_dim: # compression in output channel params += self.out_dim * out_comp return params def update_best_point_if_needed(self, in_comp, out_comp): current_param_num = self.calc_param_num(in_comp, out_comp) if current_param_num < self.best_param_num: self.best_point = (in_comp, out_comp) self.best_param_num = current_param_num logger.debug('update best_point={} best_param_num={}'.format( self.best_point, self.best_param_num, )) def factorize_in_acceptable_error(self, acceptable_error: float) \ -> Tuple[np.ndarray, Optional[np.ndarray], Optional[np.ndarray]]: """ :param W: W x H x I x O :param acceptable_error: :param initial_problems: initial problems :return: C: W x H x N x M U_in: I x N U_out: O x M """ # Search N and M whose the number of parameter is the smallest # under acceptable error. # This algorithm is based on divide and conquer algorithm and # based on the assumption that `error` monotonically decrease by increasing N or M. width, height, in_dim, out_dim = self.width, self.height, self.in_dim, self.out_dim self.add_problem(ProblemData(range(in_dim, in_dim), range(out_dim, out_dim))) if 2 <= in_dim and 2 <= out_dim: self.add_problem(ProblemData(range(1, in_dim - 1), range(1, out_dim - 1))) if 1 <= in_dim: self.add_problem(ProblemData(range(in_dim, in_dim), range(1, out_dim - 1))) if 1 <= out_dim: self.add_problem(ProblemData(range(1, in_dim - 1), range(out_dim, out_dim))) while not self.prob_queue.empty(): _, current_prob = self.prob_queue.get() # type: ProblemData if self.best_param_num < self.calc_min_param_num_by(current_prob): logger.debug('no more best_param_num:{} prob:{}'.format(self.best_param_num, current_prob)) break if len(current_prob.x_range) == 0 and len(current_prob.y_range) == 0: self.update_best_point_if_needed(current_prob.x_range.start, current_prob.y_range.start) continue logger.debug('current queue.size:{} prob:{}'.format( self.prob_queue.qsize(), current_prob, )) result = self._find_edge_point( acceptable_error, current_prob, ) # type: Optional[Tuple[int,int]] logger.debug('result={} prob={}'.format( result, current_prob )) if result is None: continue x, y = result # type: Tuple[int, int] self.update_best_point_if_needed(x, y) # divide current problem to sub-problems if len(current_prob.x_range) == 0 or len(current_prob.y_range) == 0: # X. # Y logger.debug('no sub-problem:{}'.format(current_prob)) continue if len(current_prob.x_range) == 1 and len(current_prob.y_range) == 1: # X# -> (| and _) or . # Y if x == current_prob.x_range.stop and y == current_prob.y_range.stop: # right top point # _ # X# -> X and | # Y Y Y sub_prob1 = ProblemData( x_range=range(current_prob.x_range.start, x), y_range=range(y, y), ) sub_prob2 = ProblemData( x_range=range(x, x), y_range=range(current_prob.y_range.start, y), ) self.add_problem(sub_prob1) self.add_problem(sub_prob2) logger.debug('two sub-problems:{}, (x,y)=({},{}) -> {},{}'.format( current_prob, x, y, sub_prob1, sub_prob2 )) continue else: # x == current_prob.x_range.start and y == current_prob.y_range.start logger.debug('no sub-problem:{}'.format(current_prob)) continue if len(current_prob.x_range) == 1 and len(current_prob.y_range) > 1: # X####### -> X | # Y Y sub_prob = ProblemData( x_range=current_prob.x_range, y_range=range(y, y) ) logger.debug('one row space, one sub-problem:{}, (x,y)=({},{}) -> {}'.format( current_prob, x, y, sub_prob, )) continue if len(current_prob.x_range) > 1 and len(current_prob.y_range) == 1: # # _ # X# -> X # # Y # Y sub_prob = ProblemData( x_range=range(x, x), y_range=current_prob.y_range, ) logger.debug('one column space, one sub-problem:{}, (x,y)=({},{}) -> {}'.format( current_prob, x, y, sub_prob, )) if len(current_prob.x_range) >= 2 and len(current_prob.y_range) >= 2: # ### ## # X### -> X## and X # ### # # Y Y Y sub_prob1 = ProblemData( x_range=range(current_prob.x_range.start, x), y_range=range(y, current_prob.y_range.stop), ) sub_prob2 = ProblemData( x_range=range(x, current_prob.x_range.stop), y_range=range(current_prob.y_range.start, y), ) self.add_problem(sub_prob1) self.add_problem(sub_prob2) logger.debug('two sub-problems:{}, (x,y)=({},{}) -> {},{}'.format( current_prob, x, y, sub_prob1, sub_prob2 )) continue if self.best_point is None: logger.debug('no factorization is best') return self.W, None, None in_comp, out_comp = self.best_point if in_comp >= self.in_dim: in_comp = None if out_comp >= self.out_dim: out_comp = None C, U_in, U_out = Tucker.factorize(self.W, in_comp, out_comp) return C, U_in, U_out def _find_edge_point(self, acceptable_error: float, current_prob: ProblemData) -> Optional[Tuple[int, int]]: x_range = current_prob.x_range y_range = current_prob.y_range acceptable_points = [] # consider that acceptable point doesn't exist in the current_prob space. while len(x_range) > 0 or len(y_range) > 0: if len(x_range) in [0, 1]: x = x_range.start else: x = round((x_range.start + x_range.stop) / 2) if len(y_range) in [0, 1]: y = y_range.start else: y = round((y_range.start + y_range.stop) / 2) logger.debug('binary search (x,y)=({}, {}) x_range={} y_range={} prob={}'.format( x, y, x_range, y_range, current_prob )) C, U_in, U_out = Tucker.factorize(self.W, x, y) error = self.calc_error(self.W, C, U_in, U_out) if error < acceptable_error: logger.debug('binary search: under threshold={} error={}'.format( acceptable_error, error, )) acceptable_points.append((x, y)) # update ranges x_range = range(x_range.start, x) y_range = range(y_range.start, y) else: logger.debug('binary search: over threshold={} error={}'.format( acceptable_error, error, )) # update ranges if x + 1 <= x_range.stop: new_x_start = x + 1 else: new_x_start = x x_range = range(new_x_start, x_range.stop) if y + 1 <= y_range.stop: new_y_start = y + 1 else: new_y_start = y y_range = range(new_y_start, y_range.stop) if len(acceptable_points) == 0: return None else: return acceptable_points[-1] @staticmethod def calc_error(W: np.ndarray, C: np.ndarray, U_in: np.ndarray, U_out: np.ndarray) -> float: """calculate expected bound of error of output of layer :param W: W x H x I x O :param C: W x H x N x M :param U_in: I x N :param U_out: O x M :return: """ W_hat = np.einsum('whnm,in,om->whio', C, U_in, U_out) elemental_error = np.abs(W - W_hat) error_bound = np.mean(elemental_error) / np.mean(np.abs(W)) return error_bound class TuckerFactorizer(Factorizer): factorize_target_layers = [Conv2D] @classmethod def compress(cls, old_layer: Conv2D, acceptable_error: float) -> Optional[FactorizedConv2DTucker]: '''Compress layer's kernel 4D tensor using tucker decomposition under acceptable_error. If it can't reduce the number of parameters, returns `None`. :param old_layer: :param acceptable_error: :return: ''' W = K.get_value(old_layer.kernel) searcher = TuckerParamSearcher(W) C, U_in, U_out = searcher.factorize_in_acceptable_error(acceptable_error) kernel = C if U_in is None and U_out is None: # compression failed return None if U_in is None: input_components = None pre_kernel = None else: input_components = U_in.shape[1] pre_kernel = U_in[np.newaxis, np.newaxis, :, :] if U_out is None: output_components = None post_kernel = None else: output_components = U_out.shape[1] post_kernel = U_out.T[np.newaxis, np.newaxis, :, :] base_config = old_layer.get_config() new_config = convert_config( base_config, ignore_args=[ 'kernel_constraint', ], converts={ 'kernel_regularizer': [ 'pre_kernel_regularizer', 'kernel_regularizer', 'post_kernel_regularizer', ], 'kernel_initializer': [ 'pre_kernel_initializer', 'kernel_initializer', 'post_kernel_initializer', ], }, new_kwargs={ 'input_components': input_components, 'output_components': output_components, } ) new_layer = FactorizedConv2DTucker(**new_config) new_layer.build(old_layer.get_input_shape_at(0)) # to initialize weight variables K.set_value(new_layer.kernel, kernel) if pre_kernel is not None: K.set_value(new_layer.pre_kernel, pre_kernel) if post_kernel is not None: K.set_value(new_layer.post_kernel, post_kernel) return new_layer
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/factorizers/svd.py
training/helpers/keras_compressor/keras_compressor/factorizers/svd.py
import logging import math from typing import Optional, Tuple import numpy as np from keras import backend as K from keras.engine import Layer from keras.layers import Dense from sklearn.utils.extmath import randomized_svd from ..factorizer import Factorizer from ..layers import FactorizedDense from ..utils import convert_config logger = logging.getLogger(__name__) class SVDFactorizer(Factorizer): factorize_target_layers = [Dense] @staticmethod def _factorize(W: np.ndarray, components: int): """ :param W: I x O :param components: K :return: U: I x K V: K x O """ u, s, v = randomized_svd(W, components) scale = np.diag(np.sqrt(s)) U, V = u.dot(scale).astype(W.dtype), scale.dot(v).astype(W.dtype) return U, V @staticmethod def _calc_error(W: np.ndarray, U: np.ndarray, V: np.ndarray): ''' :param W: I x O :param U: I x K :param V: K x O :return: ''' elemental_error = np.abs(W - U.dot(V)) error_bound = np.mean(elemental_error) / np.mean(np.abs(W)) return error_bound @classmethod def compress(cls, old_layer: Dense, acceptable_error: float) -> Optional[Layer]: '''compress old_layer under acceptable error using SVD. If it can't reduce the number of parameters, returns None, :param old_layer: :param acceptable_error: :return: ''' W = K.get_value(old_layer.kernel) logger.debug('factorization start W.shape:{}'.format(W.shape)) max_comps = math.floor(np.size(W) / sum(W.shape)) U, V = cls._factorize(W, max_comps) if cls._calc_error(W, U, V) >= acceptable_error: # Factorizer can't reduce the number of parameters in acceptable error by SVD. # So, this factorizer failed compression. return None U, V = cls._compress_in_acceptable_error( W, acceptable_error, start_param_range=range(1, max_comps), ) components = U.shape[-1] base_config = old_layer.get_config() new_config = convert_config( base_config, ignore_args=[ 'kernel_constraint', ], converts={ 'kernel_regularizer': [ 'pre_kernel_regularizer', 'post_kernel_regularizer', ], 'kernel_initializer': [ 'pre_kernel_initializer', 'post_kernel_initializer', ] }, new_kwargs={ 'components': components, }, ) new_layer = FactorizedDense(**new_config) new_layer.build(old_layer.get_input_shape_at(0)) # to initialize weight variables K.set_value(new_layer.pre_kernel, U) K.set_value(new_layer.post_kernel, V) return new_layer @classmethod def _compress_in_acceptable_error(cls, W, acceptable_error: float, start_param_range: range) \ -> Tuple[np.ndarray, np.ndarray]: param_range = start_param_range while len(param_range) > 0: # while not (param_range.start == param_range.stop) logger.debug('current param_range:{}'.format(param_range)) if len(param_range) == 1: ncomp = param_range.start else: ncomp = round((param_range.start + param_range.stop) / 2) U, V = cls._factorize(W, ncomp) error = cls._calc_error(W, U, V) if error <= acceptable_error: # smallest ncomp is equal to or smaller than ncomp # On the assumption that `error` monotonically decreasing by increasing ncomp logger.debug('under acceptable error ncomp:{} threshold:{} error:{}'.format( ncomp, acceptable_error, error, )) param_range = range(param_range.start, ncomp) else: # the best is larger than ncomp logger.debug('over acceptable error ncomp:{} threshold:{} error:{}'.format( ncomp, acceptable_error, error, )) param_range = range(ncomp + 1, param_range.stop) # param_range.start == param_range.stop smallest_ncomp = param_range.start logger.debug('smallest_ncomp:{}, W.shape:{} compress_rate:{}'.format( smallest_ncomp, W.shape, sum(W.shape) * smallest_ncomp / np.size(W), )) U, V = cls._factorize(W, smallest_ncomp) return U, V
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/training/helpers/keras_compressor/keras_compressor/factorizers/__init__.py
training/helpers/keras_compressor/keras_compressor/factorizers/__init__.py
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/create_csv.py
annotation/create_csv.py
''' annotate.py Annotate audio, text, image, or video files for use with regression modeling in Allie. All you need is a folder, which identifies the type of file within it, and then it goes through each file to annotate (as .JSON) ''' import os, sys, datetime, json, time import pandas as pd from optparse import OptionParser def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ def most_common(lst): ''' get most common item in a list ''' return max(set(lst), key=lst.count) # get all the options from the terminal parser = OptionParser() parser.add_option("-d", "--directory", dest="annotate_dir", help="the directory to annotate", metavar="annotate_DIR") parser.add_option("-s", "--sampletype", dest="sampletype", help="specify the type of model to make predictions (e.g. audio, text, image, video, csv)", metavar="SAMPLETYPE") parser.add_option("-c", "--classtype", dest="classtype", help="specify the class type (e.g. stress level", metavar="CLASSTYPE") parser.add_option("-p", "--problemtype", dest="problemtype", help="specify the problem type (-c classification or -r regression", metavar="PROBLEMTYPE") (options, args) = parser.parse_args() curdir=os.getcwd() prevdir=prev_dir(os.getcwd())+'/features/' sys.path.append(prevdir) from standard_array import make_features # get annotate directory and sampletype class_=options.classtype problemtype=options.problemtype annotate_dir=options.annotate_dir os.chdir(annotate_dir) sampletype=options.sampletype listdir=os.listdir() if sampletype == None: sampletype = classifyfolder(listdir) listdir=os.listdir() data=dict() classes=list() filepaths=list() for i in range(len(listdir)): if listdir[i].endswith('.json'): g=json.load(open(listdir[i])) labels=g['labels'] for j in range(len(labels)): try: if list(labels[j]) == [class_] and labels[j][class_]['problemtype'] == problemtype: value=labels[j][class_]['value'] filepath=labels[j][class_]['annotate_dir']+labels[j][class_]['file'] print(value) print(filepath) classes.append(value) filepaths.append(filepath) except: pass data[class_]=classes data['paths']=filepaths os.chdir(curdir) df = pd.DataFrame(data) df.to_csv('%s_data.csv'%(class_), index=False)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/annotate.py
annotation/annotate.py
''' AAA lllllll lllllll iiii A:::A l:::::l l:::::l i::::i A:::::A l:::::l l:::::l iiii A:::::::A l:::::l l:::::l A:::::::::A l::::l l::::l iiiiiii eeeeeeeeeeee A:::::A:::::A l::::l l::::l i:::::i ee::::::::::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::eeeee:::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::e e:::::e A:::::A A:::::A l::::l l::::l i::::i e:::::::eeeee::::::e A:::::AAAAAAAAA:::::A l::::l l::::l i::::i e:::::::::::::::::e A:::::::::::::::::::::A l::::l l::::l i::::i e::::::eeeeeeeeeee A:::::AAAAAAAAAAAAA:::::A l::::l l::::l i::::i e:::::::e A:::::A A:::::A l::::::ll::::::li::::::ie::::::::e A:::::A A:::::A l::::::ll::::::li::::::i e::::::::eeeeeeee A:::::A A:::::A l::::::ll::::::li::::::i ee:::::::::::::e AAAAAAA AAAAAAAlllllllllllllllliiiiiiii eeeeeeeeeeeeee ___ _ _ _ ___ ______ _____ / _ \ | | | | (_) / _ \ | ___ \_ _| / /_\ \_ __ _ __ ___ | |_ __ _| |_ _ ___ _ __ / /_\ \| |_/ / | | | _ | '_ \| '_ \ / _ \| __/ _` | __| |/ _ \| '_ \ | _ || __/ | | | | | | | | | | | | (_) | || (_| | |_| | (_) | | | | | | | || | _| |_ \_| |_/_| |_|_| |_|\___/ \__\__,_|\__|_|\___/|_| |_| \_| |_/\_| \___/ Annotate audio, text, image, or video files for use with regression modeling in Allie. All you need is a folder, which identifies the type of file within it, and then it goes through each file to annotate (as .JSON) Read more about file annotation @ https://github.com/jim-schwoebel/allie/tree/master/annotation ''' import os, sys, datetime, json, time from optparse import OptionParser from tqdm import tqdm def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ def most_common(lst): ''' get most common item in a list ''' return max(set(lst), key=lst.count) def classifyfolder(listdir): filetypes=list() for i in range(len(listdir)): if listdir[i].endswith(('.mp3', '.wav')): filetypes.append('audio') elif listdir[i].endswith(('.png', '.jpg')): filetypes.append('image') elif listdir[i].endswith(('.txt')): filetypes.append('text') elif listdir[i].endswith(('.mp4', '.avi')): filetypes.append('video') elif listdir[i].endswith(('.csv')): filetypes.append('csv') filetype=most_common(filetypes) return filetype def annotate_file(class_, filetype, file, problemtype): # now go through and annotate each file if filetype == 'audio': print('playing file... %s'%(file.upper())) os.system('play "%s"'%(file)) elif filetype == 'image': print('opening file... %s'%(file.upper())) os.system('open "%s"'%(file)) elif filetype == 'video': print('playing file... %s'%(file.upper())) os.system('open "%s"'%(file)) elif filetype == 'text': print('opening file... %s'%(file.upper())) os.system('open "%s"'%(file)) else: print('file type not supported for annotation') annotation='0' if problemtype in ['r', 'regression']: annotation = input('%s value?\n'%(class_.upper())) else: annotation = input('%s label 1 (yes) or 0 (no)?\n'%(class_.upper())) # only get a float back try: annotation=float(annotation) except: print('error annotating, annotating again...') while True: annotation = annotate_file(class_, filetype, file) try: annotation=float(annotation) break except: pass label=dict() label[class_]={'value': annotation, 'datetime': str(datetime.datetime.now()), 'filetype': filetype, 'file': file, 'problemtype': problemtype, 'annotate_dir': annotate_dir} annotation=[label] print(annotation) return annotation # get all the options from the terminal parser = OptionParser() parser.add_option("-d", "--directory", dest="annotate_dir", help="the directory to annotate", metavar="annotate_dir") parser.add_option("-s", "--sampletype", dest="sampletype", help="specify the type of model to make predictions (e.g. audio, text, image, video, csv)", metavar="sampletype") parser.add_option("-c", "--classtype", dest="classtype", help="specify the class type (e.g. stress level", metavar="classtype") parser.add_option("-p", "--problemtype", dest="problemtype", help="specify the problem type (-c classification or -r regression", metavar="problemtype") (options, args) = parser.parse_args() prevdir=prev_dir(os.getcwd())+'/features/' sys.path.append(prevdir) from standard_array import make_features # get annotate directory and sampletype class_=options.classtype problemtype=options.problemtype annotate_dir=options.annotate_dir os.chdir(annotate_dir) sampletype=options.sampletype listdir=os.listdir() if sampletype == None: sampletype = classifyfolder(listdir) listdir=os.listdir() for i in tqdm(range(len(listdir))): try: if listdir[i].endswith('.json'): pass else: jsonfilename=listdir[i][0:-4]+'.json' if jsonfilename not in listdir: annotation=annotate_file(class_, sampletype, listdir[i], problemtype) basearray=make_features(sampletype) basearray['labels']=annotation jsonfile=open(jsonfilename,'w') json.dump(basearray, jsonfile) jsonfile.close() elif jsonfilename in listdir: g=json.load(open(jsonfilename)) labels=g['labels'] classin=False for j in range(len(labels)): try: print(list(labels[j])) print(labels[j][class_]['problemtype']) if list(labels[j]) == [class_] and labels[j][class_]['problemtype'] == problemtype: classin=True except: pass if classin == True: print('skipping %s, already annotated'%(listdir[i])) else: annotation=annotate_file(class_, sampletype, listdir[i], problemtype) labels.append(annotation) g['labels']=labels jsonfile=open(jsonfilename,'w') json.dump(g, jsonfile) jsonfile.close() except: print('error - file %s not recognized'%(listdir[i]))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/annotate_audio.py
annotation/helpers/annotate_audio.py
import os, time, shutil from tqdm import tqdm listdir=os.listdir() try: os.mkdir('coconut') except: pass try: os.mkdir('other') except: pass wavfiles=list() for i in range(len(listdir)): if listdir[i].endswith('.wav'): wavfiles.append(listdir[i]) for i in tqdm(range(len(wavfiles))): wavfile=wavfiles[i] os.system('play %s \n'%(wavfile)) yesorno=input('coconut? -y or -n \n') if yesorno == 'y': shutil.move(os.getcwd()+'/'+wavfile, os.getcwd()+'/coconut/'+wavfile) else: shutil.move(os.getcwd()+'/'+wavfile, os.getcwd()+'/other/'+wavfile)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/balancedelete.py
annotation/helpers/helpers/balancedelete.py
import os, random, shutil ## helper functions def get_wav(): # get all .WAV or .MP3 files in folder and count the number of them listdir=os.listdir() count=0 for i in range(len(listdir)): if listdir[i][-4:] in ['.wav', '.mp3']: count = count+1 return count def random_remove(remove_num): # remove a number of files to balnace classes. listdir=os.listdir() wavfiles=list() random.shuffle(listdir) for i in range(len(listdir)): if listdir[i][-4:] in ['.wav', '.mp3']: wavfiles.append(listdir[i]) for i in range(remove_num): os.remove(wavfiles[i]) print('removed %s .wav or .mp3 files'%(remove_num)) # now go to main script listdir=os.listdir() # find all folders folders=list() for i in range(len(listdir)): if listdir[i].find('.') < 0: folders.append(listdir[i]) curdir=os.getcwd() counts=list() for i in range(len(folders)): os.chdir(curdir) os.chdir(folders[i]) count=get_wav() counts.append(count) # now find minimum min_=min(counts) for i in range(len(folders)): os.chdir(curdir) os.chdir(folders[i]) count=get_wav() if count > min_: remove_num=count-min_ random_remove(remove_num)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/data-preprocess.py
annotation/helpers/helpers/data-preprocess.py
''' Data pre-processing for keras and sklearn (good tutorial here) ''' import numpy as np from numpy import argmax, array from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import OneHotEncoder from keras.utils import to_categorical # define example data = ['cold', 'cold', 'warm', 'cold', 'hot', 'hot', 'warm', 'cold', 'warm', 'hot'] # one hot encoding via sklearn library values = np.array(data) print(values) # integer encode label_encoder = LabelEncoder() integer_encoded = label_encoder.fit_transform(values) print(integer_encoded) # binary encode onehot_encoder = OneHotEncoder(sparse=False) integer_encoded = integer_encoded.reshape(len(integer_encoded), 1) onehot_encoded = onehot_encoder.fit_transform(integer_encoded) print(onehot_encoded) # invert first example inverted = label_encoder.inverse_transform([argmax(onehot_encoded[0, :])]) print(inverted) # one hot encoding via keras library data = [1, 3, 2, 0, 3, 2, 2, 1, 0, 1] data = array(data) print(data) # one hot encode encoded = to_categorical(data) print(encoded) # invert encoding inverted = argmax(encoded[0]) print(inverted) # to get back original list inverted=list() for i in range(len(encoded)): invert=argmax(encoded[i]) inverted.append(invert) inverted=array(inverted) print(inverted)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/musicgenre_download.py
annotation/helpers/helpers/musicgenre_download.py
import os import json import pafy import json import time import wave import random import ffmpy import soundfile as sf import getpass os.chdir("/Users/"+getpass.getuser()+"/Desktop/genres") optionlist=list() one="'Feel it Still' by Portugal. The Man. https://www.youtube.com/watch?v=pBkHHoOIIn8" #tone=duration of track one in seconds two="'Feels like Summer' by Wheezer. https://www.youtube.com/watch?v=efPWrIvzGgc" #ttwo=duration of track two in seconds three="'The Man' by the Killers. https://www.youtube.com/watch?v=w3xcybdis1k" four="'Thunder' by Imagine Dragons. https://www.youtube.com/watch?v=fKopy74weus" five="'Wish I knew You' by the Rivavlists. https://www.youtube.com/watch?v=o0Pt7M0weUI" six="'Believer' by Imagine Dragons. https://www.youtube.com/watch?v=7wtfhZwyrcc" seven="'Suit and Jacket' by Judah & the Lion. https://www.youtube.com/watch?v=AigOUsOEhSY" eight="'High' by Sir Sly. https://www.youtube.com/watch?v=qIOaU7Sm-ZE" nine="'Dig Down' by Muse. https://www.youtube.com/watch?v=b4ozdiGys5g" ten="'Lay It On Me' by Vance Joy. https://www.youtube.com/watch?v=VXXD1Qxpisw" eleven="'Whole Wide World' by Cage The Elephant. https://www.youtube.com/watch?v=cYGakznEllM" twelve="'Everything Now' by Arcade Fire. https://www.youtube.com/watch?v=zC30BYR3CUk" thirteen="'Ahead of Myself' by X Ambassadors. https://www.youtube.com/watch?v=z-Jsc8TmcAU" fourteen="'The Way You Used To Do' by Queens of the Stnoe Age. https://www.youtube.com/watch?v=GvyNyFXHj4k" fifteen="'Angela' by the Lumineers. https://www.youtube.com/watch?v=_II0fc7hgNY" sixteen="'Walk On Water' by Thirty Seconds to Mars. https://www.youtube.com/watch?v=FA2w-PMKspo" seventeen="'Perfect Places' by Lorde. https://www.youtube.com/watch?v=J0DjcsK_-HY" eighteen="'Champion' by Fall Out Boy. https://www.youtube.com/watch?v=JJJpRl2cTJc" nineteen="'Vacation' by Thr Dirty Heads. https://www.youtube.com/watch?v=7zok9co_8E4" twenty="'Lights Out' by Royal Blood. https://www.youtube.com/watch?v=ZSznpyG9CHY" twentyone="'Revolution Radio' by Green Day. https://www.youtube.com/watch?v=B4zc-f0TIZ4" twentytwo="'Less Than' by Nine Inch Nails. https://www.youtube.com/watch?v=gDV-dOvqKzQ" twentythree="'So Close' by Andrew McMahon in the Wilderness. https://www.youtube.com/watch?v=e5ZUfzJoG1E" twentyfour="'The Violence' by Rise Against. https://www.youtube.com/watch?v=Y7bvMlfmTm4" twentyfive="'The Sky Is a Neighborhood' by Foo Fighters. https://www.youtube.com/watch?v=TRqiFPpw2fY" twentysix="'Golden Dandelions' by Barns Courtney. https://www.youtube.com/watch?v=u8ymzuwrS6M" twentyseven="'Its a Trip!' by Joywave. https://www.youtube.com/watch?v=0xfXcZSb8LE" twentyeight="'Run' by Foo Fighters. https://www.youtube.com/watch?v=ifwc5xgI3QM" twentynine="'Little One' by Highly Suspect. https://www.youtube.com/watch?v=eKcIedFBiVU" thirty="'The Wanting' by J Roddy Walston & The Business. https://www.youtube.com/watch?v=S6VhNqKaKj0" option1=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option1)): optionlist.append(option1[i]) one="Edvard Grieg – Peer Gynt Suite No. 1, Op. 46: Morning Mood. https://www.youtube.com/watch?v=kzTQ9fjforY" two="Ludwig van Beethoven – Symphony No. 5 In C Minor, Op. 67, Fate: I. Allegro Con Brio. https://www.youtube.com/watch?v=tV6CpVfU7ig" three="Antonio Vivaldi – The Four Seasons, Op. 8, Spring: Allegro. https://www.youtube.com/watch?v=ygpf6mxTUeY" four="Samuel Barber – Adagio for Strings. https://www.youtube.com/watch?v=izQsgE0L450" five="Richard Wagner – The Valkyrie: Ride of the Valkyries. https://www.youtube.com/watch?v=GGU1P6lBW6Q" six="Frédéric Chopin – Nocturne No. 2 In E-Flat Major, Op. 9. https://www.youtube.com/watch?v=9E6b3swbnWg" seven="Johann Pachelbel – Canon in D major. https://www.youtube.com/watch?v=NlprozGcs80" eight="Carl Orff – Carmina Burana: O Fortuna. https://www.youtube.com/watch?v=GXFSK0ogeg4" nine="Johann Sebastian Bach – Orchestral Suite No. 3 in D major, BWV 1068: Air. https://www.youtube.com/watch?v=GkWjO8ZJcpc" ten="Gustav Holst – The Planets, Op. 32: Jupiter, the Bringer of Jollity. https://www.youtube.com/watch?v=Nz0b4STz1lo" eleven="Claude Debussy – Suite bergamasque, L 75: Clair de Lune. https://www.youtube.com/watch?v=vG-vmVrHOGE" twelve="Giuseppe Verdi – Nabucco: Chorus of the Hebrew Slaves (Va', Pensiero, Sull'ali Dorate). https://www.youtube.com/watch?v=-DIcS5-8RD8" thirteen="Wolfgang Amadeus Mozart – Piano Concerto No. 21 in C major, K. 467: II. Andante. https://www.youtube.com/watch?v=LA_BYwlzDVQ" fourteen="Johann Sebastian Bach – Brandenburg Concerto No. 3 in G major, BWV 1048: Allegro. https://www.youtube.com/watch?v=Xq2WTXtKurk" fifteen="Jules Massenet – Thaïs: Meditation. https://www.youtube.com/watch?v=luL1T1WQC2k" sixteen="Antonín Dvořák – Symphony No. 9 In E Minor, Op. 95, From the New World: II. Largo. https://www.youtube.com/watch?v=Aa-aD0SdxTY" seventeen="Johann Strauss II – On the Beautiful Blue Danube, Op. 314. https://www.youtube.com/watch?v=EHt2tW_nvp8" eighteen="Johannes Brahms – Hungarian Dance No. 5 In G Minor. https://www.youtube.com/watch?v=3X9LvC9WkkQ" nineteen="Pyotr Ilyich Tchaikovsky – Swan Lake Suite, Op. 20: Scene. https://www.youtube.com/watch?v=wEgOM9iYETg" twenty="Erik Satie – Gymnopédie No. 1. https://www.youtube.com/watch?v=S-Xm7s9eGxU" twentyone="Wolfgang Amadeus Mozart – Requiem, K. 626: Lacrimosa Dies Illa. https://www.youtube.com/watch?v=5IFtGSaFzbM" twentytwo="Ludwig van Beethoven – Bagatelle In A Minor, WoO 59, Für Elise. https://www.youtube.com/watch?v=NShTqdkCrXQ" twentythree="Edward Elgar – Pomp and Circumstance, Op. 39: Land of Hope and Glory. https://www.youtube.com/watch?v=0bknTe9nM8A" twentyfour="Georges Bizet – Carmen Suite No. 2: Habanera. https://www.youtube.com/watch?v=EcFJTc28soQ" twentyfive="Ludwig van Beethoven – Symphony No. 9 In D Minor, Op. 125, Choral: Ode an Die Freude" twentysix="Jacques Offenbach – The Tales of Hoffmann: Barcarolle. https://www.youtube.com/watch?v=g7czptgEvvU" twentyseven="Remo Giazotto – Adagio In G Minor for Strings and Organ (after T. Albinoni). https://www.youtube.com/watch?v=fzOLoIu4Gw8" twentyeight="Wolfgang Amadeus Mozart – Serenade No. 13 In G Major, K. 525, Eine Kleine Nachtmusik: I. Allegro. https://www.youtube.com/watch?v=z4Hfv00eqoI" twentynine="Gioachino Rossini – The Barber of Seville: Overture. https://www.youtube.com/watch?v=OloXRhesab0" thirty="W. A. Mozart - Die Zauberflöte (The Magic Flute), K. 620 - Overture. https://www.youtube.com/watch?v=_kgcu8yhojc" option2=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option2)): optionlist.append(option2[i]) one="'Body Like A Back Road' by Sam Hunt. https://www.youtube.com/watch?v=Mdh2p03cRfw" two="'Small Town Boy' by Dustin Lynch. https://www.youtube.com/watch?v=pz9yRC-LWhU" three="'What Ifs' by Kane Brown. https://www.youtube.com/watch?v=fM8V1XOI-14" four="'No Such Thing As A Broken Heart' by Old Dominion. https://www.youtube.com/watch?v=MZdpppFRLqw" five="'In Case You Didnt Know' by Brett Young. https://www.youtube.com/watch?v=7qaHdHpSjX8" six="'When It Rains It Pours' by Luke Combs. https://www.youtube.com/watch?v=uXyxFMbqKYA" seven="'Heartache On the Dance Floor' by Jon Pardi. https://www.youtube.com/watch?v=Cg4Eui4sGlk" eight="'Unforgettable' by Thomas Rhett. https://www.youtube.com/watch?v=84hwjCnAVjU" nine="'Drinkin' Problem' by Midland. https://www.youtube.com/watch?v=g7f6HiQ2LuU" ten="'All The Pretty Girls' by Kenny Chesney. https://www.youtube.com/watch?v=embINtOC_i0" eleven="'Light It Up' by Lukey Bryan. https://www.youtube.com/watch?v=8wpUv4I_fQk" twelve="'Craving You' by Thomas Rhett. https://www.youtube.com/watch?v=zruDce8zbEo" thirteen="'They Don't Know' by Jason Aldean. https://www.youtube.com/watch?v=SQZPIsGoR1M" fourteen="'Every Little Thing' by Carly Pearce. https://www.youtube.com/watch?v=pm6DHCpmIWg" fifteen="'More Girls Like You' by Kip Moore. https://www.youtube.com/watch?v=6-UHrVz1pR8" sixteen="'It Ain't my Fault' by Brothers Osborne. https://www.youtube.com/watch?v=MyOGVk_ypnM" seventeen="'For Her' by Chris Lane. https://www.youtube.com/watch?v=tyFY2I2Qlyk" eighteen="'Greatest Love Story' by LANCO. https://www.youtube.com/watch?v=aHl0tlUYDBI" nineteen="'I Could Use A Love Song' by Maren Morris. https://www.youtube.com/watch?v=ErdZ_W35xRs" twenty="'Fix A Drink' by Chris Janson. https://www.youtube.com/watch?v=-_Op0bQfMoo" twentyone="'Round Here Buzz' by Eric Church. https://www.youtube.com/watch?v=Z3Zj2RgNyj0" twentytwo="'Ring on Every Finger' by LOCASH. https://www.youtube.com/watch?v=LMDDdj_M0js" twentythree="'Ask Me How I Know' by Garth Brooks. https://www.youtube.com/watch?v=Dsxiw-d9Qmg" twentyfour="'Losing Sleep' by Chris Young. https://www.youtube.com/watch?v=NShYb80xnjw" twentyfive="'You Broke Up With Me' by Walker Hayes. https://www.youtube.com/watch?v=z6jfEH7D5bg" twentysix="'Yours' by Russell Dickerson. https://www.youtube.com/watch?v=gFccdvKehQI" twentyseven="'Smooth' by Florida Georgia Line. https://www.youtube.com/watch?v=QZD5TCFpNXg" twentyeight="'Last Time For Everything' by Brad Paisley. https://www.youtube.com/watch?v=LWkoquUvD98" twentynine="'Like I Loved You' by Brett Young. https://www.youtube.com/watch?v=PG2azZM4w4o" thirty="'All On Me' by Devin Dawson. https://www.youtube.com/watch?v=ntCMoh-0ogo" option3=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option3)): optionlist.append(option3[i]) one="'Feels' by Calvin Harris. https://www.youtube.com/watch?v=ozv4q2ov3Mk" two="'Something Just Like This' by the Chainsmokers & Coldplay. https://www.youtube.com/watch?v=FM7MFYoylVs" three="'Stay' by Zedd & Alessia Cara. https://www.youtube.com/watch?v=h--P8HzYZ74" four="'No Promises' by Cheat Codes. https://www.youtube.com/watch?v=jn40gqhxoSY" five="'It Ain't Me' by Kygo x Selena Gomez. https://www.youtube.com/watch?v=D5drYkLiLI8" six="'Swish Swish' by Katy Perry Featuring Nicki Minaj. https://www.youtube.com/watch?v=iGk5fR-t5AU" seven="'Silence' by Marshmello, featuring Khalid. https://www.youtube.com/watch?v=tk36ovCMsU8" eight="'Rockabye' by Clean Bandit featuring Sean Paul & Anne-Marie. https://www.youtube.com/watch?v=papuvlVeZg8" nine="'2U' by David Guetta featuring Justin Bieber. https://www.youtube.com/watch?v=RqcjBLMaWCg" ten="'Slide' by Calvin Harris. https://www.youtube.com/watch?v=8Ee4QjCEHHc" eleven="'Mama' by Jonas Blue. https://www.youtube.com/watch?v=qPTfXwPf_HM" twelve="'Know No Better' by Major Lazer. https://www.youtube.com/watch?v=Sgp0WDMH88g" thirteen="'Honest' by the Chainsmokers. https://www.youtube.com/watch?v=Lsv5IeI8bA8" fourteen="'Get Low' by Zedd & Liam Payne. https://www.youtube.com/watch?v=cSX0-MP6tjw" fifteen="'Rollin' by Calvin Harris. https://www.youtube.com/watch?v=5f_JiibvQAM" sixteen="'Symphony' by Clean Bandit. https://www.youtube.com/watch?v=aatr_2MstrI" seventeen="'Would You Ever' by Skrillex. https://www.youtube.com/watch?v=r-SurvChGFk" eighteen="'More Than You Know' by Axwell & Ingrosso. https://www.youtube.com/watch?v=GsF05B8TFWg" nineteen="'Lonely Together' by Avicii. https://www.youtube.com/watch?v=ruDrVMBCLaw" twenty="'Without You' by Avicii. https://www.youtube.com/watch?v=jUe8uoKdHao" twentyone="'Rich Love' by OneRepublic. https://www.youtube.com/watch?v=sJ6hAQjW9Aw" twentytwo="'Instruction' by Jax Jones. https://www.youtube.com/watch?v=MQXLpSl26q4" twentythree="'There For You' by Martin Garrix x Troye Sivan. https://www.youtube.com/watch?v=pNNMr5glICM" twentyfour="'All My Love' by Cash Cash. https://www.youtube.com/watch?v=HYrvDBgKAPo" twentyfive="'First Time' by Kygo & Ellie Goulding. https://www.youtube.com/watch?v=OlH1RCs96JA" twentysix="'OK' by Robin Schultz. https://www.youtube.com/watch?v=P9-4xHVc7uk" twentyseven="'Moving On' by Marshmello. https://www.youtube.com/watch?v=yU0tnrEk8H4" twentyeight="'Tired' by Alan Walker. https://www.youtube.com/watch?v=g4hGRvs6HHU" twentynine="'Find Me' by Marshmello. https://www.youtube.com/watch?v=ymq1WdGUcw8" thirty="'Pizza' by Martin Garrix. https://www.youtube.com/watch?v=JsKIAO11q1Y" option4=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option4)): optionlist.append(option4[i]) one="'Caminando' by David Benoit and Marc Antoine. https://www.youtube.com/watch?v=zo_RELAS7V0" two="'Frankie B' by Gerald Albright. https://www.youtube.com/watch?v=3GemcfAYuvw" three="'Girl Talk' by Cindy Bradley. https://www.youtube.com/watch?v=oHrm8dvPDWs" four="'Next To Me' by Lindsey Webster. https://www.youtube.com/watch?v=-zdMwQ2cWnY" five="'Let's Take A Ride' by Brian Culbertson. https://www.youtube.com/watch?v=dTmsemPB9TY" six="'Between You and I' by Riley Richard. https://www.youtube.com/watch?v=OBHHy2LJFZc" seven="'Uncle Nick' by Nick Colionne. https://www.youtube.com/watch?v=_ozECYHNaJk" eight="'Piccadilly Circus' by Paul Brown. https://www.youtube.com/watch?v=zJosV9Bw3DE" nine="'Trininty' by Jackiem Joyner. https://www.youtube.com/watch?v=EHG_43UczAo" ten="'Let It Go' by Jonathan Fritzen. https://www.youtube.com/watch?v=FJ9x-CZGq1k" eleven="'Down The Road' by Paul Jackson, Jr. https://www.youtube.com/watch?v=MnlFINIQpJI" twelve="'Vivid' by Blake Aron. https://www.youtube.com/watch?v=qgBJI8KEeC8" thirteen="'Happy Hour' by Chuck Loeb. https://www.youtube.com/watch?v=Zom2UbDm52Q" fourteen="'I Don't Mind' by Adam Hawley. https://www.youtube.com/watch?v=P4MlGsVh24c" fifteen="'Tick Tock' by Boney James. https://www.youtube.com/watch?v=UnC8bkwPo2w" sixteen="'Deixa' by Marc Antoine. https://www.youtube.com/watch?v=via8CmbNA8Q" seventeen="'Here to Stay' by Darryl Williams. https://www.youtube.com/watch?v=ejRX6KaG2e4" eighteen="'Now What' by Walter Beasley. https://www.youtube.com/watch?v=JyOBHB8NQTc" nineteen="'Road Trip' by Andre Cavor. https://www.youtube.com/watch?v=eU_-xKtutdI" twenty="'Going Out' by Julian Vaughn. https://www.youtube.com/watch?v=dBaIaZy1-I4" twentyone="'Baby Coffee' by Michael J. Thomas. https://www.youtube.com/watch?v=cM4Q0v01wQU" twentytwo="'Early Arrival' by Ragan Whiteside. https://www.youtube.com/watch?v=GSqHRblW9UI" twentythree="'Carmanology' by the Allen Carman Project. https://www.youtube.com/watch?v=3BVDhAN41DE" twentyfour="'Where I Left Off' by Oli Silk. https://www.youtube.com/watch?v=eC6ikAWRq9E" twentyfive="'Let's Take It Back' by Najee. https://www.youtube.com/watch?v=N_UqrRjs9r0" twentysix="'Lay Lady Lay' by Jack DeJohnette, Larry Grenadier, John Medeski, & John Scofield. https://www.youtube.com/watch?v=BjZC9G4p4Hw" twentyseven="'The Edge of Twilight' by Keiko Matsui. https://www.youtube.com/watch?v=csg-7f539CI&list=PL6-ukl_LI7_TM1msq2UW9XpbHYsyMImeq&index=15" twentyeight="'Water Lily' by Keiko Matsui. https://www.youtube.com/watch?v=g2ia7ogeEwE&list=PL6-ukl_LI7_TM1msq2UW9XpbHYsyMImeq&index=22" twentynine="'Angel of the South' by Acoustic Alchemy. https://www.youtube.com/watch?v=qKgRrqgAvho&index=2&list=PLvBIYgM9CkyQqvJJqVKRJdbu_w5VG6dS4" thirty="'Passion Play' by Acoustic Alchemy. https://www.youtube.com/watch?v=vkXpS8oBt2A&index=16&list=PLvBIYgM9CkyQqvJJqVKRJdbu_w5VG6dS4" option5=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option5)): optionlist.append(option5[i]) one="'Bodak Yellow (Money Moves)' by Cardi B. https://www.youtube.com/watch?v=PEGccV-NOm8" two="'Unforgettable' by French Montana. https://www.youtube.com/watch?v=CTFtOOh47oo" three="'1-800-273-8255' by Logic. https://www.youtube.com/watch?v=Kb24RrHIbFk" four="'Bank Account' by 21 Savage. https://www.youtube.com/watch?v=eCK772REqw0" five="'Rake It Up' by Yo. https://www.youtube.com/watch?v=OrSadmwmmAs" six="'Humble' by Kendrick Lamar. https://www.youtube.com/watch?v=tvTRZJ-4EyI" seven="'XO Tour Llif3' by Lil Uzi Vert. https://www.youtube.com/watch?v=WrsFXgQk5UI" eight="'Congratulations' by Post Malone. https://www.youtube.com/watch?v=SC4xMk98Pdc" nine="'I'm The One' by DJ Khaled. https://www.youtube.com/watch?v=weeI1G46q0o" ten="'Jocelyn Flores' by XXXTENTACION. https://www.youtube.com/watch?v=C1D3G2VGQ_8" eleven="'Loyalty' by Kendrick Lamar. https://www.youtube.com/watch?v=Dlh-dzB2U4Y" twelve="'The Way Life Goes' by Lil Uzi Vert. https://www.youtube.com/watch?v=oKu2FVy0oUo" thirteen="'I Get The Bag' by Gucci Mane. https://www.youtube.com/watch?v=uo14xGYwWd4" fourteen="'It's A Vibe' by 2 Chainz. https://www.youtube.com/watch?v=tU3p6mz-uxU" fifteen="'Sauce It Up' by Lil Uzi Vert. https://www.youtube.com/watch?v=BnRNXWAGENE" sixteen="'Crew' by Goldlink. https://www.youtube.com/watch?v=nhNqbe6QENY" seventeen="'Everyday We Lit' by YFN Lucci. https://www.youtube.com/watch?v=44Vk5KyQbiA" eighteen="'Drowning' by A Boogie Wit da Hoodie. https://www.youtube.com/watch?v=rvaJ7QlhH0g" nineteen="'Magnolia' by Playboi Carti. https://www.youtube.com/watch?v=oCveByMXd_0" twenty="'Everybody Dies in Their Nightmares' by XXXTENTACION. https://www.youtube.com/watch?v=Tg6HGcj7pGo" twentyone="'Transportin' by Kodk Black. https://www.youtube.com/watch?v=Ns167_llTiA" twentytwo="'Butterfly Effect' by Travis Scott. https://www.youtube.com/watch?v=_EyZUTDAH0U" twentythree="'444+222' by Lil Uzi Vert. https://www.youtube.com/watch?v=9qfUYeMfl2Q" twentyfour="'Roll In Peace' by Kodak Black. https://www.youtube.com/watch?v=hwCLVBCNIt0" twentyfive="'DNA' by Kendrick Lamar. https://www.youtube.com/watch?v=NLZRYQMLDW4" twentysix="'Versace On The Floor' by Bruno Mars. https://www.youtube.com/watch?v=-FyjEnoIgTM" twentyseven="'Revenge' by XXXTENTACION. https://www.youtube.com/watch?v=3CJsqGa1ltM" twentyeight="'B.E.D. by Jacquees. https://www.youtube.com/watch?v=ul1H_p_FeaA" twentynine="'For Real' by Lil Uzi Vert. https://www.youtube.com/watch?v=xnnfQ0inQp8" thirty="'Questions' by Chris Brown. https://www.youtube.com/watch?v=mH76VvWkNwA" option6=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option6)): optionlist.append(option6[i]) one="'All I Want For Christmas Is You' by Mariah Carey. https://www.youtube.com/watch?v=yXQViqx6GMY" two="'Hallelujah' by Pentatonix. https://www.youtube.com/watch?v=LRP8d7hhpoQ" three="'Rockin' Around the Christmas Tree.' by Brenda Lee. https://www.youtube.com/watch?v=_6xNuUEnh2g" four="'Jingle Bell Rock' by Bobby Helms. https://www.youtube.com/watch?v=itcMLwMEeMQ" five="'Feliz Navidad' by Jose Feliciano. https://www.youtube.com/watch?v=xMtuVP8Mj4o" six="'Mary, Did You Know?' by Pentatonix. https://www.youtube.com/watch?v=ifCWN5pJGIE" seven="'A Holly Jolly Christmas' by Burl Ives. https://www.youtube.com/watch?v=nVMCUtsmWmQ" eight="'The Christmas Song' by Nat King Cole. https://www.youtube.com/watch?v=hwacxSnc4tI" nine="'It's the Most Wonderful Time Of The Year' by Andy Williams. https://www.youtube.com/watch?v=gFtb3EtjEic" ten="'Last Christmas' by Wham! https://www.youtube.com/watch?v=4hhew1QKi-U" eleven="'Rudolph The Red-Nosed Reindeer' by Gene Autry. https://www.youtube.com/watch?v=7ara3-hDH6I" twelve="'White Christmas' by Bing Crosby. https://www.youtube.com/watch?v=GJSUT8Inl14" thirteen="'Christmas Eve' by Trans-Siberian Orchestra. https://www.youtube.com/watch?v=xJ_OnN5F4Yw" fourteen="'You're One Mean One', Mr. Grinch' https://www.youtube.com/watch?v=WxVqliZCNw0" fifteen="'Let It Snow, Let It Snow, Let It Snow' by Dean Martin. https://www.youtube.com/watch?v=mN7LW0Y00kE" sixteen="'Blue Chrismas' by Elvis Presley. https://www.youtube.com/watch?v=Uwfz5mMLSDM" seventeen="'Christmas Time Is Here' by Vince Guaraldi Trio. https://www.youtube.com/watch?v=YvI_FNrczzQ" eighteen="'It's Beginning To Look A Lot Like Christmas' by Michael Buble. https://www.youtube.com/watch?v=EyKMPXqKlFk" nineteen="'Wonderful Christmastime' by Paul McCartney. https://www.youtube.com/watch?v=V9BZDpni56Y" twenty="'Sleigh Ride' by Ronettes. https://www.youtube.com/watch?v=Y6rDA2Czz0E" twentyone="'Happy Xmas' by John Lennon & Yoko Ono. https://www.youtube.com/watch?v=z8Vfp48laS8" twentytwo="'Christmas Soon' by the Trans-Siberian Orhcestra. https://www.youtube.com/watch?v=4cP26ndrmtg" twentythree="'Santa Tell Me' by Ariana Grande. https://www.youtube.com/watch?v=nlR0MkrRklg" twentyfour="'Mistletoe' by Justin Bieber. https://www.youtube.com/watch?v=LUjn3RpkcKY" twentyfive="'Have Yourself A Merry Little Christmas' by Frank Sinatra. https://www.youtube.com/watch?v=nZ6yQgBvuoI" twentysix="'Merry Christmas Darling' by Carpenters. https://www.youtube.com/watch?v=YR1ujXx2p-I" twentyseven="'Linus & Lucy' and Vince Guaraldi Trio. https://www.youtube.com/watch?v=x6zypc_LhnM" twentyeight="'St. Brick Intro' by Guci Mane. https://www.youtube.com/watch?v=ywaX-veaB40" twentynine="'I Want A Hippopotamus For Christmas' by Gayla Peevey. https://www.youtube.com/watch?v=7oOzszFIBcE" thirty="'Little Saint Nick' by the Beach Boys. https://www.youtube.com/watch?v=aSynDh_K0EE" option7=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option7)): optionlist.append(option7[i]) one="'Feel It Still' by Portugal. The Man. https://www.youtube.com/watch?v=pBkHHoOIIn8" two="'Do I Wanna Know' by Artic Monkeys. https://www.youtube.com/watch?v=bpOSxM0rNPM" three="'Beezeblocks' by atl-J. https://www.youtube.com/watch?v=rVeMiVU77wo" four="'Sweater Weather' by The Neighbourhood. https://www.youtube.com/watch?v=GCdwKhTtNNw" five="'Take a Walk' by Passion Pit. https://www.youtube.com/watch?v=dZX6Q-Bj_xg" six="'Chocolate' by The 1975. https://www.youtube.com/watch?v=CHk5SWVO4p8" seven="'Angels' by The xx. https://www.youtube.com/watch?v=_nW5AF0m9Zw" eight="'My Number' by Foals. https://www.youtube.com/watch?v=bAsGFnLl2u0" nine="'Entertainment' by Phoenix. https://www.youtube.com/watch?v=qaMyr36uIv8" ten="'Falling' by HAIM. https://www.youtube.com/watch?v=AIjVpRAXK18" eleven="'Sun' by Two Door Cinema Club. https://www.youtube.com/watch?v=sKyK1Mme9Sc" twelve="'Feels Like We Only Go Backwards' by Tame Impala." thirteen="'Recovery' by Frank Turner. https://www.youtube.com/watch?v=F1L5zJ2afLs" fourteen="'Don't Save Me' by HAIM. https://www.youtube.com/watch?v=kiqIush2nTA" fifteen="'Global Concepts' by Robert DeLong. https://www.youtube.com/watch?v=JND-sxlg7YU" sixteen="'Miracle Mile' by Cold War Kids. https://www.youtube.com/watch?v=1F6gAN6MOII" seventeen="'Song For Zula' by Phosphorescent. https://www.youtube.com/watch?v=ZPxQYhGpdvg" eighteen="'Tessellate' by alt-J. https://www.youtube.com/watch?v=Qg6BwvDcANg" nineteen="'Oblivion' by M83. https://www.youtube.com/watch?v=822P87a773c" twenty="'Tap Out' by The Strokes. https://www.youtube.com/watch?v=-7PINAYE4z4" twentyone="'Matilda' by alt-J. https://www.youtube.com/watch?v=Q06wFUi5OM8" twentytwo="'Carried Away' by Passion Pit. https://www.youtube.com/watch?v=Evz03lIJ3f0" twentythree="'Heavy Feet' by Local Natives. https://www.youtube.com/watch?v=h2zWfxW60z0" twentyfour="'Trying To Be Cool' by Phoenix. https://www.youtube.com/watch?v=OePvsCfKHJg" twentyfive="'Trembling Hands' by The Temper Trap. https://www.youtube.com/watch?v=iW0uYfq3VLU" twentysix="'Kemosabe' by Everything Everything. https://www.youtube.com/watch?v=TKKMfJ8cZoQ" twentyseven="'The Way I Tend To Be' by Frank Turner. https://www.youtube.com/watch?v=Cf5O2M5GaEA" twentyeight="'Sunset' by The xx. https://www.youtube.com/watch?v=M2JrAhmZmpM" twentynine="'The Bay' by Metronomy. https://www.youtube.com/watch?v=9PnOG67flRA" thirty="'Time To Run' by Lord Huron. https://www.youtube.com/watch?v=5_e8RRTT0r8" option8=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option8)): optionlist.append(option8[i]) #toby mac one="'Real Love' by Blanca. https://www.youtube.com/watch?v=AKG_3u_kdJ8" two="'Old Church Choir' by Zach Williams. https://www.youtube.com/watch?v=yOEviTLJOqo" three="'O Come to the Alatar' by Elevation Worship. https://www.youtube.com/watch?v=ycWDFd0yCHA" four="'Love Broke Thru' by Toby Mac. https://www.youtube.com/watch?v=44l9PRI4c2M" five="'Full of Faith' by Cody Carnes. https://www.youtube.com/watch?v=IPPGvudBaiM" six="'Gracefully Broken' by Matt Redman. https://www.youtube.com/watch?v=__haUJns_b8" seven="'I'll Find You' by Lecrae. https://www.youtube.com/watch?v=Jv8IqJm6q7w" eight="'So Will I (100 Billion X)' Hillsong United. https://www.youtube.com/watch?v=EuYOnYL6G0Y" nine="'Broken Things' by Matthew West. https://www.youtube.com/watch?v=WdUu6ZsdVfM&list=RDWdUu6ZsdVfM" ten="'The Gospel' by Ryan Stevenson. https://www.youtube.com/watch?v=NTdFEZhjiko" eleven="'Hard Love' by NEEDTOBREATHE. https://www.youtube.com/watch?v=tE3Fp8C_ufg" twelve="'Hills and Valleys' by Tauren Wells. https://www.youtube.com/watch?v=p4rRCjrAyCs" thirteen="'Bleed the Same' by Mandisa. https://www.youtube.com/watch?v=UEzCQBwQkdA" fourteen="'The Answer' by Jeremy Camp. https://www.youtube.com/watch?v=rQHXJi1EhDM" fifteen="'A Million Lights' by Michael W. Smith. https://www.youtube.com/watch?v=DaTcRrINSXo" sixteen="'Death Was Arrested' by North Point InsideOut. https://www.youtube.com/watch?v=uMsMiluCUUI" seventeen="'Hold You Down' by Deraj. https://www.youtube.com/watch?v=ZXgiea7hW90" eighteen="'What A Beautiful Name' by Hillsong Worship. https://www.youtube.com/watch?v=nQWFzMvCfLE" nineteen="'Jesus & You' by Matthew West. https://www.youtube.com/watch?v=kJhVA8YW5yY" twenty="'Details' by Sarah Reeves. https://www.youtube.com/watch?v=gTICoQC8PRw" twentyone="'Bulletproof' by Citizen Way. https://www.youtube.com/watch?v=RzqpK7ZaH6o" twentytwo="'Rescuer (Good News)' by Rend Collective. https://www.youtube.com/watch?v=sAg7rn7fH3Q" twentythree="'You Belong' by Jasmine Murray. https://www.youtube.com/watch?v=4-GIcCOHoMw" twentyfour="'Even If' by MercyMe. https://www.youtube.com/watch?v=B6fA35Ved-Y" twentyfive="'Build My Life' by Christy Nockels. https://www.youtube.com/watch?v=QJCxe0cd15A" twentysix="'Mountain - Radio Version' by Bryan & Katie Torwalt. https://www.youtube.com/watch?v=IUAOF5LLXDc" twentyseven="'Control (Somehow You Want Me)' by Tenth Avenue North. https://www.youtube.com/watch?v=kFfztu8-bBQ" twentyeight="'Different' by Micah Tyler. https://www.youtube.com/watch?v=3MtfLap4qcc" twentynine="'I know' by Kim Walker Smith. https://www.youtube.com/watch?v=mAq-74wRSFQ" thirty="'Home' by Chris Tomlin. https://www.youtube.com/watch?v=BCiBQqfHSvQ" option9=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option9)): optionlist.append(option9[i]) one="'Bella y Sensual' by Romeo Santos. https://www.youtube.com/watch?v=ybzDgGCL1Xk" two="'Felices los 4' by Maluma. https://www.youtube.com/watch?v=n6WUcjyagN8" three="'Mayores' by Becky G'Krippy Kush' by Farruko. https://www.youtube.com/watch?v=GMFewiplIbw" four="'Krippy Kush' by Farruko. https://www.youtube.com/watch?v=j1_JW7An2l0" five="'Se Preparo' by Ozuna. https://www.youtube.com/watch?v=KWGrPNqz4uc" six="'Escapate Conmigo' by Wisin. https://www.youtube.com/watch?v=3X9wEwulYhk" seven="Una Lady Como Tu' by Mnaual Turizo. https://www.youtube.com/watch?v=T3pstB1gWyo" eight="'Robarte un Beso' by Carlos Vives. https://www.youtube.com/watch?v=Mtau4v6foHA" nine="'Mi Gente' by J. Balvin. https://www.youtube.com/watch?v=wnJ6LuUFpMo" ten="'Explicale' by Yandel. https://www.youtube.com/watch?v=U516oP9nt2o" eleven="'Ni Tu Ni Yo' by Jennifer Lopez. https://www.youtube.com/watch?v=V5_tnpdnNz4" twelve="Hey DJ' by CNCO. https://www.youtube.com/watch?v=X6wQOW9ihDA" thirteen="Doble Personalidad' by Noriel. https://www.youtube.com/watch?v=1qpKJJgBXmw" fourteen="'Perro Fiel' by Shakira. https://www.youtube.com/watch?v=o5U5ivOnJjs" fifteen="'Ganas Locas' by Prince Royce. https://www.youtube.com/watch?v=Ztf7QEetikY" sixteen="Imitadora' by Romeo Santos. https://www.youtube.com/watch?v=mhHqonzsuoA" seventeen="'Si Tu La Ves' by Nicky Jam. https://www.youtube.com/watch?v=mcGBVy3-W4s" eighteen="'Internacionales' by Bomba Estereo. https://www.youtube.com/watch?v=tWwWoDFoubw" nineteen="'Loco Enamorado' by Abrham Mateo. https://www.youtube.com/watch?v=cmIKUyUrKl4" twenty="'Unforgettable' by French Montana. https://www.youtube.com/watch?v=CTFtOOh47oo" twentyone="'Sastre de Tu Amor' by Orishas. https://www.youtube.com/watch?v=n_CLTHgFF4c" twentytwo="'Me Rehuso' by Danny Ocean. https://www.youtube.com/watch?v=aDCcLQto5BM" twentythree="'Hey Guapo' by Play-N-Skillz. https://www.youtube.com/watch?v=Vzs9JgtW_lI" twentyfour="'No Le Hablen de Amor' by CD9. https://www.youtube.com/watch?v=_ixn-FRppEk" twentyfive="'Criminal' by Natti Natasha. https://www.youtube.com/watch?v=VqEbCxg2bNI" twentysix="'Muevete' by MIX5. https://www.youtube.com/watch?v=KfaU_RY14HQ" twentyseven="'Que Me Has Hecho' by Chayanne. https://www.youtube.com/watch?v=q_OGqmx3DNQ" twentyeight="'Just As I am' by Spiff TV. https://www.youtube.com/watch?v=i1dVohtr4Uk" twentynine="'SUBEME LA RADIO' by Enrique Iglesias. https://www.youtube.com/watch?v=9sg-A-eS6Ig" thirty="'El Amante' by Nicky Jam. https://www.youtube.com/watch?v=YG2p6XBuSKA" option10=[one,two,three,four,five,six,seven,eight,nine,ten,eleven,twelve,thirteen,fourteen,fifteen,sixteen,seventeen,eighteen,nineteen,twenty,twentyone,twentytwo,twentythree,twentyfour,twentyfive,twentysix,twentyseven,twentyeight,twentynine,thirty] for i in range(len(option10)): optionlist.append(option10[i]) #phillip wesley #ludivico einvaldi one="'River Flows In You' by Yuruma. https://www.youtube.com/watch?v=XsTjI75uEUQ" two="'Comptine d'un autre ete, l'apres-midi' by Yann Tierson. https://www.youtube.com/watch?v=NvryolGa19A" three="'Watermark' by Enya. https://www.youtube.com/watch?v=NO5tb20qQnA" four="'Song For Sienna' by Brian Crain. https://www.youtube.com/watch?v=2MYXZi2q02A" five="'Nuvole bianche' by Ludovico Einaudi. https://www.youtube.com/watch?v=kcihcYEOeic" six="'One Man's Dream' by Yanni. https://www.youtube.com/watch?v=STSzCX36U6o" seven="'Sundial Dreams' by Kevin Kern. https://www.youtube.com/watch?v=ERGGPB_ok18" eight="'Chrisofori's Dream' by David Lanz. https://www.youtube.com/watch?v=9wxrB41PMhw" nine="'Near light' by Olafur Arnalds. https://www.youtube.com/watch?v=0kYc55bXJFI" ten="'Waterfall' by Jon Schmidt. https://www.youtube.com/watch?v=8P9hAN-teOU" eleven="'Opus 28' by Dustin O'Halloran. https://www.youtube.com/watch?v=iQTWbS2SlVY" twelve="'Angel Eyes' by Jim Brickman. https://www.youtube.com/watch?v=3jcN20Efpq0" thirteen="'The Approaching Night.' by Philip Wesley. https://www.youtube.com/watch?v=MsTQjB1f4-A" fourteen="'A Beautiful Distraction' by Michele McLaughlin. https://www.youtube.com/watch?v=erbuUytTB44" fifteen="'Breathe' by Greg Maroney. https://www.youtube.com/watch?v=9eqyir5JGpA" sixteen="'Winter Walk' by David Nevue. https://www.youtube.com/watch?v=g9J4GPURT0s" seventeen="'Rococo' by Brique a Braq. https://www.youtube.com/watch?v=XpZn_8Nx9w4" eighteen="'Love's River' by Laura Sullivan. https://www.youtube.com/watch?v=bzHgyYj2-8I" nineteen="'Simply Satie' by Michael Dulin. https://www.youtube.com/watch?v=xcArvm3yCOI" twenty="'Walk With Me' by Joe Bongiorno. https://www.youtube.com/watch?v=IY7btFAPl1M" twentyone="'Surrender' by Solomon Keal. https://www.youtube.com/watch?v=y2u0IE_adDM"
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
true
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/deletejsonfolders.py
annotation/helpers/helpers/deletejsonfolders.py
import os from tqdm import tqdm os.chdir('/Users/jimschwoebel/desktop/deletejson') hostdir=os.getcwd() listdir=os.listdir() folders=list() for i in range(len(listdir)): if listdir[i].find('.') < 0: folders.append(listdir[i]) for i in tqdm(range(len(folders))): os.chdir(folders[i]) listdir=os.listdir() for j in range(len(listdir)): if listdir[j][-5:]=='.json': os.remove(listdir[j]) os.chdir(hostdir)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/random20secsplice.py
annotation/helpers/helpers/random20secsplice.py
import soundfile as sf import os import ffmpy import random import getpass genre=input('what folder do you want to create 20 sec splices for?') dir1='/Users/'+getpass.getuser()+'/Desktop/genres/'+genre dir2='/Users/'+getpass.getuser()+'/Desktop/genres/'+genre+'_snipped' os.chdir(dir1) os.mkdir(dir2) listdir=os.listdir() for i in range(len(listdir)): try: os.chdir(dir1) file=listdir[i] data, samplerate = sf.read(file) totalframes=len(data) totalseconds=int(totalframes/samplerate) startsec=random.randint(0,totalseconds-21) endsec=startsec+20 startframe=samplerate*startsec endframe=samplerate*endsec #write file to resave wave file at those frames os.chdir(dir2) sf.write('snipped_'+file, data[int(startframe):int(endframe)], samplerate) except: print('error, skipping...')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/make_controls.py
annotation/helpers/helpers/make_controls.py
''' Make_controls.py Generate control data from a list of folders filled with .wav files. ''' import soundfile as sf import os, ffmpy, random, shutil # CONVERT FILE def convert_file(filename): #take in an audio file and convert with ffpeg file type #types of input files: .ogg #output file type: .wav newfile=filename[0:-4]+'.wav' ff = ffmpy.FFmpeg( inputs={filename:None}, outputs={newfile: None} ) ff.run() os.remove(filename) return newfile ############################################################################## default_dir=os.getcwd()+'/' os.chdir(default_dir) g=os.listdir() h=0 ############################################################################## # CONVERT ALL FILES TO WAV (COMMENTED OUT) convertfiles='n' if convertfiles in ['y','yes']: #check first if all the files are .wav files and if not convert them and delete the other file type for i in range(len(g)): try: if g[i] not in ['.DS_Store']: os.chdir(default_dir+g[i]) h=os.listdir() for j in range(len(h)): try: if h[j][-4:]!='.wav': print('converting %s'%(h[j])) new_file=convert_file(h[j]) except: print('error') except: print('error') else: pass ############################################################################## class_default=input('what is the class that will not be used as a control?') control_dir=class_default+'_controls' os.mkdir(control_dir) os.chdir(default_dir+class_default) q=os.listdir() filenum=len(q) movedlist=list() movedlist.append('') ind=g.index('.DS_Store') if ind>=0: del g[ind] else: pass #need equal amount of controls #loop over total number of files in class over the number of classes -1 (not including class) count=0 for i in range(int(len(q)/(len(g)-1))): for j in range(len(g)): print(g[j]) if g[j] not in [class_default]: try: print('changing to %s directory'%(default_dir+g[j])) os.chdir(default_dir+g[j]) h=os.listdir() file_num=len(h) cur_file='' count=0 while cur_file in movedlist: if count>file_num: break randint=random.randint(0,file_num-1) cur_file=h[randint] count=count+1 print('copying file: %s'%(cur_file)) shutil.copy(default_dir+g[j]+'/'+cur_file,default_dir+control_dir+'/'+cur_file) movedlist.append(cur_file) count=count+1 except: print('error') if count==0: for j in range(len(g)): print(g[j]) if g[j] not in [class_default]: try: print('changing to %s directory'%(default_dir+g[j])) os.chdir(default_dir+g[j]) h=os.listdir() file_num=len(h) cur_file='' count=0 while cur_file in movedlist: if count>file_num: break randint=random.randint(0,file_num-1) cur_file=h[randint] count=count+1 print('copying file: %s'%(cur_file)) shutil.copy(default_dir+g[j]+'/'+cur_file,default_dir+control_dir+'/'+cur_file) movedlist.append(cur_file) count=count+1 except: print('error')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/audio-network.py
annotation/helpers/helpers/audio-network.py
''' Draw graphs to visualize audio data with network theory. Network theory - https://github.com/networkx/networkx Documentation - https://networkx.github.io/documentation/networkx-1.10/tutorial/tutorial.html#drawing-graphs >>> import networkx as nx >>> G = nx.Graph() >>> G.add_edge('A', 'B', weight=4) >>> G.add_edge('B', 'D', weight=2) >>> G.add_edge('A', 'C', weight=3) >>> G.add_edge('C', 'D', weight=4) >>> nx.shortest_path(G, 'A', 'D', weight='weight') ['A', 'B', 'D'] ''' import networkx as nx import numpy.linalg import matplotlib.pyplot as plt n = 1000 # 1000 nodes m = 5000 # 5000 edges G = nx.gnm_random_graph(n,m) L = nx.normalized_laplacian_matrix(G) e = numpy.linalg.eigvals(L.A) print("Largest eigenvalue:", max(e)) print("Smallest eigenvalue:", min(e)) plt.hist(e,bins=100) # histogram with 100 bins plt.xlim(0,2) # eigenvalues between 0 and 2 plt.show() # mlab.show() # interactive window #TUTORIAL # can do this for many audio features (mfcc coefficients) ## ##G = nx.Graph(day="Friday") ##G.add_nodes_from(range(100,110)) ##G.add_node("spam") ##G.remove_nodes_from("spam") ## ##G.add_path(range(100,110)) ## ##G.add_edge('A', 'B', weight=1) ##G.add_edge('A', 'C', weight=1) ##G.add_edge('C', 'D', weight=1) ##G.add_edge('B', 'D', weight=1) ## ##G.remove_edge('B','D') ## ##nx.draw(G) ##plt.savefig("path.png") ## ## ##print(G.edges(data='weight')) ##print(G.number_of_nodes()) ##print(G.number_of_edges())
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/yscrape.py
annotation/helpers/helpers/yscrape.py
import os import json import pafy import json import time import wave import ffmpy import pandas as pd import soundfile as sf import shutil filename=input('what is the file name? \n') desktop="/Users/jim/Desktop/" os.chdir(desktop) foldername=filename[0:-5] destfolder=desktop+foldername+'/' try: os.mkdir(foldername) os.chdir(destfolder) except: os.chdir(destfolder) #move file to destfolder shutil.move(desktop+filename,destfolder+filename) #load xls sheet loadfile=pd.read_excel(filename) link=loadfile.iloc[:,0] length=loadfile.iloc[:,1] times=loadfile.iloc[:,2] label=loadfile.iloc[:,3] #initialize lists links=list() lengths=list() start_times=list() end_times=list() labels=list() #only make links that are in youtube processable for i in range(len(link)): if str(link[i]).find('youtube.com/watch') != -1: links.append(str(link[i])) lengths.append(str(length[i])) #find the dash for start/stop times time=str(times[i]) index=time.find('-') start_time=time[0:index] #get start time in seconds start_minutes=int(start_time[0]) start_seconds=int(start_time[-2:]) start_total=start_minutes*60+start_seconds #get end time in seconds end_time=time[index+1:] end_minutes=int(end_time[0]) end_seconds=int(end_time[-2:]) end_total=end_minutes*60+end_seconds #update lists start_times.append(start_total) end_times.append(end_total) #labels labels.append(str(label[i])) files=list() for i in range(len(links)): try: video=pafy.new(links[i]) bestaudio=video.getbestaudio() filename=bestaudio.download() start=start_times[i] end=end_times[i] extension=bestaudio.extension #get file extension and convert to .wav for processing later os.rename(filename,'%s_start_%s_end_%s%s'%(str(i),start,end,extension)) filename='%s_start_%s_end_%s%s'%(str(i),start,end,extension) if extension not in ['.wav']: xindex=filename.find(extension) filename=filename[0:xindex] ff=ffmpy.FFmpeg( inputs={filename+extension:None}, outputs={filename+'.wav':None} ) ff.run() os.remove(filename+extension) file=filename+'.wav' data,samplerate=sf.read(file) totalframes=len(data) totalseconds=totalframes/samplerate startsec=int(start_times[i]) startframe=samplerate*startsec endsec=int(end_times[i]) endframe=samplerate*endsec sf.write('snipped'+file, data[startframe:endframe], samplerate) os.remove(file) #can write json too except: print('no urls')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/remove_json.py
annotation/helpers/helpers/remove_json.py
''' Remove_json.py Remove all json files in sub-directories. Useful when you are cloning directories that have already been featurized to get new feature embeddings with nlx-model repo. ''' import os def removejson(listdir): for i in range(len(listdir)): if listdir[i][-5:]=='.json': os.remove(listdir[i]) listdir=os.listdir() hostdir=os.getcwd() for i in range(len(listdir)): try: os.chdir(hostdir+'/'+listdir[i]) listdir2=os.listdir() removejson(listdir2) except: pass
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/pickclass_byprobability.py
annotation/helpers/helpers/pickclass_byprobability.py
import pickle, os, json import numpy as np def pick_class(classlist): names=['teens','twenties','thirties','fourties','fifties','sixties','seventies'] probabilities=[.0666,.48888,.2296296,.08888,.08888,.0296,.0074] freqs=list() for i in range(len(classlist)): try: index=names.index(classlist[i]) freq=probabilities[index] freqs.append(freqs) except: freq=0 freqs.append(freq) #now pick the maxinum frequency maxfreq=np.amax(freqs) index=freqs.index(maxfreq) return classlist[index] def classify(features): listdir=os.listdir() model_list=['teens.pickle','twenties.pickle','thirties.pickle','fourties.pickle', 'fifties.pickle','sixties.pickle','seventies.pickle'] classlist=list() model_acc=list() deviations=list() modeltypes=list() modelslist=list() for i in range(len(model_list)): modelname=model_list[i] loadmodel=open(modelname,'rb') model=pickle.load(loadmodel) loadmodel.close() output=str(model.predict(features)[0]) classname=output if classname.count('controls')>0: pass else: classlist.append(classname) if len(classlist)>1: winclass=pick_class(classlist) modelslist.append(winclass+'.pickle') g=json.load(open(winclass+'.json')) model_acc.append(g['accuracy']) deviations.append(g['deviation']) modeltypes.append(g['modeltype']) elif len(classlist)==1: winclass=classlist[0] modelslist.append(winclass+'.pickle') g=json.load(open(winclass+'.json')) model_acc.append(g['accuracy']) deviations.append(g['deviation']) modeltypes.append(g['modeltype']) else: winclass='n/a' g=json.load(open(winclass+'.json')) model_acc.append(0) deviations.append(0) modeltypes.append('n/a') return winclass
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/automata.py
annotation/helpers/helpers/automata.py
''' Cellular automata Could be useful for audio applications. REFERENCES https://faingezicht.com/articles/2017/01/23/wolfram/ http://mathworld.wolfram.com/Rule30.html ''' def window(iterable, stride=3): for index in range(len(iterable) - stride + 1): yield iterable[index:index + stride] def generate_pattern(state, rule, MAX_TIME): for time in range(MAX_TIME): print(state) patterns = window(state) state = ''.join(rule[pat] for pat in patterns) state = '0{}0'.format(state) print(state) ''' window function creates list of all possible states list(window('footbar')) ['foo', 'oot', 'otb', 'tba', 'bar']''' #rule 30, 90, 110, 184 RULES = {30: {"111": '0', "110": '0', "101": '0', "000": '0', "100": '1', "011": '1', "010": '1', "001": '1'}, 90: {"111": "0", "110": "1", "101": "0", "100": "1", "011": "1", "010": "0", "001": "1", "000": "0"}, 110: {"111": '0', "110": '1', "101": '1', "100": '0', "011": '1', "010": '1', "001": '1', "000": '0'}, 184: {"111": "1", "110": "0", "101": "1", "100": "1", "011": "1", "010": "0", "001": "0", "000": "0"} } initial_state = '00000000000000000000100000000000000000000' list(window(initial_state)) generate_pattern(initial_state, RULES[30], 30) # figure out the rules of the system in speech propagation....
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/githubtable.py
annotation/helpers/helpers/githubtable.py
''' Load all model accuracies, names, and standard deviations and output them in a spreadsheet. This is intended for any model file directory using the nlx-model repository.''' import json, os, xlsxwriter, getpass def sort_list(list1, list2): zipped_pairs = zip(list2, list1) z = [x for _, x in sorted(zipped_pairs)] return z os.chdir(os.getcwd()[0:-(len('nlx-datalabeling'))]+'nlx-audiomodel/models') listdir=os.listdir() names=list() accs=list() stds=list() modeltypes=list() for i in range(len(listdir)): if listdir[i][-5:]=='.json': try: g=json.load(open(listdir[i])) acc=g['accuracy'] name=g['model'] std=g['deviation'] modeltype=g['modeltype'] names.append(name) accs.append(acc) stds.append(std) modeltypes.append(modeltype) except: print('error %s'%(listdir[i])) names=sort_list(names, accs) stds=sort_list(stds, accs) modeltypes=sort_list(modeltypes, accs) accs=sort_list(accs, accs) file = open('table.txt','w') file.write('| '+'Model Name' + ' |') file.write(' Accuracy'+' |') file.write(' Standard Deviation' + ' |') file.write(' Modeltype'+ ' |') file.write('\n') file.write('|-----|-----|-----|-----|') file.write('\n') print(names) for j in range(len(names)): file.write('| '+str(names[j])+' |') file.write(' '+str(accs[j])+' |') file.write(' '+str(stds[j])+' |') file.write(' '+str(modeltypes[j])+' |') file.write('\n') file.close() os.system('open %s'%(os.getcwd()+'/table.txt'))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/facedetect.py
annotation/helpers/helpers/facedetect.py
import numpy as np import cv2 #put these files on the desktop face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml') eye_cascade = cv2.CascadeClassifier('haarcascade_eye.xml') img = cv2.imread('face.png') gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) faces = face_cascade.detectMultiScale(gray, 1.3, 5) increment=0 for (x,y,w,h) in faces: cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2) newimg=img[y:y+h,x:x+w] #save only the face cv2.imwrite('only_face' + str(increment) + '.jpg', newimg) roi_gray = gray[y:y+h, x:x+w] roi_color = img[y:y+h, x:x+w] eyes = eye_cascade.detectMultiScale(roi_gray) for (ex,ey,ew,eh) in eyes: cv2.rectangle(roi_color,(ex,ey),(ex+ew,ey+eh),(0,255,0),2) increment=increment+1 cv2.imshow('img',img) cv2.imwrite('faces.png',img) #detect smile throughout all images (and get timestamp annotated) #detect sadness throughout all images (and get timestamp annotated)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/excel2json.py
annotation/helpers/helpers/excel2json.py
import librosa import os import soundfile as sf import xlsxwriter import pandas as pd import time import json exceldirectory='/Users/jim/Desktop/neurolex/voicemails/' jsondirectory='/Users/jim/Desktop/neurolex/voicemails/jsonfiles/' jsonexceldirectory='/Users/jim/Desktop/neurolex/voicemails/jsonfiles-excel/' os.chdir(exceldirectory) ls=pd.read_excel('voicemails.xlsx') filename=ls.iloc[:,0] #name of file to align with json files gender=ls.iloc[:,2] #0=music, 1=male, 2=femaile, 3=male/female, 4=child, 5=multi-child) age=ls.iloc[:,3] #0=music, 1=adult, 2=child sadness=ls.iloc[:,4] #0=music, 1=least, 10=most happiness=ls.iloc[:,5] #0=music, 1=least, 10=most stress=ls.iloc[6] #0=music, 1=lest, 10=most dialect=ls.iloc[7] #0=music, 1=american dialect, 2=foreign dialect voicemusic=ls.iloc[8] #1=voice, 2=music, 3=multi-music and voice fatigue=ls.iloc[9] #0=music, 1=least, 10=most audioquality=ls.iloc[10] #0=nothing, 1=lowest, 10=highest sickness=ls.iloc[11] #1=natural, 2=non-natural, 3=music, 4=sick ##os.chdir(jsondirectory) ##jsonfiles=os.listdir() os.chdir(jsonexceldirectory) for g in range(len(filename)): #try: #find json file recording ID field ## jsonfileread=open(jsonfiles[i],'r').read() ## jsonfile=json.loads(jsonfileread) ## jsonfileid=jsonfile['recordingID'] ## #search for this recording ID in excel file ## for g in range(len(filename)): ## if filename[g]==jsonfileid: ## indval=int(g) #looks for the index of the filename matching row #creates array of new data to add to json newdata={ 'filename':filename[g], 'gender': int(gender[g]), 'age': int(age[g]), 'sadness': int(sadness[g]), 'happiness': int(happiness[g]), 'stress': stress[g], 'dialect': int(dialect[g]), 'voicemusic': int(voicemusic[g]), 'fatigue': int(fatigue[g]), 'audioquality': int(audioquality[g]), 'sickness': int(sickness[g]), } #dump to new directory json.dump(jsonfile) #except: #if not in excel file and no match, print this # print('no file match found')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/markov-chain.py
annotation/helpers/helpers/markov-chain.py
''' Make markov chain. Following tutorial on Datacamp https://www.datacamp.com/community/tutorials/markov-chains-python-tutorial ''' import numpy as np import random # the statespace states = ["Sleep","Icecream","Run"] # Possible sequences of events transitionName = [["SS","SR","SI"],["RS","RR","RI"],["IS","IR","II"]] # Probabilities matrix (transition matrix) transitionMatrix = [[0.2,0.6,0.2],[0.1,0.6,0.3],[0.2,0.7,0.1]] if sum(transitionMatrix[0])+sum(transitionMatrix[1])+sum(transitionMatrix[1]) != 3: print("Somewhere, something went wrong. Transition matrix, perhaps?") else: print("All is gonna be okay, you should move on!! ;)") # A function that implements the Markov model to forecast the state/mood. def activity_forecast(days): # Choose the starting state activityToday = "Sleep" activityList = [activityToday] i = 0 prob = 1 while i != days: if activityToday == "Sleep": change = np.random.choice(transitionName[0],replace=True,p=transitionMatrix[0]) if change == "SS": prob = prob * 0.2 activityList.append("Sleep") pass elif change == "SR": prob = prob * 0.6 activityToday = "Run" activityList.append("Run") else: prob = prob * 0.2 activityToday = "Icecream" activityList.append("Icecream") elif activityToday == "Run": change = np.random.choice(transitionName[1],replace=True,p=transitionMatrix[1]) if change == "RR": prob = prob * 0.5 activityList.append("Run") pass elif change == "RS": prob = prob * 0.2 activityToday = "Sleep" activityList.append("Sleep") else: prob = prob * 0.3 activityToday = "Icecream" activityList.append("Icecream") elif activityToday == "Icecream": change = np.random.choice(transitionName[2],replace=True,p=transitionMatrix[2]) if change == "II": prob = prob * 0.1 activityList.append("Icecream") pass elif change == "IS": prob = prob * 0.2 activityToday = "Sleep" activityList.append("Sleep") else: prob = prob * 0.7 activityToday = "Run" activityList.append("Run") i += 1 return activityList # To save every activityList list_activity = [] count = 0 # `Range` starts from the first count up until but excluding the last count for iterations in range(1,10000): list_activity.append(activity_forecast(2)) # Check out all the `activityList` we collected #print(list_activity) # Iterate through the list to get a count of all activities ending in state:'Run' for smaller_list in list_activity: if(smaller_list[2] == "Run"): count += 1 # Calculate the probability of starting from state:'Sleep' and ending at state:'Run' percentage = (count/10000) * 100 print("The probability of starting at state:'Sleep' and ending at state:'Run'= " + str(percentage) + "%")
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/visualizemodels.py
annotation/helpers/helpers/visualizemodels.py
''' Load all model accuracies, names, and standard deviations and output them in a spreadsheet. This is intended for any model file directory using the nlx-model repository.''' import json, os, xlsxwriter, getpass os.chdir('/Users/'+getpass.getuser()+'/nlx-model/nlx-audiomodel/models') listdir=os.listdir() names=list() accs=list() stds=list() modeltypes=list() for i in range(len(listdir)): if listdir[i][-5:]=='.json': try: g=json.load(open(listdir[i])) acc=g['accuracy'] name=g['model'] std=g['deviation'] modeltype=g['modeltype'] names.append(name) accs.append(acc) stds.append(std) modeltypes.append(modeltype) except: print('error %s'%(listdir[i])) workbook = xlsxwriter.Workbook('summary.xlsx') worksheet = workbook.add_worksheet() worksheet.write('A1', 'Model Name') worksheet.write('B1', 'Accuracy') worksheet.write('C1', 'Standard Deviation') worksheet.write('D1', 'Modeltype') for j in range(len(names)): worksheet.write('A%s'%(str(j+2)), names[j]) worksheet.write('B%s'%(str(j+2)), accs[j]) worksheet.write('C%s'%(str(j+2)), stds[j]) worksheet.write('D%s'%(str(j+2)), modeltypes[j]) workbook.close() os.system('open %s'%(os.getcwd()+'/summary.xlsx'))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/helpers/extract_noise.py
annotation/helpers/helpers/extract_noise.py
import shutil, os, random from pydub import AudioSegment try: os.mkdir('noise') except: shutil.rmtree('noise') os.mkdir('noise') def extract_noise(filename, length): song = AudioSegment.from_mp3(filename) first = song[100:100+length] first.export(filename[0:-4]+'_noise.mp3') shutil.move(os.getcwd()+'/'+filename[0:-4]+'_noise.mp3', os.getcwd()+'/noise/'+filename[0:-4]+'_noise.mp3') listdir=os.listdir() mp3files=list() for i in range(len(listdir)): if listdir[i][-4:]=='.mp3': mp3files.append(listdir[i]) random.shuffle(mp3files) for i in range(len(mp3files)): extract_noise(mp3files[i],300) if i == 100: break os.chdir('noise') listdir=os.listdir() for i in range(len(listdir)): if listdir[i][-4:]=='.mp3': os.system('play %s'%(listdir[i])) remove=input('should remove? type y to remove') if remove=='y': os.remove(listdir[i])
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/youtube_scrape/download_playlist.py
annotation/helpers/youtube_scrape/download_playlist.py
''' ================================================ YOUTUBE_SCRAPE REPOSITORY ================================================ repository name: youtube_scrape repository version: 1.0 repository link: https://github.com/jim-schwoebel/youtube_scrape author: Jim Schwoebel author contact: js@neurolex.co description: Library for scraping youtube videos. Alternative to pafy, pytube, and youtube-dl. license category: opensource license: Apache 2.0 license organization name: NeuroLex Laboratories, Inc. location: Seattle, WA website: https://neurolex.ai release date: 2018-07-23 This code (youtube_scrape) is hereby released under a Apache 2.0 license license. For more information, check out the license terms below. ================================================ LICENSE TERMS ================================================ Copyright 2018 NeuroLex Laboratories, Inc. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ================================================ SERVICE STATEMENT ================================================ If you are using the code written for a larger project, we are happy to consult with you and help you with deployment. Our team has >10 world experts in kafka distributed architectures, microservices built on top of Node.JS / python / docker, and applying machine learning to model speech and text data. We have helped a wide variety of enterprises - small businesses, researchers, enterprises, and/or independent developers. If you would like to work with us let us know @ js@neurolex.co. ================================================ NOTE ================================================ Download a playlist from the URLS previously generated with make_playlist.py script. Make sure you have at least around 10GB of disk space before bulk downloading videos, as they can take up a lot of space. ''' import requests, json, os, shutil from bs4 import BeautifulSoup from pytube import YouTube playlist_name=input('what is the name of the playlist to download?') hostdir=os.getcwd() os.chdir(os.getcwd()+'/playlists/') try: if playlist_name[-5:] != '.json': g=json.load(open(playlist_name+'.json')) entries=g['entries'] links=list() elif playlist_name[-5:] == '.json': g=json.load(open(playlist_name)) entries=g['entries'] links=list() except: print('error loading playlist. Please make sure it is in the playlists folder and you type in the name properly. \n\n For example yc_podcast.json ==> yc_podcast or yc_podcast.json') if playlist_name[-5:]=='.json': foldername=playlist_name[0:-5] else: foldername=playlist_name foldername try: os.mkdir(foldername) os.chdir(foldername) except: shutil.rmtree(foldername) os.mkdir(foldername) os.chdir(foldername) for i in range(len(entries)): link=entries[i]['link'] links.append(link) print(link) # download files for i in range(len(links)): try: link=links[i] print('downloading %s'%(link)) YouTube(link).streams.first().download() except: print('error') # rename videos in order listdir=os.listdir() for i in range(len(listdir)): if listdir[i][-5:] in ['.webm']: os.rename(listdir[i],str(i)+'.webm') os.system('ffmpeg -i %s %s'%(str(i)+'.webm',str(i)+'.mp4')) os.remove(str(i)+'.webm') elif listdir[i][-4:] in ['.mp4']: os.rename(listdir[i],str(i)+'.mp4') # now make audio for each .mp4 file listdir=os.listdir() for i in range(len(listdir)): if listdir[i][-4:]=='.mp4': os.system('ffmpeg -i %s %s'%(listdir[i],listdir[i][0:-4]+'.wav'))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/youtube_scrape/make_playlist.py
annotation/helpers/youtube_scrape/make_playlist.py
''' ================================================ YOUTUBE_SCRAPE REPOSITORY ================================================ repository name: youtube_scrape repository version: 1.0 repository link: https://github.com/jim-schwoebel/youtube_scrape author: Jim Schwoebel author contact: js@neurolex.co description: Library for scraping youtube videos. Alternative to pafy, pytube, and youtube-dl. license category: opensource license: Apache 2.0 license organization name: NeuroLex Laboratories, Inc. location: Seattle, WA website: https://neurolex.ai release date: 2018-07-23 This code (youtube_scrape) is hereby released under a Apache 2.0 license license. For more information, check out the license terms below. ================================================ LICENSE TERMS ================================================ Copyright 2018 NeuroLex Laboratories, Inc. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ================================================ SERVICE STATEMENT ================================================ If you are using the code written for a larger project, we are happy to consult with you and help you with deployment. Our team has >10 world experts in kafka distributed architectures, microservices built on top of Node.JS / python / docker, and applying machine learning to model speech and text data. We have helped a wide variety of enterprises - small businesses, researchers, enterprises, and/or independent developers. If you would like to work with us let us know @ js@neurolex.co. ''' ########################################################################### # IMPORT STATEMENTS ## ########################################################################### import requests, json, os from bs4 import BeautifulSoup from pytube import YouTube ########################################################################### # HELPER FUNCTIONS ## ########################################################################### def scrapelinks(playlist, links): #https://www.youtube.com/playlist?list=PL1v-PVIZFDsqbzPIsEPZPnvcgIQ8bNTKS page=requests.get(playlist) base='https://www.youtube.com/watch?v=' soup=BeautifulSoup(page.content, 'lxml') g=soup.find_all('tr',class_='pl-video yt-uix-tile ') entries=list() totaltime=0 for i in range(len(g)): try: h=str(g[i]) # get titles h1=h.find('data-title="')+len('data-title="') h2=h[h1:].find('"') title=h[h1:h1+h2] # get links h3=h.find('data-video-id="')+len('data-video-id="') h4=h[h3:].find('"') link=base+h[h3:h3+h4] # get duration (in seconds) h5=h.find('<div class="timestamp"><span aria-label="') h6=h[h5:] hsoup=BeautifulSoup(h6,'lxml') htext=hsoup.text.replace('\n','').replace(' ','') hmin=htext.split(':') duration=int(hmin[0])*60+int(hmin[1]) totaltime=totaltime+duration if link not in links: # avoids duplicate links links.append(link) entry={ 'title':title, 'link':link, 'duration':duration } entries.append(entry) except: print('error') return entries, len(entries), totaltime, links ########################################################################### ## MAIN CODE BASE ## ########################################################################### playlists=list() entries=list() t=1 totalnum=0 totaltime=0 links=list() playlist_name=input('what do you want to name this playlist (e.g. angry)?') while t>0: #try: playlist=input('what is the playlist id or URL?') if playlist.find('playlist?list=')>0: playlists.append(playlist) entry, enum, nowtime, link=scrapelinks(playlist, links) links=links+link totalnum=totalnum+enum totaltime=totaltime+nowtime entries=entries+entry elif playlist not in ['', 'n']: playlist='https://www.youtube.com/playlist?list='+playlist playlists.append(playlist) entry, enum, nowtime, link=scrapelinks(playlist, links) links=links+link totalnum=totalnum+enum totaltime=totaltime+nowtime entries=entries+entry else: break #except: #print('error') os.chdir(os.getcwd()+'/playlists') data={ 'entrynum':totalnum, 'total time':totaltime, 'playlist url':playlists, 'entries':entries, } jsonfile=open(playlist_name+'.json','w') json.dump(data,jsonfile) jsonfile.close()
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/youtube_scrape/archived/get_audio.py
annotation/helpers/youtube_scrape/archived/get_audio.py
''' get audio from each file (for processing) ''' import os folder=input('what playlist do you want audio?') os.chdir(folder) listdir=os.listdir() for i in range(len(listdir)): if listdir[i][-4:]=='.mp4': os.system('ffmpeg -i %s %s'%(listdir[i],listdir[i][0:-4]+'.wav'))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/youtube_scrape/archived/download_playlist_url.py
annotation/helpers/youtube_scrape/archived/download_playlist_url.py
''' Extract playlist URLs (for further processing) ''' import requests, json, os from bs4 import BeautifulSoup from pytube import YouTube base='https://www.youtube.com/watch?v=' playlist_name=input('what do you want to name this playlist (e.g. angry)?') #angry playlist=input('what is the playlist url?') #https://www.youtube.com/playlist?list=PL1v-PVIZFDsqbzPIsEPZPnvcgIQ8bNTKS page=requests.get(playlist) soup=BeautifulSoup(page.content, 'lxml') g=soup.find_all('tr',class_='pl-video yt-uix-tile ') entries=list() links=list() totaltime=0 for i in range(len(g)): try: h=str(g[i]) # get titles h1=h.find('data-title="')+len('data-title="') h2=h[h1:].find('"') title=h[h1:h1+h2] # get links h3=h.find('data-video-id="')+len('data-video-id="') h4=h[h3:].find('"') link=base+h[h3:h3+h4] # get duration (in seconds) h5=h.find('<div class="timestamp"><span aria-label="') h6=h[h5:] hsoup=BeautifulSoup(h6,'lxml') htext=hsoup.text.replace('\n','').replace(' ','') hmin=htext.split(':') duration=int(hmin[0])*60+int(hmin[1]) totaltime=totaltime+duration if link not in links: # avoids duplicate links links.append(link) entry={ 'title':title, 'link':link, 'duration':duration } entries.append(entry) except: print('error') os.mkdir(playlist_name) os.chdir(os.getcwd()+'/'+playlist_name) data={ 'entrynum':len(entries), 'total time':totaltime, 'playlist url':playlist, 'entries':entries, } jsonfile=open('entries.json','w') json.dump(data,jsonfile) jsonfile.close() for i in range(len(entries)): try: link=entries[i]['link'] print('downloading %s'%(link)) YouTube(link).streams.first().download() except: print('error') # rename videos in order listdir=os.listdir() for i in range(len(listdir)): if listdir[i][-5:] in ['.webm']: os.rename(listdir[i],str(i)+'.webm') os.system('ffmpeg -i %s %s'%(str(i)+'.webm',str(i)+'.mp4')) os.remove(str(i)+'.webm') elif listdir[i][-4:] in ['.mp4']: os.rename(listdir[i],str(i)+'.mp4')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/annotation/helpers/youtube_scrape/archived/download_playlist_id.py
annotation/helpers/youtube_scrape/archived/download_playlist_id.py
''' Extract playlist URLs (for further processing) ''' import requests, json, os from bs4 import BeautifulSoup from pytube import YouTube base='https://www.youtube.com/watch?v=' playlist_name=input('what do you want to name this playlist (e.g. angry)?') #angry playlist=input('what is the playlist id?') playlist='https://www.youtube.com/playlist?list='+playlist #https://www.youtube.com/playlist?list=PL1v-PVIZFDsqbzPIsEPZPnvcgIQ8bNTKS page=requests.get(playlist) soup=BeautifulSoup(page.content, 'lxml') g=soup.find_all('tr',class_='pl-video yt-uix-tile ') entries=list() links=list() totaltime=0 for i in range(len(g)): try: h=str(g[i]) # get titles h1=h.find('data-title="')+len('data-title="') h2=h[h1:].find('"') title=h[h1:h1+h2] # get links h3=h.find('data-video-id="')+len('data-video-id="') h4=h[h3:].find('"') link=base+h[h3:h3+h4] # get duration (in seconds) h5=h.find('<div class="timestamp"><span aria-label="') h6=h[h5:] hsoup=BeautifulSoup(h6,'lxml') htext=hsoup.text.replace('\n','').replace(' ','') hmin=htext.split(':') duration=int(hmin[0])*60+int(hmin[1]) totaltime=totaltime+duration if link not in links: # avoids duplicate links links.append(link) entry={ 'title':title, 'link':link, 'duration':duration } entries.append(entry) except: print('error') os.mkdir(playlist_name) os.chdir(os.getcwd()+'/'+playlist_name) data={ 'entrynum':len(entries), 'total time':totaltime, 'playlist url':playlist, 'entries':entries, } jsonfile=open('entries.json','w') json.dump(data,jsonfile) jsonfile.close() for i in range(len(entries)): try: link=entries[i]['link'] print('downloading %s'%(link)) YouTube(link).streams.first().download() except: print('error') # rename videos in order listdir=os.listdir() for i in range(len(listdir)): if listdir[i][-5:] in ['.webm']: os.rename(listdir[i],str(i)+'.webm') os.system('ffmpeg -i %s %s'%(str(i)+'.webm',str(i)+'.mp4')) os.remove(str(i)+'.webm') elif listdir[i][-4:] in ['.mp4']: os.rename(listdir[i],str(i)+'.mp4')
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/visualize/visualize.py
visualize/visualize.py
''' AAA lllllll lllllll iiii A:::A l:::::l l:::::l i::::i A:::::A l:::::l l:::::l iiii A:::::::A l:::::l l:::::l A:::::::::A l::::l l::::l iiiiiii eeeeeeeeeeee A:::::A:::::A l::::l l::::l i:::::i ee::::::::::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::eeeee:::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::e e:::::e A:::::A A:::::A l::::l l::::l i::::i e:::::::eeeee::::::e A:::::AAAAAAAAA:::::A l::::l l::::l i::::i e:::::::::::::::::e A:::::::::::::::::::::A l::::l l::::l i::::i e::::::eeeeeeeeeee A:::::AAAAAAAAAAAAA:::::A l::::l l::::l i::::i e:::::::e A:::::A A:::::A l::::::ll::::::li::::::ie::::::::e A:::::A A:::::A l::::::ll::::::li::::::i e::::::::eeeeeeee A:::::A A:::::A l::::::ll::::::li::::::i ee:::::::::::::e AAAAAAA AAAAAAAlllllllllllllllliiiiiiii eeeeeeeeeeeeee | | | (_) | (_) / _ \ | ___ \_ _| | | | |_ ___ _ _ __ _| |_ _______ / /_\ \| |_/ / | | | | | | / __| | | |/ _` | | |_ / _ \ | _ || __/ | | \ \_/ / \__ \ |_| | (_| | | |/ / __/ | | | || | _| |_ \___/|_|___/\__,_|\__,_|_|_/___\___| \_| |_/\_| \___/ Takes in a folder or set of folders of featurized files and outputs visualizations to look deeper at the data. This is often useful as a precursor before building machine learning models to uncover relationships in the data. Note that this automatically happens as part of the modeling process if visualize==True in settings. This is also restricted to classification problems for version 1.0 of Allie. Usage: python3 visualize.py [problemtype] [folder A] [folder B] ... [folder N] Example: python3 visualize.py audio males females ''' import os, sys, json, time, shutil os.system('pip3 install yellowbrick==1.1 scikit-plot==0.3.7 umap==0.1.1 umap-learn==0.4.4') from tqdm import tqdm from yellowbrick.features import Rank1D, Rank2D, Manifold, FeatureImportances from yellowbrick.features.pca import PCADecomposition from sklearn.ensemble import ExtraTreesClassifier import matplotlib.pyplot as plt from sklearn import preprocessing import numpy as np from yellowbrick.text import UMAPVisualizer import pandas as pd from sklearn.linear_model import LogisticRegression from sklearn.naive_bayes import GaussianNB from yellowbrick.classifier import precision_recall_curve, discrimination_threshold from sklearn.linear_model import Ridge from sklearn.linear_model import Lasso from yellowbrick.regressor import residuals_plot from yellowbrick.regressor import prediction_error from sklearn.cluster import KMeans from yellowbrick.cluster import silhouette_visualizer from sklearn.cluster import MiniBatchKMeans from yellowbrick.cluster import intercluster_distance from sklearn.metrics import auc, roc_curve from yellowbrick.classifier.rocauc import roc_auc from yellowbrick.regressor import cooks_distance from yellowbrick.features import RadViz from yellowbrick.target.feature_correlation import feature_correlation import seaborn as sns import umap from sklearn.model_selection import train_test_split # feature selection from sklearn.feature_selection import SelectPercentile, chi2 from sklearn.model_selection import cross_val_score from sklearn.pipeline import Pipeline from sklearn.preprocessing import StandardScaler from sklearn.svm import SVC from sklearn.preprocessing import MinMaxScaler # other things in scikitlearn import scikitplot as skplt import matplotlib.pyplot as plt from sklearn.decomposition import PCA from sklearn.datasets import load_digits from sklearn.ensemble import AdaBoostClassifier, RandomForestClassifier, GradientBoostingClassifier from sklearn.neighbors import KNeighborsClassifier from sklearn.linear_model import LogisticRegression from sklearn.model_selection import cross_val_predict from sklearn.tree import DecisionTreeClassifier from sklearn.svm import LinearSVC from sklearn.naive_bayes import GaussianNB from sklearn.metrics import roc_curve from sklearn.cluster import KMeans from sklearn import metrics from itertools import cycle import random def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ def get_classes(): count=2 classes=list() while True: try: class_=sys.argv[count] classes.append(class_) print(classes) count=count+1 except: break # added this in for the CLI if classes == []: classnum=input('how many classes do you want to model? (e.g. 2)\n') classnum = int(classnum) for i in range(classnum): classes.append(input('what is class #%s\n'%(str(i+1)))) return classes def get_features(classes, problem_type, default_features, balance_data): features=list() feature_labels=list() class_labels=list() curdir=os.getcwd() lengths=list() minlength=0 if balance_data == True: for i in range(len(classes)): os.chdir(curdir+'/'+classes[i]) listdir=os.listdir() jsonfiles=list() for j in range(len(listdir)): if listdir[j].endswith('.json'): jsonfiles.append(listdir[j]) lengths.append(len(jsonfiles)) minlength=np.amin(lengths) print('minimum length is...') print(minlength) time.sleep(2) for i in range(len(classes)): print('----------LOADING %s----------'%(classes[i].upper())) os.chdir(curdir+'/'+classes[i]) listdir=os.listdir() jsonfiles=list() for j in range(len(listdir)): if listdir[j].endswith('.json'): jsonfiles.append(listdir[j]) g=json.load(open(jsonfiles[0])) feature_list=list(g['features'][problem_type]) for j in tqdm(range(len(jsonfiles))): if balance_data==True: if class_labels.count(classes[i]) > minlength: break else: try: g=json.load(open(jsonfiles[j])) feature_=list() label_=list() for k in range(len(feature_list)): if feature_list[k] in default_features: feature_=feature_+g['features'][problem_type][feature_list[k]]['features'] label_=label_+g['features'][problem_type][feature_list[k]]['labels'] # quick quality check to only add to list if the feature_labels match in length the features_ if len(feature_) == len(label_): features.append(feature_) feature_labels.append(label_) class_labels.append(classes[i]) except: print('error loading feature embedding: %s'%(feature_list[k].upper())) else: try: g=json.load(open(jsonfiles[j])) feature_=list() label_=list() for k in range(len(feature_list)): if feature_list[k] in default_features: feature_=feature_+g['features'][problem_type][feature_list[k]]['features'] label_=label_+g['features'][problem_type][feature_list[k]]['labels'] # quick quality check to only add to list if the feature_labels match in length the features_ if len(feature_) == len(label_): features.append(feature_) feature_labels.append(label_) class_labels.append(classes[i]) except: print('error loading feature embedding: %s'%(feature_list[k].upper())) return features, feature_labels, class_labels def plot_roc_curve(y_test, probs, clf_names): cycol = cycle('bgrcmyk') for i in range(len(probs)): try: fper, tper, thresholds = roc_curve(y_test, probs[i]) plt.plot(fper, tper, color=next(cycol), label=clf_names[i]+' = %s'%(str(round(metrics.auc(fper, tper), 2)))) plt.plot([0, 1], [0, 1], color='darkblue', linestyle='--') except: print('passing %s'%(clf_names[i])) plt.xlabel('False Positive Rate') plt.ylabel('True Positive Rate') plt.title('Receiver Operating Characteristic (ROC) Curve') plt.legend() plt.savefig('roc_curve.png') plt.close() def get_indices(selectedlabels, labels): ''' takes in an list of labels and gets back indices for these labels; useful to restructure arrays around the right features for heatmaps. selectedfeatures = string of selected features from univariate feature selection labels = string of all potential labels in right order ''' indices=list() for i in range(len(selectedlabels)): indices.append(labels.index(selectedlabels[i])) return indices def restructure_features(selectedlabels, features, labels): # get top 20 features instead of all features for the heatmaps # get top 20 feature indices indices=get_indices(selectedlabels, labels) # now that we have the indices, only select these indices in all numpy array features newfeatures=list() for i in range(len(features)): feature=dict() for j in range(len(indices)): feature[selectedlabels[j]]=features[i][indices[j]] newfeatures.append(feature) newfeatures=pd.DataFrame(newfeatures) newfeatures.to_csv('data.csv',index=False) return newfeatures, selectedlabels def visualize_features(classes, problem_type, curdir, default_features, balance_data, test_size): # make features into label encoder here features, feature_labels, class_labels = get_features(classes, problem_type, default_features, balance_data) # now preprocess features for all the other plots os.chdir(curdir) le = preprocessing.LabelEncoder() le.fit(class_labels) tclass_labels = le.transform(class_labels) # process features to help with clustering se=preprocessing.StandardScaler() t_features=se.fit_transform(features) X_train, X_test, y_train, y_test = train_test_split(features, tclass_labels, test_size=test_size, random_state=42) # print(len(features)) # print(len(feature_labels)) # print(len(class_labels)) # print(class_labels) # GET TRAINING DATA DURING MODELING PROCESS ################################## # get filename # csvfile='' # print(classes) # for i in range(len(classes)): # csvfile=csvfile+classes[i]+'_' # get training and testing data for later # try: # print('loading training files...') # X_train=pd.read_csv(prev_dir(curdir)+'/models/'+csvfile+'train.csv') # y_train=X_train['class_'] # X_train.drop(['class_'], axis=1) # X_test=pd.read_csv(prev_dir(curdir)+'/models/'+csvfile+'test.csv') # y_test=X_test['class_'] # X_test.drop(['class_'], axis=1) # y_train=le.inverse_transform(y_train) # y_test=le.inverse_transform(y_test) # except: # print('error loading in training files, making new test data') # Visualize each class (quick plot) ################################## visualization_dir='visualization_session' try: os.mkdir(visualization_dir) os.chdir(visualization_dir) except: shutil.rmtree(visualization_dir) os.mkdir(visualization_dir) os.chdir(visualization_dir) objects = tuple(set(class_labels)) y_pos = np.arange(len(objects)) performance=list() for i in range(len(objects)): performance.append(class_labels.count(objects[i])) plt.bar(y_pos, performance, align='center', alpha=0.5) plt.xticks(y_pos, objects) plt.xticks(rotation=90) plt.title('Counts per class') plt.ylabel('Count') plt.xlabel('Class') plt.tight_layout() plt.savefig('classes.png') plt.close() # set current directory curdir=os.getcwd() # ################################## # # CLUSTERING!!! # ################################## ################################## # Manifold type options ################################## ''' "lle" Locally Linear Embedding (LLE) uses many local linear decompositions to preserve globally non-linear structures. "ltsa" LTSA LLE: local tangent space alignment is similar to LLE in that it uses locality to preserve neighborhood distances. "hessian" Hessian LLE an LLE regularization method that applies a hessian-based quadratic form at each neighborhood "modified" Modified LLE applies a regularization parameter to LLE. "isomap" Isomap seeks a lower dimensional embedding that maintains geometric distances between each instance. "mds" MDS: multi-dimensional scaling uses similarity to plot points that are near to each other close in the embedding. "spectral" Spectral Embedding a discrete approximation of the low dimensional manifold using a graph representation. "tsne" (default) t-SNE: converts the similarity of points into probabilities then uses those probabilities to create an embedding. ''' os.mkdir('clustering') os.chdir('clustering') # tSNE plt.figure() viz = Manifold(manifold="tsne", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="tsne.png") plt.close() # os.system('open tsne.png') # viz.show() # PCA plt.figure() visualizer = PCADecomposition(scale=True, classes=set(classes)) visualizer.fit_transform(np.array(features), tclass_labels) visualizer.poof(outpath="pca.png") plt.close() # os.system('open pca.png') # spectral embedding plt.figure() viz = Manifold(manifold="spectral", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="spectral.png") plt.close() # lle embedding plt.figure() viz = Manifold(manifold="lle", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="lle.png") plt.close() # ltsa # plt.figure() # viz = Manifold(manifold="ltsa", classes=set(classes)) # viz.fit_transform(np.array(features), tclass_labels) # viz.poof(outpath="ltsa.png") # plt.close() # hessian # plt.figure() # viz = Manifold(manifold="hessian", method='dense', classes=set(classes)) # viz.fit_transform(np.array(features), tclass_labels) # viz.poof(outpath="hessian.png") # plt.close() # modified plt.figure() viz = Manifold(manifold="modified", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="modified.png") plt.close() # isomap plt.figure() viz = Manifold(manifold="isomap", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="isomap.png") plt.close() # mds plt.figure() viz = Manifold(manifold="mds", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="mds.png") plt.close() # spectral plt.figure() viz = Manifold(manifold="spectral", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="spectral.png") plt.close() # UMAP embedding plt.figure() umap = UMAPVisualizer(metric='cosine', classes=set(classes), title="UMAP embedding") umap.fit_transform(np.array(features), class_labels) umap.poof(outpath="umap.png") plt.close() # alternative UMAP # import umap.plot # plt.figure() # mapper = umap.UMAP().fit(np.array(features)) # fig=umap.plot.points(mapper, labels=np.array(tclass_labels)) # fig = fig.get_figure() # fig.tight_layout() # fig.savefig('umap2.png') # plt.close(fig) ################################# # FEATURE RANKING!! ################################# os.chdir(curdir) os.mkdir('feature_ranking') os.chdir('feature_ranking') # You can get the feature importance of each feature of your dataset # by using the feature importance property of the model. plt.figure(figsize=(12,12)) model = ExtraTreesClassifier() model.fit(np.array(features),tclass_labels) # print(model.feature_importances_) feat_importances = pd.Series(model.feature_importances_, index=feature_labels[0]) feat_importances.nlargest(20).plot(kind='barh') plt.title('Feature importances (ExtraTrees)', size=16) plt.title('Feature importances with %s features'%(str(len(features[0])))) plt.tight_layout() plt.savefig('feature_importance.png') plt.close() # os.system('open feature_importance.png') # get selected labels for top 20 features selectedlabels=list(dict(feat_importances.nlargest(20))) new_features, new_labels = restructure_features(selectedlabels, t_features, feature_labels[0]) new_features_, new_labels_ = restructure_features(selectedlabels, features, feature_labels[0]) # Shapiro rank algorithm (1D) plt.figure(figsize=(28,12)) visualizer = Rank1D(algorithm='shapiro', classes=set(classes), features=new_labels) visualizer.fit(np.array(new_features), tclass_labels) visualizer.transform(np.array(new_features)) # plt.tight_layout() visualizer.poof(outpath="shapiro.png") plt.title('Shapiro plot (top 20 features)', size=16) plt.close() # os.system('open shapiro.png') # visualizer.show() # pearson ranking algorithm (2D) plt.figure(figsize=(12,12)) visualizer = Rank2D(algorithm='pearson', classes=set(classes), features=new_labels) visualizer.fit(np.array(new_features), tclass_labels) visualizer.transform(np.array(new_features)) plt.tight_layout() visualizer.poof(outpath="pearson.png") plt.title('Pearson ranking plot (top 20 features)', size=16) plt.close() # os.system('open pearson.png') # visualizer.show() # feature importances with top 20 features for Lasso plt.figure(figsize=(12,12)) viz = FeatureImportances(Lasso(), labels=new_labels_) viz.fit(np.array(new_features_), tclass_labels) plt.tight_layout() viz.poof(outpath="lasso.png") plt.close() # correlation plots with feature removal if corr > 0.90 # https://towardsdatascience.com/feature-selection-correlation-and-p-value-da8921bfb3cf # now remove correlated features # --> p values # --> https://towardsdatascience.com/the-next-level-of-data-visualization-in-python-dd6e99039d5e / https://github.com/WillKoehrsen/Data-Analysis/blob/master/plotly/Plotly%20Whirlwind%20Introduction.ipynb- plotly for correlation heatmap and scatterplot matrix # --> https://seaborn.pydata.org/tutorial/distributions.html data=new_features corr = data.corr() plt.figure(figsize=(12,12)) fig=sns.heatmap(corr) fig = fig.get_figure() plt.title('Heatmap with correlated features (top 20 features)', size=16) fig.tight_layout() fig.savefig('heatmap.png') plt.close(fig) columns = np.full((corr.shape[0],), True, dtype=bool) for i in range(corr.shape[0]): for j in range(i+1, corr.shape[0]): if corr.iloc[i,j] >= 0.9: if columns[j]: columns[j] = False selected_columns = data.columns[columns] data = data[selected_columns] corr=data.corr() plt.figure(figsize=(12,12)) fig=sns.heatmap(corr) fig = fig.get_figure() plt.title('Heatmap without correlated features (top 20 features)', size=16) fig.tight_layout() fig.savefig('heatmap_clean.png') plt.close(fig) # radviz # Instantiate the visualizer plt.figure(figsize=(12,12)) visualizer = RadViz(classes=classes, features=new_labels) visualizer.fit(np.array(new_features), tclass_labels) visualizer.transform(np.array(new_features)) visualizer.poof(outpath="radviz.png") visualizer.show() plt.close() # feature correlation plot plt.figure(figsize=(28,12)) visualizer = feature_correlation(np.array(new_features), tclass_labels, labels=new_labels) visualizer.poof(outpath="correlation.png") visualizer.show() plt.tight_layout() plt.close() os.mkdir('feature_plots') os.chdir('feature_plots') newdata=new_features_ newdata['classes']=class_labels for j in range(len(new_labels_)): fig=sns.violinplot(x=newdata['classes'], y=newdata[new_labels_[j]]) fig = fig.get_figure() fig.tight_layout() fig.savefig('%s_%s.png'%(str(j), new_labels_[j])) plt.close(fig) os.mkdir('feature_plots_transformed') os.chdir('feature_plots_transformed') newdata=new_features newdata['classes']=class_labels for j in range(len(new_labels)): fig=sns.violinplot(x=newdata['classes'], y=newdata[new_labels[j]]) fig = fig.get_figure() fig.tight_layout() fig.savefig('%s_%s.png'%(str(j), new_labels[j])) plt.close(fig) ################################################## # PRECISION-RECALL CURVES ################################################## os.chdir(curdir) os.mkdir('model_selection') os.chdir('model_selection') plt.figure() visualizer = precision_recall_curve(GaussianNB(), np.array(features), tclass_labels) visualizer.poof(outpath="precision-recall.png") plt.close() plt.figure() visualizer = roc_auc(LogisticRegression(), np.array(features), tclass_labels) visualizer.poof(outpath="roc_curve_train.png") plt.close() plt.figure() visualizer = discrimination_threshold( LogisticRegression(multi_class="auto", solver="liblinear"), np.array(features), tclass_labels) visualizer.poof(outpath="thresholds.png") plt.close() plt.figure() visualizer = residuals_plot( Ridge(), np.array(features), tclass_labels, train_color="maroon", test_color="gold" ) visualizer.poof(outpath="residuals.png") plt.close() plt.figure() visualizer = prediction_error(Lasso(), np.array(features), tclass_labels) visualizer.poof(outpath='prediction_error.png') plt.close() # outlier detection plt.figure() visualizer = cooks_distance(np.array(features), tclass_labels, draw_threshold=True, linefmt="C0-", markerfmt=",") visualizer.poof(outpath='outliers.png') plt.close() # cluster numbers plt.figure() visualizer = silhouette_visualizer(KMeans(len(set(tclass_labels)), random_state=42), np.array(features)) visualizer.poof(outpath='siloutte.png') plt.close() # cluster distance plt.figure() visualizer = intercluster_distance(KMeans(len(set(tclass_labels)), random_state=777), np.array(features)) visualizer.poof(outpath='cluster_distance.png') plt.close() # plot percentile of features plot with SVM to see which percentile for features is optimal features=preprocessing.MinMaxScaler().fit_transform(features) clf = Pipeline([('anova', SelectPercentile(chi2)), ('scaler', StandardScaler()), ('logr', LogisticRegression())]) score_means = list() score_stds = list() percentiles = (1, 3, 6, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100) for percentile in percentiles: clf.set_params(anova__percentile=percentile) this_scores = cross_val_score(clf, np.array(features), class_labels) score_means.append(this_scores.mean()) score_stds.append(this_scores.std()) plt.errorbar(percentiles, score_means, np.array(score_stds)) plt.title('Performance of the LogisticRegression-Anova varying the percent features selected') plt.xticks(np.linspace(0, 100, 11, endpoint=True)) plt.xlabel('Percentile') plt.ylabel('Accuracy Score') plt.axis('tight') plt.savefig('logr_percentile_plot.png') plt.close() # get PCA pca = PCA(random_state=1) pca.fit(X_train) skplt.decomposition.plot_pca_component_variance(pca) plt.savefig('pca_explained_variance.png') plt.close() # estimators rf = RandomForestClassifier() skplt.estimators.plot_learning_curve(rf, X_train, y_train) plt.title('Learning Curve (Random Forest)') plt.savefig('learning_curve.png') plt.close() # elbow plot kmeans = KMeans(random_state=1) skplt.cluster.plot_elbow_curve(kmeans, X_train, cluster_ranges=range(1, 30), title='Elbow plot (KMeans clustering)') plt.savefig('elbow.png') plt.close() # KS statistic (only if 2 classes) lr = LogisticRegression() lr = lr.fit(X_train, y_train) y_probas = lr.predict_proba(X_test) skplt.metrics.plot_ks_statistic(y_test, y_probas) plt.savefig('ks.png') plt.close() # precision-recall nb = GaussianNB() nb.fit(X_train, y_train) y_probas = nb.predict_proba(X_test) skplt.metrics.plot_precision_recall(y_test, y_probas) plt.tight_layout() plt.savefig('precision-recall.png') plt.close() ## plot calibration curve rf = RandomForestClassifier() lr = LogisticRegression() nb = GaussianNB() svm = LinearSVC() dt = DecisionTreeClassifier(random_state=0) ab = AdaBoostClassifier(n_estimators=100) gb = GradientBoostingClassifier(n_estimators=100, learning_rate=1.0, max_depth=1, random_state=0) knn = KNeighborsClassifier(n_neighbors=7) rf_probas = rf.fit(X_train, y_train).predict_proba(X_test) lr_probas = lr.fit(X_train, y_train).predict_proba(X_test) nb_probas = nb.fit(X_train, y_train).predict_proba(X_test) # svm_scores = svm.fit(X_train, y_train).predict_proba(X_test) dt_scores= dt.fit(X_train, y_train).predict_proba(X_test) ab_scores= ab.fit(X_train, y_train).predict_proba(X_test) gb_scores= gb.fit(X_train, y_train).predict_proba(X_test) knn_scores= knn.fit(X_train, y_train).predict_proba(X_test) probas_list = [rf_probas, lr_probas, nb_probas, # svm_scores, dt_scores, ab_scores, gb_scores, knn_scores] clf_names = ['Random Forest', 'Logistic Regression', 'Gaussian NB', # 'SVM', 'Decision Tree', 'Adaboost', 'Gradient Boost', 'KNN'] skplt.metrics.plot_calibration_curve(y_test,probas_list, clf_names) plt.savefig('calibration.png') plt.tight_layout() plt.close() # pick classifier type by ROC (without optimization) probs = [rf_probas[:, 1], lr_probas[:, 1], nb_probas[:, 1], # svm_scores[:, 1], dt_scores[:, 1], ab_scores[:, 1], gb_scores[:, 1], knn_scores[:, 1]] plot_roc_curve(y_test, probs, clf_names) # more elaborate ROC example with CV = 5 fold # https://scikit-learn.org/stable/auto_examples/model_selection/plot_roc_crossval.html#sphx-glr-auto-examples-model-selection-plot-roc-crossval-py os.chdir(curdir) return '' def visualize_features_csv(csvdata, target, classes, problem_type, curdir, balance_data, test_size): # make features into label encoder here g=csvdata data=dict() csv_labels=g[target] csv_features=g.drop([target], axis=1) csv_feature_labels=list(csv_features) feature_labels=[csv_feature_labels] classes=list(set(list(csv_labels))) for i in range(len(classes)): class_type = classes[i] feature_list=list() label_list=list() for i in range(len(csv_features)): if csv_labels[i] == class_type: feature_list.append(list(csv_features.iloc[i,:])) label_list.append(csv_feature_labels) data[class_type]=feature_list g=data lengths=list() alldata=list() labels=list() # check to see all classes are same length and reshape if necessary for i in range(len(classes)): class_=g[classes[i]] lengths.append(len(class_)) lengths=np.array(lengths) minlength=np.amin(lengths) # now load all the classes for i in range(len(classes)): class_=g[classes[i]] random.shuffle(class_) # only balance if specified in settings if balance_data==True: if len(class_) > minlength: print('%s greater than minlength (%s) by %s, equalizing...'%(classes[i], str(minlength), str(len(class_)-minlength))) class_=class_[0:minlength] for j in range(len(class_)): alldata.append(class_[j]) labels.append(i) # process features to help with clustering se=preprocessing.StandardScaler() t_features=se.fit_transform(alldata) # make regular features features=alldata tclass_labels=labels class_labels=labels X_train, X_test, y_train, y_test = train_test_split(alldata, labels, test_size=test_size, random_state=42) # print(len(features)) # print(len(feature_labels)) # print(len(class_labels)) # print(class_labels) # GET TRAINING DATA DURING MODELING PROCESS ################################## # get filename # csvfile='' # print(classes) # for i in range(len(classes)): # csvfile=csvfile+classes[i]+'_' # get training and testing data for later # try: # print('loading training files...') # X_train=pd.read_csv(prev_dir(curdir)+'/models/'+csvfile+'train.csv') # y_train=X_train['class_'] # X_train.drop(['class_'], axis=1) # X_test=pd.read_csv(prev_dir(curdir)+'/models/'+csvfile+'test.csv') # y_test=X_test['class_'] # X_test.drop(['class_'], axis=1) # y_train=le.inverse_transform(y_train) # y_test=le.inverse_transform(y_test) # except: # print('error loading in training files, making new test data') # Visualize each class (quick plot) ################################## visualization_dir='visualization_session' try: os.mkdir(visualization_dir) os.chdir(visualization_dir) except: shutil.rmtree(visualization_dir) os.mkdir(visualization_dir) os.chdir(visualization_dir) objects = tuple(set(class_labels)) y_pos = np.arange(len(objects)) performance=list() for i in range(len(objects)): performance.append(class_labels.count(objects[i])) plt.bar(y_pos, performance, align='center', alpha=0.5) plt.xticks(y_pos, objects) plt.xticks(rotation=90) plt.title('Counts per class') plt.ylabel('Count') plt.xlabel('Class') plt.tight_layout() plt.savefig('classes.png') plt.close() # set current directory curdir=os.getcwd() # ################################## # # CLUSTERING!!! # ################################## ################################## # Manifold type options ################################## ''' "lle" Locally Linear Embedding (LLE) uses many local linear decompositions to preserve globally non-linear structures. "ltsa" LTSA LLE: local tangent space alignment is similar to LLE in that it uses locality to preserve neighborhood distances. "hessian" Hessian LLE an LLE regularization method that applies a hessian-based quadratic form at each neighborhood "modified" Modified LLE applies a regularization parameter to LLE. "isomap" Isomap seeks a lower dimensional embedding that maintains geometric distances between each instance. "mds" MDS: multi-dimensional scaling uses similarity to plot points that are near to each other close in the embedding. "spectral" Spectral Embedding a discrete approximation of the low dimensional manifold using a graph representation. "tsne" (default) t-SNE: converts the similarity of points into probabilities then uses those probabilities to create an embedding. ''' os.mkdir('clustering') os.chdir('clustering') # tSNE plt.figure() viz = Manifold(manifold="tsne", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="tsne.png") plt.close() # os.system('open tsne.png') # viz.show() # PCA plt.figure() visualizer = PCADecomposition(scale=True, classes=set(classes)) visualizer.fit_transform(np.array(features), tclass_labels) visualizer.poof(outpath="pca.png") plt.close() # os.system('open pca.png') # spectral embedding plt.figure() viz = Manifold(manifold="spectral", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="spectral.png") plt.close() # lle embedding plt.figure() viz = Manifold(manifold="lle", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="lle.png") plt.close() # ltsa # plt.figure() # viz = Manifold(manifold="ltsa", classes=set(classes)) # viz.fit_transform(np.array(features), tclass_labels) # viz.poof(outpath="ltsa.png") # plt.close() # hessian # plt.figure() # viz = Manifold(manifold="hessian", method='dense', classes=set(classes)) # viz.fit_transform(np.array(features), tclass_labels) # viz.poof(outpath="hessian.png") # plt.close() # modified plt.figure() viz = Manifold(manifold="modified", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="modified.png") plt.close() # isomap plt.figure() viz = Manifold(manifold="isomap", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="isomap.png") plt.close() # mds plt.figure() viz = Manifold(manifold="mds", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="mds.png") plt.close() # spectral plt.figure() viz = Manifold(manifold="spectral", classes=set(classes)) viz.fit_transform(np.array(features), tclass_labels) viz.poof(outpath="spectral.png") plt.close() # UMAP embedding plt.figure() umap = UMAPVisualizer(metric='cosine', classes=set(classes), title="UMAP embedding") umap.fit_transform(np.array(features), class_labels) umap.poof(outpath="umap.png") plt.close() # alternative UMAP # import umap.plot # plt.figure() # mapper = umap.UMAP().fit(np.array(features)) # fig=umap.plot.points(mapper, labels=np.array(tclass_labels)) # fig = fig.get_figure() # fig.tight_layout() # fig.savefig('umap2.png') # plt.close(fig) ################################# # FEATURE RANKING!! ################################# os.chdir(curdir) os.mkdir('feature_ranking') os.chdir('feature_ranking') # You can get the feature importance of each feature of your dataset # by using the feature importance property of the model. plt.figure(figsize=(12,12)) model = ExtraTreesClassifier() model.fit(np.array(features),tclass_labels) # print(model.feature_importances_)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
true
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/augment_vidaug.py
augmentation/video_augmentation/augment_vidaug.py
''' AAA lllllll lllllll iiii A:::A l:::::l l:::::l i::::i A:::::A l:::::l l:::::l iiii A:::::::A l:::::l l:::::l A:::::::::A l::::l l::::l iiiiiii eeeeeeeeeeee A:::::A:::::A l::::l l::::l i:::::i ee::::::::::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::eeeee:::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::e e:::::e A:::::A A:::::A l::::l l::::l i::::i e:::::::eeeee::::::e A:::::AAAAAAAAA:::::A l::::l l::::l i::::i e:::::::::::::::::e A:::::::::::::::::::::A l::::l l::::l i::::i e::::::eeeeeeeeeee A:::::AAAAAAAAAAAAA:::::A l::::l l::::l i::::i e:::::::e A:::::A A:::::A l::::::ll::::::li::::::ie::::::::e A:::::A A:::::A l::::::ll::::::li::::::i e::::::::eeeeeeee A:::::A A:::::A l::::::ll::::::li::::::i ee:::::::::::::e AAAAAAA AAAAAAAlllllllllllllllliiiiiiii eeeeeeeeeeeeee / _ \ | | | | (_) / /_\ \_ _ __ _ _ __ ___ ___ _ __ | |_ __ _| |_ _ ___ _ __ | _ | | | |/ _` | '_ ` _ \ / _ \ '_ \| __/ _` | __| |/ _ \| '_ \ | | | | |_| | (_| | | | | | | __/ | | | || (_| | |_| | (_) | | | | \_| |_/\__,_|\__, |_| |_| |_|\___|_| |_|\__\__,_|\__|_|\___/|_| |_| __/ | |___/ ___ ______ _____ _ _ _ _ / _ \ | ___ \_ _| _ | | | (_) | | / /_\ \| |_/ / | | (_) | | | |_ __| | ___ ___ | _ || __/ | | | | | | |/ _` |/ _ \/ _ \ | | | || | _| |_ _ \ \_/ / | (_| | __/ (_) | \_| |_/\_| \___/ (_) \___/|_|\__,_|\___|\___/ ''' import helpers.vidaug.vidaug.augmentors as va from PIL import Image, ImageSequence import os import moviepy.editor as mp def augment_vidaug(videofile, basedir): def gif_loader(path, modality="RGB"): frames = [] with open(path, 'rb') as f: with Image.open(f) as video: index = 1 for frame in ImageSequence.Iterator(video): frames.append(frame.convert(modality)) index += 1 return frames file = videofile # convert file to gif if file[-4:] != '.gif': # only take first 10 seconds. os.system('ffmpeg -i %s %s'%(file, file[0:-4]+'.gif')) file=file[0:-4]+'.gif' frames = gif_loader(os.getcwd()+"/%s"%(file)) sometimes = lambda aug: va.Sometimes(0.75, aug) # Used to apply augmentor with 75% probability seq = va.Sequential([ va.RandomCrop(size=(240, 180)), # randomly crop video with a size of (240 x 180) va.RandomRotate(degrees=10), # randomly rotates the video with a degree randomly choosen from [-10, 10] sometimes(va.HorizontalFlip()) # horizontally flip the video with 50% probability ]) #augment the frames video_aug = seq(frames) # save augmentad frames as gif video_aug[0].save(file[0:-4]+'.gif', save_all=True, append_images=video_aug[1:], duration=100, loop=0) clip = mp.VideoFileClip(file[0:-4]+'.gif') clip.write_videofile('augmented_'+file[0:-4]+'.mp4') os.remove(file[0:-4]+'.gif') return [videofile, 'augmented_'+file[0:-4]+'.mp4']
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/augment.py
augmentation/video_augmentation/augment.py
''' AAA lllllll lllllll iiii A:::A l:::::l l:::::l i::::i A:::::A l:::::l l:::::l iiii A:::::::A l:::::l l:::::l A:::::::::A l::::l l::::l iiiiiii eeeeeeeeeeee A:::::A:::::A l::::l l::::l i:::::i ee::::::::::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::eeeee:::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::e e:::::e A:::::A A:::::A l::::l l::::l i::::i e:::::::eeeee::::::e A:::::AAAAAAAAA:::::A l::::l l::::l i::::i e:::::::::::::::::e A:::::::::::::::::::::A l::::l l::::l i::::i e::::::eeeeeeeeeee A:::::AAAAAAAAAAAAA:::::A l::::l l::::l i::::i e:::::::e A:::::A A:::::A l::::::ll::::::li::::::ie::::::::e A:::::A A:::::A l::::::ll::::::li::::::i e::::::::eeeeeeee A:::::A A:::::A l::::::ll::::::li::::::i ee:::::::::::::e AAAAAAA AAAAAAAlllllllllllllllliiiiiiii eeeeeeeeeeeeee / _ \ | | | | (_) / /_\ \_ _ __ _ _ __ ___ ___ _ __ | |_ __ _| |_ _ ___ _ __ | _ | | | |/ _` | '_ ` _ \ / _ \ '_ \| __/ _` | __| |/ _ \| '_ \ | | | | |_| | (_| | | | | | | __/ | | | || (_| | |_| | (_) | | | | \_| |_/\__,_|\__, |_| |_| |_|\___|_| |_|\__\__,_|\__|_|\___/|_| |_| __/ | |___/ ___ ______ _____ _ _ _ _ / _ \ | ___ \_ _| _ | | | (_) | | / /_\ \| |_/ / | | (_) | | | |_ __| | ___ ___ | _ || __/ | | | | | | |/ _` |/ _ \/ _ \ | | | || | _| |_ _ \ \_/ / | (_| | __/ (_) | \_| |_/\_| \___/ (_) \___/|_|\__,_|\___|\___/ This section of Allie's API augments folders of video files using the default_video_augmenters. Usage: python3 augment.py [folder] [augment_type] All augment_type options include: ["augment_vidaug"] Read more @ https://github.com/jim-schwoebel/allie/tree/master/augmentation/video_augmentation ''' ################################################ ## IMPORT STATEMENTS ## ################################################ import json, os, sys, time, random import numpy as np # import helpers.transcribe as ts # import speech_recognition as sr from tqdm import tqdm def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ ################################################ ## Helper functions ## ################################################ def video_augment(augmentation_set, videofile, basedir): # only load the relevant featuresets for featurization to save memory if augmentation_set=='augment_vidaug': augment_vidaug.augment_vidaug(videofile, basedir) ################################################ ## Load main settings ## ################################################ # directory=sys.argv[1] basedir=os.getcwd() settingsdir=prev_dir(basedir) settingsdir=prev_dir(settingsdir) settings=json.load(open(settingsdir+'/settings.json')) os.chdir(basedir) video_transcribe=settings['transcribe_video'] default_video_transcribers=settings['default_video_transcriber'] try: # assume 1 type of feature_set augmentation_sets=[sys.argv[2]] except: # if none provided in command line, then load deafult features augmentation_sets=settings['default_video_augmenters'] ################################################ ## Import According to settings ## ################################################ # only load the relevant featuresets for featurization to save memory if 'augment_vidaug' in augmentation_sets: import augment_vidaug ################################################ ## Get featurization folder ## ################################################ foldername=sys.argv[1] os.chdir(foldername) listdir=os.listdir() random.shuffle(listdir) cur_dir=os.getcwd() help_dir=basedir+'/helpers/' # get class label from folder name labelname=foldername.split('/') if labelname[-1]=='': labelname=labelname[-2] else: labelname=labelname[-1] ################################################ ## NOW AUGMENT!! ## ################################################ listdir=os.listdir() random.shuffle(listdir) # featurize all files accoridng to librosa featurize for i in tqdm(range(len(listdir)), desc=labelname): if listdir[i][-4:] in ['.mp4']: filename=[listdir[i]] for j in range(len(augmentation_sets)): augmentation_set=augmentation_sets[j] for k in range(len(filename)): filename=video_augment(augmentation_set, filename[k], basedir)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/setup.py
augmentation/video_augmentation/helpers/vidaug/setup.py
import setuptools setuptools.setup(name='vidaug', version='0.1', description='Video Augmentation Library', url='https://github.com/okankop/vidaug', author='Okan Kopuklu', author_email='okankopuklu@gmail.com', license='MIT', packages=setuptools.find_packages(), zip_safe=False)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/__init__.py
augmentation/video_augmentation/helpers/vidaug/vidaug/__init__.py
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/flip.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/flip.py
""" Augmenters that apply video flipping horizontally and vertically. To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.HorizontalFlip(), va.VerticalFlip() ]) List of augmenters: * HorizontalFlip * VerticalFlip """ import numpy as np import PIL class HorizontalFlip(object): """ Horizontally flip the video. """ def __call__(self, clip): if isinstance(clip[0], np.ndarray): return [np.fliplr(img) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.transpose(PIL.Image.FLIP_LEFT_RIGHT) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + ' but got list of {0}'.format(type(clip[0]))) class VerticalFlip(object): """ Vertically flip the video. """ def __call__(self, clip): if isinstance(clip[0], np.ndarray): return [np.flipud(img) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.transpose(PIL.Image.FLIP_TOP_BOTTOM) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + ' but got list of {0}'.format(type(clip[0])))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/crop.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/crop.py
""" Augmenters that apply video flipping horizontally and vertically. To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.HorizontalFlip(), va.VerticalFlip() ]) List of augmenters: * CenterCrop * CornerCrop * RandomCrop """ import numpy as np import PIL import numbers import random class CenterCrop(object): """ Extract center crop of thevideo. Args: size (sequence or int): Desired output size for the crop in format (h, w). """ def __init__(self, size): if isinstance(size, numbers.Number): if size < 0: raise ValueError('If size is a single number, it must be positive') size = (size, size) else: if len(size) != 2: raise ValueError('If size is a sequence, it must be of len 2.') self.size = size def __call__(self, clip): crop_h, crop_w = self.size if isinstance(clip[0], np.ndarray): im_h, im_w, im_c = clip[0].shape elif isinstance(clip[0], PIL.Image.Image): im_w, im_h = clip[0].size else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) if crop_w > im_w or crop_h > im_h: error_msg = ('Initial image size should be larger then' + 'cropped size but got cropped sizes : ' + '({w}, {h}) while initial image is ({im_w}, ' + '{im_h})'.format(im_w=im_w, im_h=im_h, w=crop_w, h=crop_h)) raise ValueError(error_msg) w1 = int(round((im_w - crop_w) / 2.)) h1 = int(round((im_h - crop_h) / 2.)) if isinstance(clip[0], np.ndarray): return [img[h1:h1 + crop_h, w1:w1 + crop_w, :] for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.crop((w1, h1, w1 + crop_w, h1 + crop_h)) for img in clip] class CornerCrop(object): """ Extract corner crop of the video. Args: size (sequence or int): Desired output size for the crop in format (h, w). crop_position (str): Selected corner (or center) position from the list ['c', 'tl', 'tr', 'bl', 'br']. If it is non, crop position is selected randomly at each call. """ def __init__(self, size, crop_position=None): if isinstance(size, numbers.Number): if size < 0: raise ValueError('If size is a single number, it must be positive') size = (size, size) else: if len(size) != 2: raise ValueError('If size is a sequence, it must be of len 2.') self.size = size if crop_position is None: self.randomize = True else: if crop_position not in ['c', 'tl', 'tr', 'bl', 'br']: raise ValueError("crop_position should be one of " + "['c', 'tl', 'tr', 'bl', 'br']") self.randomize = False self.crop_position = crop_position self.crop_positions = ['c', 'tl', 'tr', 'bl', 'br'] def __call__(self, clip): crop_h, crop_w = self.size if isinstance(clip[0], np.ndarray): im_h, im_w, im_c = clip[0].shape elif isinstance(clip[0], PIL.Image.Image): im_w, im_h = clip[0].size else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) if self.randomize: self.crop_position = self.crop_positions[random.randint(0,len(self.crop_positions) - 1)] if self.crop_position == 'c': th, tw = (self.size, self.size) x1 = int(round((im_w - crop_w) / 2.)) y1 = int(round((im_h - crop_h) / 2.)) x2 = x1 + crop_w y2 = y1 + crop_h elif self.crop_position == 'tl': x1 = 0 y1 = 0 x2 = crop_w y2 = crop_h elif self.crop_position == 'tr': x1 = im_w - crop_w y1 = 0 x2 = im_w y2 = crop_h elif self.crop_position == 'bl': x1 = 0 y1 = im_h - crop_h x2 = crop_w y2 = im_h elif self.crop_position == 'br': x1 = im_w - crop_w y1 = im_h - crop_h x2 = im_w y2 = im_h if isinstance(clip[0], np.ndarray): return [img[y1:y2, x1:x2, :] for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.crop((x1, y1, x2, y2)) for img in clip] class RandomCrop(object): """ Extract random crop of the video. Args: size (sequence or int): Desired output size for the crop in format (h, w). crop_position (str): Selected corner (or center) position from the list ['c', 'tl', 'tr', 'bl', 'br']. If it is non, crop position is selected randomly at each call. """ def __init__(self, size): if isinstance(size, numbers.Number): if size < 0: raise ValueError('If size is a single number, it must be positive') size = (size, size) else: if len(size) != 2: raise ValueError('If size is a sequence, it must be of len 2.') self.size = size def __call__(self, clip): crop_h, crop_w = self.size if isinstance(clip[0], np.ndarray): im_h, im_w, im_c = clip[0].shape elif isinstance(clip[0], PIL.Image.Image): im_w, im_h = clip[0].size else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) if crop_w > im_w or crop_h > im_h: error_msg = ('Initial image size should be larger then' + 'cropped size but got cropped sizes : ' + '({w}, {h}) while initial image is ({im_w}, ' + '{im_h})'.format(im_w=im_w, im_h=im_h, w=crop_w, h=crop_h)) raise ValueError(error_msg) w1 = random.randint(0, im_w - crop_w) h1 = random.randint(0, im_h - crop_h) if isinstance(clip[0], np.ndarray): return [img[h1:h1 + crop_h, w1:w1 + crop_w, :] for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.crop((w1, h1, w1 + crop_w, h1 + crop_h)) for img in clip]
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/geometric.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/geometric.py
""" Augmenters that apply geometric transformations. To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.RandomRotate(30), va.RandomResize(0.2) ]) List of augmenters: * GaussianBlur * ElasticTransformation * PiecewiseAffineTransform * Superpixel """ from skimage import segmentation, measure import numpy as np import random import numbers import scipy import PIL import cv2 class GaussianBlur(object): """ Augmenter to blur images using gaussian kernels. Args: sigma (float): Standard deviation of the gaussian kernel. """ def __init__(self, sigma): self.sigma = sigma def __call__(self, clip): if isinstance(clip[0], np.ndarray): return [scipy.ndimage.gaussian_filter(img, sigma=self.sigma, order=0) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.filter(PIL.ImageFilter.GaussianBlur(radius=self.sigma)) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) class ElasticTransformation(object): """ Augmenter to transform images by moving pixels locally around using displacement fields. See Simard, Steinkraus and Platt Best Practices for Convolutional Neural Networks applied to Visual Document Analysis in Proc. of the International Conference on Document Analysis and Recognition, 2003 for a detailed explanation. Args: alpha (float): Strength of the distortion field. Higher values mean more "movement" of pixels. sigma (float): Standard deviation of the gaussian kernel used to smooth the distortion fields. order (int): Interpolation order to use. Same meaning as in `scipy.ndimage.map_coordinates` and may take any integer value in the range 0 to 5, where orders close to 0 are faster. cval (int): The constant intensity value used to fill in new pixels. This value is only used if `mode` is set to "constant". For standard uint8 images (value range 0-255), this value may also come from the range 0-255. It may be a float value, even for integer image dtypes. mode : Parameter that defines the handling of newly created pixels. May take the same values as in `scipy.ndimage.map_coordinates`, i.e. "constant", "nearest", "reflect" or "wrap". """ def __init__(self, alpha=0, sigma=0, order=3, cval=0, mode="constant", name=None, deterministic=False): self.alpha = alpha self.sigma = sigma self.order = order self.cval = cval self.mode = mode def __call__(self, clip): is_PIL = isinstance(clip[0], PIL.Image.Image) if is_PIL: clip = [np.asarray(img) for img in clip] result = [] nb_images = len(clip) for i in range(nb_images): image = clip[i] image_first_channel = np.squeeze(image[..., 0]) indices_x, indices_y = self._generate_indices(image_first_channel.shape, alpha=self.alpha, sigma=self.sigma) result.append(self._map_coordinates( clip[i], indices_x, indices_y, order=self.order, cval=self.cval, mode=self.mode)) if is_PIL: return [PIL.Image.fromarray(img) for img in result] else: return result def _generate_indices(self, shape, alpha, sigma): assert (len(shape) == 2),"shape: Should be of size 2!" dx = scipy.ndimage.gaussian_filter((np.random.rand(*shape) * 2 - 1), sigma, mode="constant", cval=0) * alpha dy = scipy.ndimage.gaussian_filter((np.random.rand(*shape) * 2 - 1), sigma, mode="constant", cval=0) * alpha x, y = np.meshgrid(np.arange(shape[0]), np.arange(shape[1]), indexing='ij') return np.reshape(x+dx, (-1, 1)), np.reshape(y+dy, (-1, 1)) def _map_coordinates(self, image, indices_x, indices_y, order=1, cval=0, mode="constant"): assert (len(image.shape) == 3),"image.shape: Should be of size 3!" result = np.copy(image) height, width = image.shape[0:2] for c in range(image.shape[2]): remapped_flat = scipy.ndimage.interpolation.map_coordinates( image[..., c], (indices_x, indices_y), order=order, cval=cval, mode=mode ) remapped = remapped_flat.reshape((height, width)) result[..., c] = remapped return result class PiecewiseAffineTransform(object): """ Augmenter that places a regular grid of points on an image and randomly moves the neighbourhood of these point around via affine transformations. Args: displacement (init): gives distorted image depending on the valuse of displacement_magnification and displacement_kernel displacement_kernel (init): gives the blury effect displacement_magnification (float): it magnify the image """ def __init__(self, displacement=0, displacement_kernel=0, displacement_magnification=0): self.displacement = displacement self.displacement_kernel = displacement_kernel self.displacement_magnification = displacement_magnification def __call__(self, clip): ret_img_group = clip if isinstance(clip[0], np.ndarray): im_size = clip[0].shape image_w, image_h = im_size[1], im_size[0] elif isinstance(clip[0], PIL.Image.Image): im_size = clip[0].size image_w, image_h = im_size[0], im_size[1] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) displacement_map = np.random.rand(image_h, image_w, 2) * 2 * self.displacement - self.displacement displacement_map = cv2.GaussianBlur(displacement_map, None, self.displacement_kernel) displacement_map *= self.displacement_magnification * self.displacement_kernel displacement_map = np.floor(displacement_map).astype('int32') displacement_map_rows = displacement_map[..., 0] + np.tile(np.arange(image_h), (image_w, 1)).T.astype('int32') displacement_map_rows = np.clip(displacement_map_rows, 0, image_h - 1) displacement_map_cols = displacement_map[..., 1] + np.tile(np.arange(image_w), (image_h, 1)).astype('int32') displacement_map_cols = np.clip(displacement_map_cols, 0, image_w - 1) if isinstance(clip[0], np.ndarray): return [img[(displacement_map_rows.flatten(), displacement_map_cols.flatten())].reshape(img.shape) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [PIL.Image.fromarray(np.asarray(img)[(displacement_map_rows.flatten(), displacement_map_cols.flatten())].reshape(np.asarray(img).shape)) for img in clip] class Superpixel(object): """ Completely or partially transform images to their superpixel representation. Args: p_replace (int) : Defines the probability of any superpixel area being replaced by the superpixel. n_segments (int): Target number of superpixels to generate. Lower numbers are faster. interpolation (str): Interpolation to use. Can be one of 'nearest', 'bilinear' defaults to nearest """ def __init__(self, p_replace=0, n_segments=0, max_size=360, interpolation="bilinear"): self.p_replace = p_replace self.n_segments = n_segments self.interpolation = interpolation def __call__(self, clip): is_PIL = isinstance(clip[0], PIL.Image.Image) if is_PIL: clip = [np.asarray(img) for img in clip] # TODO this results in an error when n_segments is 0 replace_samples = np.tile(np.array([self.p_replace]), self.n_segments) avg_image = np.mean(clip, axis=0) segments = segmentation.slic(avg_image, n_segments=self.n_segments, compactness=10) if not np.max(replace_samples) == 0: print("Converting") clip = [self._apply_segmentation(img, replace_samples, segments) for img in clip] if is_PIL: return [PIL.Image.fromarray(img) for img in clip] else: return clip def _apply_segmentation(self, image, replace_samples, segments): nb_channels = image.shape[2] image_sp = np.copy(image) for c in range(nb_channels): # segments+1 here because otherwise regionprops always misses # the last label regions = measure.regionprops(segments + 1, intensity_image=image[..., c]) for ridx, region in enumerate(regions): # with mod here, because slic can sometimes create more # superpixel than requested. replace_samples then does # not have enough values, so we just start over with the # first one again. if replace_samples[ridx % len(replace_samples)] == 1: mean_intensity = region.mean_intensity image_sp_c = image_sp[..., c] image_sp_c[segments == ridx] = mean_intensity return image_sp
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/temporal.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/temporal.py
""" Augmenters that apply temporal transformations. To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.RandomRotate(30), va.RandomResize(0.2) ]) List of augmenters: * TemporalBeginCrop * TemporalCenterCrop * TemporalRandomCrop * InverseOrder * Downsample * Upsample * TemporalFit * TemporalElasticTransformation """ import numpy as np import PIL import random import math class TemporalBeginCrop(object): """ Temporally crop the given frame indices at a beginning. If the number of frames is less than the size, loop the indices as many times as necessary to satisfy the size. Args: size (int): Desired output size of the crop. """ def __init__(self, size): self.size = size def __call__(self, clip): out = clip[:self.size] for img in out: if len(out) >= self.size: break out.append(img) return out class TemporalCenterCrop(object): """ Temporally crop the given frame indices at a center. If the number of frames is less than the size, loop the indices as many times as necessary to satisfy the size. Args: size (int): Desired output size of the crop. """ def __init__(self, size): self.size = size def __call__(self, clip): center_index = len(clip) // 2 begin_index = max(0, center_index - (self.size // 2)) end_index = min(begin_index + self.size, len(clip)) out = clip[begin_index:end_index] for img in out: if len(out) >= self.size: break out.append(img) return out class TemporalRandomCrop(object): """ Temporally crop the given frame indices at a random location. If the number of frames is less than the size, loop the indices as many times as necessary to satisfy the size. Args: size (int): Desired output size of the crop. """ def __init__(self, size): self.size = size def __call__(self, clip): rand_end = max(0, len(clip) - self.size - 1) begin_index = random.randint(0, rand_end) end_index = min(begin_index + self.size, len(clip)) out = clip[begin_index:end_index] for img in out: if len(out) >= self.size: break out.append(img) return out class InverseOrder(object): """ Inverts the order of clip frames. """ def __call__(self, clip): for i in range(len(clip)): nb_images = len(clip) return [clip[img] for img in reversed(range(1, nb_images))] class Downsample(object): """ Temporally downsample a video by deleting some of its frames. Args: ratio (float): Downsampling ratio in [0.0 <= ratio <= 1.0]. """ def __init__(self , ratio=1.0): if ratio < 0.0 or ratio > 1.0: raise TypeError('ratio should be in [0.0 <= ratio <= 1.0]. ' + 'Please use upsampling for ratio > 1.0') self.ratio = ratio def __call__(self, clip): nb_return_frame = np.floor(self.ratio * len(clip)) return_ind = [int(i) for i in np.linspace(1, len(clip), num=nb_return_frame)] return [clip[i-1] for i in return_ind] class Upsample(object): """ Temporally upsampling a video by deleting some of its frames. Args: ratio (float): Upsampling ratio in [1.0 < ratio < infinity]. """ def __init__(self , ratio=1.0): if ratio < 1.0: raise TypeError('ratio should be 1.0 < ratio. ' + 'Please use downsampling for ratio <= 1.0') self.ratio = ratio def __call__(self, clip): nb_return_frame = np.floor(self.ratio * len(clip)) return_ind = [int(i) for i in np.linspace(1, len(clip), num=nb_return_frame)] return [clip[i-1] for i in return_ind] class TemporalFit(object): """ Temporally fits a video to a given frame size by downsampling or upsampling. Args: size (int): Frame size to fit the video. """ def __init__(self, size): if size < 0: raise TypeError('size should be positive') self.size = size def __call__(self, clip): return_ind = [int(i) for i in np.linspace(1, len(clip), num=self.size)] return [clip[i-1] for i in return_ind] class TemporalElasticTransformation(object): """ Stretches or schrinks a video at the beginning, end or middle parts. In normal operation, augmenter stretches the beggining and end, schrinks the center. In inverse operation, augmenter shrinks the beggining and end, stretches the center. """ def __call__(self, clip): nb_images = len(clip) new_indices = self._get_distorted_indices(nb_images) return [clip[i] for i in new_indices] def _get_distorted_indices(self, nb_images): inverse = random.randint(0, 1) if inverse: scale = random.random() scale *= 0.21 scale += 0.6 else: scale = random.random() scale *= 0.6 scale += 0.8 frames_per_clip = nb_images indices = np.linspace(-scale, scale, frames_per_clip).tolist() if inverse: values = [math.atanh(x) for x in indices] else: values = [math.tanh(x) for x in indices] values = [x / values[-1] for x in values] values = [int(round(((x + 1) / 2) * (frames_per_clip - 1), 0)) for x in values] return values
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/group.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/group.py
""" Augmenters that apply to a group of augmentations, like selecting an augmentation from a list, or applying all the augmentations in a list sequentially To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.HorizontalFlip(), va.VerticalFlip() ]) List of augmenters: * Sequential * OneOf * SomeOf * Sometimes """ import numpy as np import PIL import random class Sequential(object): """ Composes several augmentations together. Args: transforms (list of "Augmentor" objects): The list of augmentations to compose. random_order (bool): Whether to apply the augmentations in random order. """ def __init__(self, transforms, random_order=False): self.transforms = transforms self.rand = random_order def __call__(self, clip): if self.rand: rand_transforms = self.transforms[:] random.shuffle(rand_transforms) for t in rand_transforms: clip = t(clip) else: for t in self.transforms: clip = t(clip) return clip class OneOf(object): """ Selects one augmentation from a list. Args: transforms (list of "Augmentor" objects): The list of augmentations to compose. """ def __init__(self, transforms): self.transforms = transforms def __call__(self, clip): select = random.choice(self.transforms) clip = select(clip) return clip class SomeOf(object): """ Selects a given number of augmentation from a list. Args: transforms (list of "Augmentor" objects): The list of augmentations. N (int): The number of augmentations to select from the list. random_order (bool): Whether to apply the augmentations in random order. """ def __init__(self, transforms, N, random_order=True): self.transforms = transforms self.rand = random_order if N > len(transforms): raise TypeError('The number of applied augmentors should be smaller than the given augmentation number') else: self.N = N def __call__(self, clip): if self.rand: tmp = self.transforms[:] selected_trans = [tmp.pop(random.randrange(len(tmp))) for _ in range(self.N)] for t in selected_trans: clip = t(clip) return clip else: indices = [i for i in range(len(self.transforms))] selected_indices = [indices.pop(random.randrange(len(indices))) for _ in range(self.N)] selected_indices.sort() selected_trans = [self.transforms[i] for i in selected_indices] for t in selected_trans: clip = t(clip) return clip class Sometimes(object): """ Applies an augmentation with a given probability. Args: p (float): The probability to apply the augmentation. transform (an "Augmentor" object): The augmentation to apply. Example: Use this this transform as follows: sometimes = lambda aug: va.Sometimes(0.5, aug) sometimes(va.HorizontalFlip) """ def __init__(self, p, transform): self.transform = transform if (p > 1.0) | (p < 0.0): raise TypeError('Expected p to be in [0.0 <= 1.0], ' + 'but got p = {0}'.format(p)) else: self.p = p def __call__(self, clip): if random.random() < self.p: clip = self.transform(clip) return clip
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/__init__.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/__init__.py
from __future__ import absolute_import from .affine import * from .crop import * from .flip import * from .group import * from .temporal import * from .intensity import * from .geometric import *
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/affine.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/affine.py
""" Augmenters that apply affine transformations. To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.RandomRotate(30), va.RandomResize(0.2) ]) List of augmenters: * RandomRotate * RandomResize * RandomTranslate * RandomShear """ import numpy as np import numbers import random import scipy import PIL import cv2 class RandomRotate(object): """ Rotate video randomly by a random angle within given boundsi. Args: degrees (sequence or int): Range of degrees to randomly select from. If degrees is a number instead of sequence like (min, max), the range of degrees, will be (-degrees, +degrees). """ def __init__(self, degrees): if isinstance(degrees, numbers.Number): if degrees < 0: raise ValueError('If degrees is a single number,' 'must be positive') degrees = (-degrees, degrees) else: if len(degrees) != 2: raise ValueError('If degrees is a sequence,' 'it must be of len 2.') self.degrees = degrees def __call__(self, clip): angle = random.uniform(self.degrees[0], self.degrees[1]) if isinstance(clip[0], np.ndarray): rotated = [scipy.misc.imrotate(img, angle) for img in clip] elif isinstance(clip[0], PIL.Image.Image): rotated = [img.rotate(angle) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) return rotated class RandomResize(object): """ Resize video bysoomingin and out. Args: rate (float): Video is scaled uniformly between [1 - rate, 1 + rate]. interp (string): Interpolation to use for re-sizing ('nearest', 'lanczos', 'bilinear', 'bicubic' or 'cubic'). """ def __init__(self, rate=0.0, interp='bilinear'): self.rate = rate self.interpolation = interp def __call__(self, clip): scaling_factor = random.uniform(1 - self.rate, 1 + self.rate) if isinstance(clip[0], np.ndarray): im_h, im_w, im_c = clip[0].shape elif isinstance(clip[0], PIL.Image.Image): im_w, im_h = clip[0].size new_w = int(im_w * scaling_factor) new_h = int(im_h * scaling_factor) new_size = (new_h, new_w) if isinstance(clip[0], np.ndarray): return [scipy.misc.imresize(img, size=(new_h, new_w),interp=self.interpolation) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.resize(size=(new_w, new_h), resample=self._get_PIL_interp(self.interpolation)) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) def _get_PIL_interp(self, interp): if interp == 'nearest': return PIL.Image.NEAREST elif interp == 'lanczos': return PIL.Image.LANCZOS elif interp == 'bilinear': return PIL.Image.BILINEAR elif interp == 'bicubic': return PIL.Image.BICUBIC elif interp == 'cubic': return PIL.Image.CUBIC class RandomTranslate(object): """ Shifting video in X and Y coordinates. Args: x (int) : Translate in x direction, selected randomly from [-x, +x] pixels. y (int) : Translate in y direction, selected randomly from [-y, +y] pixels. """ def __init__(self, x=0, y=0): self.x = x self.y = y def __call__(self, clip): x_move = random.randint(-self.x, +self.x) y_move = random.randint(-self.y, +self.y) if isinstance(clip[0], np.ndarray): rows, cols, ch = clip[0].shape transform_mat = np.float32([[1, 0, x_move], [0, 1, y_move]]) return [cv2.warpAffine(img, transform_mat, (cols, rows)) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.transform(img.size, PIL.Image.AFFINE, (1, 0, x_move, 0, 1, y_move)) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) class RandomShear(object): """ Shearing video in X and Y directions. Args: x (int) : Shear in x direction, selected randomly from [-x, +x]. y (int) : Shear in y direction, selected randomly from [-y, +y]. """ def __init__(self, x, y): self.x = x self.y = y def __call__(self, clip): x_shear = random.uniform(-self.x, self.x) y_shear = random.uniform(-self.y, self.y) if isinstance(clip[0], np.ndarray): rows, cols, ch = clip[0].shape transform_mat = np.float32([[1, x_shear, 0], [y_shear, 1, 0]]) return [cv2.warpAffine(img, transform_mat, (cols, rows)) for img in clip] elif isinstance(clip[0], PIL.Image.Image): return [img.transform(img.size, PIL.Image.AFFINE, (1, x_shear, 0, y_shear, 1, 0)) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0])))
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/intensity.py
augmentation/video_augmentation/helpers/vidaug/vidaug/augmentors/intensity.py
""" Augmenters that apply transformations on the pixel intensities. To use the augmenters, clone the complete repo and use `from vidaug import augmenters as va` and then e.g. : seq = va.Sequential([ va.RandomRotate(30), va.RandomResize(0.2) ]) List of augmenters: * InvertColor * Add * Multiply * Pepper * Salt """ import numpy as np import random import PIL from PIL import ImageOps class InvertColor(object): """ Inverts the color of the video. """ def __call__(self, clip): if isinstance(clip[0], np.ndarray): return [np.invert(img) for img in clip] elif isinstance(clip[0], PIL.Image.Image): inverted = [ImageOps.invert(img) for img in clip] else: raise TypeError('Expected numpy.ndarray or PIL.Image' + 'but got list of {0}'.format(type(clip[0]))) return inverted class Add(object): """ Add a value to all pixel intesities in an video. Args: value (int): The value to be added to pixel intesities. """ def __init__(self, value=0): if value > 255 or value < -255: raise TypeError('The video is blacked or whitened out since ' + 'value > 255 or value < -255.') self.value = value def __call__(self, clip): is_PIL = isinstance(clip[0], PIL.Image.Image) if is_PIL: clip = [np.asarray(img) for img in clip] data_final = [] for i in range(len(clip)): image = clip[i].astype(np.int32) image += self.value image = np.where(image > 255, 255, image) image = np.where(image < 0, 0, image) image = image.astype(np.uint8) data_final.append(image.astype(np.uint8)) if is_PIL: return [PIL.Image.fromarray(img) for img in data_final] else: return data_final class Multiply(object): """ Multiply all pixel intensities with given value. This augmenter can be used to make images lighter or darker. Args: value (float): The value with which to multiply the pixel intensities of video. """ def __init__(self, value=1.0): if value < 0.0: raise TypeError('The video is blacked out since for value < 0.0') self.value = value def __call__(self, clip): is_PIL = isinstance(clip[0], PIL.Image.Image) if is_PIL: clip = [np.asarray(img) for img in clip] data_final = [] for i in range(len(clip)): image = clip[i].astype(np.float64) image *= self.value image = np.where(image > 255, 255, image) image = np.where(image < 0, 0, image) image = image.astype(np.uint8) data_final.append(image.astype(np.uint8)) if is_PIL: return [PIL.Image.fromarray(img) for img in data_final] else: return data_final class Pepper(object): """ Augmenter that sets a certain fraction of pixel intensities to 0, hence they become black. Args: ratio (int): Determines number of black pixels on each frame of video. Smaller the ratio, higher the number of black pixels. """ def __init__(self, ratio=100): self.ratio = ratio def __call__(self, clip): is_PIL = isinstance(clip[0], PIL.Image.Image) if is_PIL: clip = [np.asarray(img) for img in clip] data_final = [] for i in range(len(clip)): img = clip[i].astype(np.float) img_shape = img.shape noise = np.random.randint(self.ratio, size=img_shape) img = np.where(noise == 0, 0, img) data_final.append(img.astype(np.uint8)) if is_PIL: return [PIL.Image.fromarray(img) for img in data_final] else: return data_final class Salt(object): """ Augmenter that sets a certain fraction of pixel intesities to 255, hence they become white. Args: ratio (int): Determines number of white pixels on each frame of video. Smaller the ratio, higher the number of white pixels. """ def __init__(self, ratio=100): self.ratio = ratio def __call__(self, clip): is_PIL = isinstance(clip[0], PIL.Image.Image) if is_PIL: clip = [np.asarray(img) for img in clip] data_final = [] for i in range(len(clip)): img = clip[i].astype(np.float) img_shape = img.shape noise = np.random.randint(self.ratio, size=img_shape) img = np.where(noise == 0, 255, img) data_final.append(img.astype(np.uint8)) if is_PIL: return [PIL.Image.fromarray(img) for img in data_final] else: return data_final
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/text_augmentation/augment_textacy.py
augmentation/text_augmentation/augment_textacy.py
''' AAA lllllll lllllll iiii A:::A l:::::l l:::::l i::::i A:::::A l:::::l l:::::l iiii A:::::::A l:::::l l:::::l A:::::::::A l::::l l::::l iiiiiii eeeeeeeeeeee A:::::A:::::A l::::l l::::l i:::::i ee::::::::::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::eeeee:::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::e e:::::e A:::::A A:::::A l::::l l::::l i::::i e:::::::eeeee::::::e A:::::AAAAAAAAA:::::A l::::l l::::l i::::i e:::::::::::::::::e A:::::::::::::::::::::A l::::l l::::l i::::i e::::::eeeeeeeeeee A:::::AAAAAAAAAAAAA:::::A l::::l l::::l i::::i e:::::::e A:::::A A:::::A l::::::ll::::::li::::::ie::::::::e A:::::A A:::::A l::::::ll::::::li::::::i e::::::::eeeeeeee A:::::A A:::::A l::::::ll::::::li::::::i ee:::::::::::::e AAAAAAA AAAAAAAlllllllllllllllliiiiiiii eeeeeeeeeeeeee / _ \ | | | | (_) / /_\ \_ _ __ _ _ __ ___ ___ _ __ | |_ __ _| |_ _ ___ _ __ | _ | | | |/ _` | '_ ` _ \ / _ \ '_ \| __/ _` | __| |/ _ \| '_ \ | | | | |_| | (_| | | | | | | __/ | | | || (_| | |_| | (_) | | | | \_| |_/\__,_|\__, |_| |_| |_|\___|_| |_|\__\__,_|\__|_|\___/|_| |_| __/ | |___/ ___ ______ _____ _____ _ / _ \ | ___ \_ _| _ |_ _| | | / /_\ \| |_/ / | | (_) | | _____ _| |_ | _ || __/ | | | |/ _ \ \/ / __| | | | || | _| |_ _ | | __/> <| |_ \_| |_/\_| \___/ (_) \_/\___/_/\_\\__| ''' import os, sys, shutil try: import textacy except: os.system('pip3 install textacy==0.8.0') import textacy import textacy.augmentation.transforms as transforms def augment_textacy(textfile, basedir): filename1=textfile text = open(textfile).read() # "The quick brown fox jumps over the lazy dog." doc = textacy.make_spacy_doc(text, lang="en") tfs = [transforms.substitute_word_synonyms, transforms.delete_words, transforms.swap_chars, transforms.delete_chars] augmenter=textacy.augmentation.augmenter.Augmenter(tfs, num=[0.5, 0.5, 0.5, 0.5]) augmented_text=augmenter.apply_transforms(doc) filename2='augmented_'+textfile textfile=open(filename2,'w') textfile.write(str(augmented_text)) textfile.close() return [filename1, filename2]
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/text_augmentation/augment.py
augmentation/text_augmentation/augment.py
''' AAA lllllll lllllll iiii A:::A l:::::l l:::::l i::::i A:::::A l:::::l l:::::l iiii A:::::::A l:::::l l:::::l A:::::::::A l::::l l::::l iiiiiii eeeeeeeeeeee A:::::A:::::A l::::l l::::l i:::::i ee::::::::::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::eeeee:::::ee A:::::A A:::::A l::::l l::::l i::::i e::::::e e:::::e A:::::A A:::::A l::::l l::::l i::::i e:::::::eeeee::::::e A:::::AAAAAAAAA:::::A l::::l l::::l i::::i e:::::::::::::::::e A:::::::::::::::::::::A l::::l l::::l i::::i e::::::eeeeeeeeeee A:::::AAAAAAAAAAAAA:::::A l::::l l::::l i::::i e:::::::e A:::::A A:::::A l::::::ll::::::li::::::ie::::::::e A:::::A A:::::A l::::::ll::::::li::::::i e::::::::eeeeeeee A:::::A A:::::A l::::::ll::::::li::::::i ee:::::::::::::e AAAAAAA AAAAAAAlllllllllllllllliiiiiiii eeeeeeeeeeeeee / _ \ | | | | (_) / /_\ \_ _ __ _ _ __ ___ ___ _ __ | |_ __ _| |_ _ ___ _ __ | _ | | | |/ _` | '_ ` _ \ / _ \ '_ \| __/ _` | __| |/ _ \| '_ \ | | | | |_| | (_| | | | | | | __/ | | | || (_| | |_| | (_) | | | | \_| |_/\__,_|\__, |_| |_| |_|\___|_| |_|\__\__,_|\__|_|\___/|_| |_| __/ | |___/ ___ ______ _____ _____ _ / _ \ | ___ \_ _| _ |_ _| | | / /_\ \| |_/ / | | (_) | | _____ _| |_ | _ || __/ | | | |/ _ \ \/ / __| | | | || | _| |_ _ | | __/> <| |_ \_| |_/\_| \___/ (_) \_/\___/_/\_\\__| This section of Allie's API augments folders of .TXT files using the default_text_augmenters. Usage: python3 augment.py [folder] [augment_type] All augment_type options include: ["augment_textacy", "augment_summary"] Read more @ https://github.com/jim-schwoebel/allie/tree/master/augmentation/text_augmentation ''' ################################################ ## IMPORT STATEMENTS ## ################################################ import json, os, sys, time, random import numpy as np # import helpers.transcribe as ts # import speech_recognition as sr from tqdm import tqdm def prev_dir(directory): g=directory.split('/') dir_='' for i in range(len(g)): if i != len(g)-1: if i==0: dir_=dir_+g[i] else: dir_=dir_+'/'+g[i] # print(dir_) return dir_ ################################################ ## Helper functions ## ################################################ def text_augment(augmentation_set, textfile, basedir): # only load the relevant featuresets for featurization to save memory if augmentation_set=='augment_eda': augment_eda.augment_eda(textfile, basedir) elif augmentation_set=='augment_textacy': augment_textacy.augment_textacy(textfile,basedir) ################################################ ## Load main settings ## ################################################ # directory=sys.argv[1] basedir=os.getcwd() settingsdir=prev_dir(basedir) settingsdir=prev_dir(settingsdir) settings=json.load(open(settingsdir+'/settings.json')) os.chdir(basedir) text_transcribe=settings['transcribe_text'] default_image_transcribers=settings['default_text_transcriber'] try: # assume 1 type of feature_set augmentation_sets=[sys.argv[2]] except: # if none provided in command line, then load deafult features augmentation_sets=settings['default_text_augmenters'] ################################################ ## Import According to settings ## ################################################ # only load the relevant featuresets for featurization to save memory if 'augment_eda' in augmentation_sets: import augment_eda if 'augment_textacy' in augmentation_sets: import augment_textacy ################################################ ## Get featurization folder ## ################################################ foldername=sys.argv[1] os.chdir(foldername) listdir=os.listdir() random.shuffle(listdir) cur_dir=os.getcwd() help_dir=basedir+'/helpers/' # get class label from folder name labelname=foldername.split('/') if labelname[-1]=='': labelname=labelname[-2] else: labelname=labelname[-1] ################################################ ## NOW AUGMENT!! ## ################################################ listdir=os.listdir() random.shuffle(listdir) # featurize all files accoridng to librosa featurize for i in tqdm(range(len(listdir)), desc=labelname): if listdir[i][-4:] in ['.txt']: filename=[listdir[i]] for j in range(len(augmentation_sets)): augmentation_set=augmentation_sets[j] for k in range(len(filename)): filename=text_augment(augmentation_set, filename[k], basedir)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/text_augmentation/helpers/augment_eda.py
augmentation/text_augmentation/helpers/augment_eda.py
import os, sys, shutil def augment_eda(textfile, basedir): arg_num= 1 text='1\t'+open(os.getcwd()+'/'+textfile).read() textfile2=open(textfile,'w') textfile2.write(text) textfile2.close() shutil.copy(os.getcwd()+'/'+textfile,basedir+'/helpers/eda_nlp/data/'+textfile) newfile='augmented_'+textfile os.system('python3 %s/helpers/eda_nlp/code/augment.py --input=%s --output=%s --num_aug=%s --alpha=0.05'%(basedir, textfile, newfile, str(arg_num))) shutil.copy(basedir+'/helpers/eda_nlp/data/'+newfile, os.getcwd()+'/'+newfile) os.remove(basedir+'/helpers/eda_nlp/data/'+textfile) os.remove(basedir+'/helpers/eda_nlp/data/'+newfile)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false
jim-schwoebel/allie
https://github.com/jim-schwoebel/allie/blob/b89f1403f63033ad406d0606b7c7a45000b43481/augmentation/text_augmentation/helpers/eda_nlp/experiments/b_1_data_process.py
augmentation/text_augmentation/helpers/eda_nlp/experiments/b_1_data_process.py
from methods import * from b_config import * if __name__ == "__main__": #generate the augmented data sets for dataset_folder in dataset_folders: #pre-existing file locations train_orig = dataset_folder + '/train_orig.txt' #file to be created train_aug_st = dataset_folder + '/train_aug_st.txt' #standard augmentation gen_standard_aug(train_orig, train_aug_st) #generate the vocab dictionary word2vec_pickle = dataset_folder + '/word2vec.p' # don't want to load the huge pickle every time, so just save the words that are actually used into a smaller dictionary gen_vocab_dicts(dataset_folder, word2vec_pickle, huge_word2vec)
python
Apache-2.0
b89f1403f63033ad406d0606b7c7a45000b43481
2026-01-05T07:09:07.495102Z
false