Detached TensorFlow code from mltu package

pythonlessons · pythonlessons · commit 9f9fcc1ce1fd · 2023-02-27T10:08:11.000+02:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,3 +1,19 @@
+## [1.0.0] - 2022-02-27
+
+### Changed
+- detaching TensorFlow from mltu, now mltu is only a collection of utilities for training machine learning models
+
+### Added:
+-
+
+## [0.1.6] - 2022-02-26
+
+### Changed
+
+### Added:
+- added 05_sound_to_text tutorial
+- added WavReader to mltu/preprocessors, used to read wav files and convert them to numpy arrays
+
 ## [0.1.5] - 2022-01-10
 
 ### Changed
diff --git a/README.md b/README.md
@@ -1,5 +1,5 @@
-# MLTU - Machine Learning Training Utilities (TensorFlow)
-Machine Learning Training Utilities with TensorFlow 2.* and Python 3
+# MLTU - Machine Learning Training Utilities
+Machine Learning Training Utilities for TensorFlow 2.* and PyTorch with Python 3
 
 # Installation:
 To use MLTU in your own project, you can install it from PyPI:
diff --git a/Tutorials/01_image_to_word/model.py b/Tutorials/01_image_to_word/model.py
@@ -1,7 +1,7 @@
 from keras import layers
 from keras.models import Model
 
-from mltu.model_utils import residual_block
+from mltu.tensorflow.model_utils import residual_block
 
 def train_model(input_dim, output_dim, activation='leaky_relu', dropout=0.2):
     
diff --git a/Tutorials/01_image_to_word/train.py b/Tutorials/01_image_to_word/train.py
@@ -10,9 +10,9 @@
 from mltu.dataProvider import DataProvider
 from mltu.preprocessors import ImageReader
 from mltu.transformers import ImageResizer, LabelIndexer, LabelPadding
-from mltu.losses import CTCloss
-from mltu.callbacks import Model2onnx, TrainLogger
-from mltu.metrics import CWERMetric
+from mltu.tensorflow.losses import CTCloss
+from mltu.tensorflow.callbacks import Model2onnx, TrainLogger
+from mltu.tensorflow.metrics import CWERMetric
 
 from model import train_model
 from configs import ModelConfigs
diff --git a/Tutorials/02_captcha_to_text/model.py b/Tutorials/02_captcha_to_text/model.py
@@ -1,7 +1,7 @@
 from keras import layers
 from keras.models import Model
 
-from mltu.model_utils import residual_block
+from mltu.tensorflow.model_utils import residual_block
 
 def train_model(input_dim, output_dim, activation='leaky_relu', dropout=0.2):
     
diff --git a/Tutorials/02_captcha_to_text/train.py b/Tutorials/02_captcha_to_text/train.py
@@ -4,13 +4,14 @@
 
 from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau, TensorBoard
 
-from mltu.dataProvider import DataProvider
+from mltu.tensorflow.dataProvider import DataProvider
+from mltu.tensorflow.losses import CTCloss
+from mltu.tensorflow.callbacks import Model2onnx, TrainLogger
+from mltu.tensorflow.metrics import CWERMetric
+
 from mltu.preprocessors import ImageReader
 from mltu.transformers import ImageResizer, LabelIndexer, LabelPadding
 from mltu.augmentors import RandomBrightness, RandomRotate, RandomErodeDilate
-from mltu.losses import CTCloss
-from mltu.callbacks import Model2onnx, TrainLogger
-from mltu.metrics import CWERMetric
 
 from model import train_model
 from configs import ModelConfigs
@@ -70,15 +71,15 @@ def download_and_unzip(url, extract_to='Datasets'):
 model.compile(
     optimizer=tf.keras.optimizers.Adam(learning_rate=configs.learning_rate), 
     loss=CTCloss(), 
-    metrics=[CWERMetric()],
+    metrics=[CWERMetric(padding_token=len(configs.vocab))],
     run_eagerly=False
 )
 model.summary(line_length=110)
 # Define path to save the model
 stow.mkdir(configs.model_path)
 
 # Define callbacks
-earlystopper = EarlyStopping(monitor='val_CER', patience=40, verbose=1)
+earlystopper = EarlyStopping(monitor='val_CER', patience=50, verbose=1)
 checkpoint = ModelCheckpoint(f"{configs.model_path}/model.h5", monitor='val_CER', verbose=1, save_best_only=True, mode='min')
 trainLogger = TrainLogger(configs.model_path)
 tb_callback = TensorBoard(f'{configs.model_path}/logs', update_freq=1)
diff --git a/Tutorials/03_handwriting_recognition/model.py b/Tutorials/03_handwriting_recognition/model.py
@@ -1,7 +1,7 @@
 from keras import layers
 from keras.models import Model
 
-from mltu.model_utils import residual_block
+from mltu.tensorflow.model_utils import residual_block
 
 def train_model(input_dim, output_dim, activation='leaky_relu', dropout=0.2):
     
diff --git a/Tutorials/03_handwriting_recognition/train.py b/Tutorials/03_handwriting_recognition/train.py
@@ -4,13 +4,14 @@
 
 from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau, TensorBoard
 
-from mltu.dataProvider import DataProvider
 from mltu.preprocessors import ImageReader
 from mltu.transformers import ImageResizer, LabelIndexer, LabelPadding, ImageShowCV2
 from mltu.augmentors import RandomBrightness, RandomRotate, RandomErodeDilate, RandomSharpen
-from mltu.losses import CTCloss
-from mltu.callbacks import Model2onnx, TrainLogger
-from mltu.metrics import CWERMetric
+
+from mltu.tensorflow.dataProvider import DataProvider
+from mltu.tensorflow.losses import CTCloss
+from mltu.tensorflow.callbacks import Model2onnx, TrainLogger
+from mltu.tensorflow.metrics import CWERMetric
 
 from model import train_model
 from configs import ModelConfigs
diff --git a/Tutorials/04_sentence_recognition/model.py b/Tutorials/04_sentence_recognition/model.py
@@ -1,7 +1,7 @@
 from keras import layers
 from keras.models import Model
 
-from mltu.model_utils import residual_block
+from mltu.tensorflow.model_utils import residual_block
 
 def train_model(input_dim, output_dim, activation='leaky_relu', dropout=0.2):
     
diff --git a/Tutorials/04_sentence_recognition/train.py b/Tutorials/04_sentence_recognition/train.py
@@ -4,13 +4,14 @@
 
 from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau, TensorBoard
 
-from mltu.dataProvider import DataProvider
 from mltu.preprocessors import ImageReader
 from mltu.transformers import ImageResizer, LabelIndexer, LabelPadding, ImageShowCV2
 from mltu.augmentors import RandomBrightness, RandomRotate, RandomErodeDilate, RandomSharpen
-from mltu.losses import CTCloss
-from mltu.callbacks import Model2onnx, TrainLogger
-from mltu.metrics import CERMetric, WERMetric
+
+from mltu.tensorflow.dataProvider import DataProvider
+from mltu.tensorflow.losses import CTCloss
+from mltu.tensorflow.callbacks import Model2onnx, TrainLogger
+from mltu.tensorflow.metrics import CERMetric, WERMetric
 
 from model import train_model
 from configs import ModelConfigs
diff --git a/Tutorials/05_sound_to_text/model.py b/Tutorials/05_sound_to_text/model.py
@@ -2,7 +2,7 @@
 from keras import layers
 from keras.models import Model
 
-from mltu.model_utils import residual_block, activation_layer
+from mltu.tensorflow.model_utils import residual_block, activation_layer
 
 def train_model(input_dim, output_dim, activation='leaky_relu', dropout=0.2):
     
diff --git a/Tutorials/05_sound_to_text/train.py b/Tutorials/05_sound_to_text/train.py
@@ -10,12 +10,13 @@
 from io import BytesIO
 
 from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau, TensorBoard
-from mltu.dataProvider import DataProvider
 from mltu.preprocessors import WavReader
+
+from mltu.tensorflow.dataProvider import DataProvider
 from mltu.transformers import LabelIndexer, LabelPadding, SpectrogramPadding
-from mltu.losses import CTCloss
-from mltu.callbacks import Model2onnx, TrainLogger
-from mltu.metrics import CERMetric, WERMetric
+from mltu.tensorflow.losses import CTCloss
+from mltu.tensorflow.callbacks import Model2onnx, TrainLogger
+from mltu.tensorflow.metrics import CERMetric, WERMetric
 
 from model import train_model
 from configs import ModelConfigs
diff --git a/mltu/__init__.py b/mltu/__init__.py
@@ -1 +1 @@
-__version__ = "0.1.6"
+__version__ = "1.0.0"
diff --git a/mltu/dataProvider.py b/mltu/dataProvider.py
@@ -4,7 +4,6 @@
 import numpy as np
 import pandas as pd
 from tqdm import tqdm
-import tensorflow as tf
 
 from .augmentors import Augmentor
 from .transformers import Transformer
@@ -14,7 +13,7 @@
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)
 
-class DataProvider(tf.keras.utils.Sequence):
+class DataProvider:
     """ Standardised object for providing data to a TensorFlow model while training.
 
     Args:
@@ -38,7 +37,6 @@ def __init__(
         augmentors: typing.List[Augmentor] = None,
         transformers: typing.List[Transformer] = None,
         skip_validation: bool = False,
-        store_in_memory: bool = False,
         limit: int = None,
         ) -> None:
         super().__init__()
@@ -142,7 +140,7 @@ def validate(self, dataset: typing.Union[str, list, pd.DataFrame], skip_validati
         else:
             raise TypeError("Dataset must be a path, list or pandas dataframe.")
 
-    def split(self, split: float = 0.9, shuffle: bool = True) -> typing.Tuple[tf.keras.utils.Sequence, tf.keras.utils.Sequence]:
+    def split(self, split: float = 0.9, shuffle: bool = True) -> typing.Tuple[typing.Any, typing.Any]:
         """ Split current data provider into training and validation data providers. 
         
         Args:
diff --git a/mltu/tensorflow/README.md b/mltu/tensorflow/README.md
@@ -0,0 +1 @@
+# Functions and objects specific for TensorFlow 2.* and Python 3
diff --git a/mltu/tensorflow/Tests/test_metrics.py b/mltu/tensorflow/Tests/test_metrics.py
@@ -1,6 +1,6 @@
 import unittest
 import numpy as np
-from mltu.metrics import CERMetric, WERMetric
+from mltu.tensorflow.metrics import CERMetric, WERMetric
 
 import numpy as np
 import tensorflow as tf
diff --git a/mltu/tensorflow/callbacks.py b/mltu/tensorflow/callbacks.py
diff --git a/mltu/tensorflow/dataProvider.py b/mltu/tensorflow/dataProvider.py
@@ -0,0 +1,7 @@
+import tensorflow as tf
+
+from ..dataProvider import DataProvider as dataProvider
+
+class DataProvider(dataProvider, tf.keras.utils.Sequence):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
diff --git a/mltu/tensorflow/losses.py b/mltu/tensorflow/losses.py
diff --git a/mltu/tensorflow/metrics.py b/mltu/tensorflow/metrics.py
diff --git a/mltu/tensorflow/model_utils.py b/mltu/tensorflow/model_utils.py
diff --git a/mltu/tensorflow/requirements.txt b/mltu/tensorflow/requirements.txt
@@ -0,0 +1 @@
+tensorflow<=2.10.1
diff --git a/requirements.txt b/requirements.txt
@@ -6,7 +6,6 @@ numpy
 opencv-python
 tf2onnx
 onnxruntime # onnxruntime-gpu for GPU support
-tensorflow<=2.10.1
 librosa==0.9.2
 importlib==1.0.4
 matplotlib
