docs

README.md

@@ -13,7 +13,7 @@
 
 ### News and Announcements
 - **2023-04-21**
-  - **ktrain 0.36.x** is released and supports **Sentiment Analysis**. See the [example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/develop/examples/text/sentiment_analysis_example.ipynb) for more information. 
+  - **ktrain 0.36.x** is released and supports a simple wrapper to **Sentiment Analysis**. See the [example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/develop/examples/text/sentiment_analysis_example.ipynb) for more information. 
 ```python
 # Example: Sentiment Analysis
 from ktrain.text.sentiment import SentimentAnalyzer
@@ -42,9 +42,6 @@ print(model.execute(prompt))
 ```
 - **2023-03-30**
   - **ktrain 0.34.x** is released and supports fast LexRank-based text summarization.
-- **2023-01-14**
-  - **ktrain 0.33.x** is released and includes fixes to support the latest version of Hugging Face`transformers`. Note that `transformers<=4.25.1` [has a bug](https://github.com/huggingface/transformers/issues/20750) related to TensorFlow 2.11. You can downgrade TensorFlow to 2.10 if you receive an error that says *"has no attribute 'expand_1d'"* (or upgrade to `transformers>4.25.1` if available).
-
 ----
 
 ### Overview
@@ -70,6 +67,7 @@ print(model.execute(prompt))
      - **Speech Transcription**: Extract text from audio files <sub><sup>[[example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/develop/examples/text/speech_transcription_example.ipynb)]</sup></sub>
      - **Universal Information Extraction**:  extract any kind of information from documents by simply phrasing it in the form of a question <sub><sup>[[example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/master/examples/text/qa_information_extraction.ipynb)]</sup></sub>
      - **Keyphrase Extraction**:  extract keywords from documents <sub><sup>[[example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/develop/examples/text/keyword_extraction_example.ipynb)]</sup></sub>
+     - **Sentiment Analysis**: easy-to-use wrapper to pretrained sentiment analysis <sub><sup>[[example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/develop/examples/text/sentiment_analysis_example.ipynb)]</sup
      - **Generative AI with GPT**: Provide instructions to a lightweight ChatGPT-like model running on your own own machine to solve various tasks. Model was fine-tuned on the [Alpaca](https://github.com/tatsu-lab/stanford_alpaca) instruction dataset ([CC By NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/deed.en_GB)) <sub><sup>[[example notebook](https://nbviewer.jupyter.org/github/amaiya/ktrain/blob/develop/examples/text/generative_ai_example.ipynb)]</sup>
   - `vision` data:
     - **image classification** (e.g., [ResNet](https://arxiv.org/abs/1512.03385), [Wide ResNet](https://arxiv.org/abs/1605.07146), [Inception](https://www.cs.unc.edu/~wliu/papers/GoogLeNet.pdf)) <sub><sup>[[example notebook](https://colab.research.google.com/drive/1WipQJUPL7zqyvLT10yekxf_HNMXDDtyR)]</sup></sub>

docs/lroptimize/optimization.html

@@ -558,7 +558,7 @@ <h2 class="section-title" id="header-classes">Classes</h2>
   # Simple example, double the gradients.
   return [(2. * g, v) for g, v in grads_and_vars]
 
-optimizer = tf.keras.optimizers.SGD(
+optimizer = tf.keras.optimizers.legacy.SGD(
     1e-3, gradient_transformers=[my_gradient_transformer])
 </code></pre>
 <h2 id="args">Args</h2>
@@ -568,9 +568,9 @@ <h2 id="args">Args</h2>
 by the optimizer.</dd>
 <dt><strong><code>gradient_aggregator</code></strong></dt>
 <dd>The function to use to aggregate gradients across
-devices (when using <code>tf.distribute.Strategy</code>). If <code>None</code>, defaults to
-summing the gradients across devices. The function should accept and
-return a list of <code>(gradient, variable)</code> tuples.</dd>
+devices (when using <code>tf.distribute.Strategy</code>). If <code>None</code>, defaults
+to summing the gradients across devices. The function should accept
+and return a list of <code>(gradient, variable)</code> tuples.</dd>
 <dt><strong><code>gradient_transformers</code></strong></dt>
 <dd>Optional. List of functions to use to transform
 gradients before applying updates to Variables. The functions are
@@ -582,9 +582,10 @@ <h2 id="args">Args</h2>
 If <code>clipvalue</code> (float) is set, the gradient of each weight
 is clipped to be no higher than this value.
 If <code>clipnorm</code> (float) is set, the gradient of each weight
-is individually clipped so that its norm is no higher than this value.
-If <code>global_clipnorm</code> (float) is set the gradient of all weights is
-clipped so that their global norm is no higher than this value.</dd>
+is individually clipped so that its norm is no higher than this
+value. If <code>global_clipnorm</code> (float) is set the gradient of all
+weights is clipped so that their global norm is no higher than this
+value.</dd>
 </dl>
 <h2 id="raises">Raises</h2>
 <dl>
@@ -743,7 +744,7 @@ <h3>Ancestors</h3>
 <ul class="hlist">
 <li>keras.optimizers.optimizer_v2.adam.Adam</li>
 <li>keras.optimizers.optimizer_v2.optimizer_v2.OptimizerV2</li>
-<li>tensorflow.python.training.tracking.base.Trackable</li>
+<li>tensorflow.python.trackable.base.Trackable</li>
 </ul>
 <h3>Static methods</h3>
 <dl>
@@ -776,8 +777,8 @@ <h3>Methods</h3>
 <p>This is the second part of <code>minimize()</code>. It returns an <code>Operation</code> that
 applies gradients.</p>
 <p>The method sums gradients from all replicas in the presence of
-<code>tf.distribute.Strategy</code> by default. You can aggregate gradients yourself by
-passing <code>experimental_aggregate_gradients=False</code>.</p>
+<code>tf.distribute.Strategy</code> by default. You can aggregate gradients
+yourself by passing <code>experimental_aggregate_gradients=False</code>.</p>
 <p>Example:</p>
 <pre><code class="language-python">grads = tape.gradient(loss, vars)
 grads = tf.distribute.get_replica_context().all_reduce('sum', grads)
@@ -791,12 +792,13 @@ <h2 id="args">Args</h2>
 <dt><strong><code>grads_and_vars</code></strong></dt>
 <dd>List of (gradient, variable) pairs.</dd>
 <dt><strong><code>name</code></strong></dt>
-<dd>Optional name for the returned operation. Default to the name passed
-to the <code>Optimizer</code> constructor.</dd>
+<dd>Optional name for the returned operation. Default to the name
+passed to the <code>Optimizer</code> constructor.</dd>
 <dt><strong><code>experimental_aggregate_gradients</code></strong></dt>
-<dd>Whether to sum gradients from different
-replicas in the presence of <code>tf.distribute.Strategy</code>. If False, it's
-user responsibility to aggregate the gradients. Default to True.</dd>
+<dd>Whether to sum gradients from
+different replicas in the presence of <code>tf.distribute.Strategy</code>. If
+False, it's user responsibility to aggregate the gradients. Default
+to True.</dd>
 </dl>
 <h2 id="returns">Returns</h2>
 <p>An <code>Operation</code> that applies the specified gradients. The <code>iterations</code>

docs/lroptimize/triangular.html

@@ -761,19 +761,19 @@ <h3>Methods</h3>
 </code></dt>
 <dd>
 <div class="desc"><p>Called at the end of an epoch.</p>
-<p>Subclasses should override for any actions to run. This function should only
-be called during TRAIN mode.</p>
+<p>Subclasses should override for any actions to run. This function should
+only be called during TRAIN mode.</p>
 <h2 id="args">Args</h2>
 <dl>
 <dt><strong><code>epoch</code></strong></dt>
 <dd>Integer, index of epoch.</dd>
 <dt><strong><code>logs</code></strong></dt>
 <dd>Dict, metric results for this training epoch, and for the
-validation epoch if validation is performed. Validation result keys
-are prefixed with <code>val_</code>. For training epoch, the values of the</dd>
-</dl>
-<p><code>Model</code>'s metrics are returned. Example : <code>{'loss': 0.2, 'accuracy':
-0.7}</code>.</p></div>
+validation epoch if validation is performed. Validation result
+keys are prefixed with <code>val_</code>. For training epoch, the values of
+the <code>Model</code>'s metrics are returned. Example:
+<code>{'loss': 0.2, 'accuracy': 0.7}</code>.</dd>
+</dl></div>
 <details class="source">
 <summary>
 <span>Expand source code</span>
@@ -839,8 +839,8 @@ <h2 id="args">Args</h2>
 <h2 id="args">Args</h2>
 <dl>
 <dt><strong><code>logs</code></strong></dt>
-<dd>Dict. Currently no data is passed to this argument for this method
-but that may change in the future.</dd>
+<dd>Dict. Currently no data is passed to this argument for this
+method but that may change in the future.</dd>
 </dl></div>
 <details class="source">
 <summary>

docs/text/index.html

@@ -157,6 +157,10 @@ <h2 class="section-title" id="header-submodules">Sub-modules</h2>
 <dd>
 <div class="desc"></div>
 </dd>
+<dt><code class="name"><a title="ktrain.text.sentiment" href="sentiment/index.html">ktrain.text.sentiment</a></code></dt>
+<dd>
+<div class="desc"></div>
+</dd>
 <dt><code class="name"><a title="ktrain.text.shallownlp" href="shallownlp/index.html">ktrain.text.shallownlp</a></code></dt>
 <dd>
 <div class="desc"></div>
@@ -6624,6 +6628,7 @@ <h1>Index</h1>
 <li><code><a title="ktrain.text.predictor" href="predictor.html">ktrain.text.predictor</a></code></li>
 <li><code><a title="ktrain.text.preprocessor" href="preprocessor.html">ktrain.text.preprocessor</a></code></li>
 <li><code><a title="ktrain.text.qa" href="qa/index.html">ktrain.text.qa</a></code></li>
+<li><code><a title="ktrain.text.sentiment" href="sentiment/index.html">ktrain.text.sentiment</a></code></li>
 <li><code><a title="ktrain.text.shallownlp" href="shallownlp/index.html">ktrain.text.shallownlp</a></code></li>
 <li><code><a title="ktrain.text.speech" href="speech/index.html">ktrain.text.speech</a></code></li>
 <li><code><a title="ktrain.text.summarization" href="summarization/index.html">ktrain.text.summarization</a></code></li>

docs/text/ner/anago/callbacks.html

@@ -89,16 +89,17 @@ <h2 class="section-title" id="header-classes">Classes</h2>
 <p>Callbacks can be passed to keras methods such as <code>fit</code>, <code>evaluate</code>, and
 <code>predict</code> in order to hook into the various stages of the model training and
 inference lifecycle.</p>
-<p>To create a custom callback, subclass <code>keras.callbacks.Callback</code> and override
-the method associated with the stage of interest. See
+<p>To create a custom callback, subclass <code>keras.callbacks.Callback</code> and
+override the method associated with the stage of interest. See
 <a href="https://www.tensorflow.org/guide/keras/custom_callback">https://www.tensorflow.org/guide/keras/custom_callback</a> for more information.</p>
 <p>Example:</p>
 <pre><code class="language-python-repl">&gt;&gt;&gt; training_finished = False
 &gt;&gt;&gt; class MyCallback(tf.keras.callbacks.Callback):
 ...   def on_train_end(self, logs=None):
 ...     global training_finished
 ...     training_finished = True
-&gt;&gt;&gt; model = tf.keras.Sequential([tf.keras.layers.Dense(1, input_shape=(1,))])
+&gt;&gt;&gt; model = tf.keras.Sequential([
+...     tf.keras.layers.Dense(1, input_shape=(1,))])
 &gt;&gt;&gt; model.compile(loss='mean_squared_error')
 &gt;&gt;&gt; model.fit(tf.constant([[1.0]]), tf.constant([[1.0]]),
 ...           callbacks=[MyCallback()])
@@ -111,22 +112,21 @@ <h2 class="section-title" id="header-classes">Classes</h2>
 <li>You will need to manually call all the <code>on_*</code> methods at the appropriate
 locations in your loop. Like this:</li>
 </ol>
-<p>```
-callbacks =
-tf.keras.callbacks.CallbackList([&hellip;])
-callbacks.append(&hellip;)</p>
-<p>callbacks.on_train_begin(&hellip;)
-for epoch in range(EPOCHS):
-callbacks.on_epoch_begin(epoch)
-for i, data in dataset.enumerate():
-callbacks.on_train_batch_begin(i)
-batch_logs = model.train_step(data)
-callbacks.on_train_batch_end(i, batch_logs)
-epoch_logs = &hellip;
-callbacks.on_epoch_end(epoch, epoch_logs)
-final_logs=&hellip;
-callbacks.on_train_end(final_logs)
-```</p>
+<p>Example:</p>
+<pre><code class="language-python">   callbacks =  tf.keras.callbacks.CallbackList([...])
+   callbacks.append(...)
+   callbacks.on_train_begin(...)
+   for epoch in range(EPOCHS):
+     callbacks.on_epoch_begin(epoch)
+     for i, data in dataset.enumerate():
+       callbacks.on_train_batch_begin(i)
+       batch_logs = model.train_step(data)
+       callbacks.on_train_batch_end(i, batch_logs)
+     epoch_logs = ...
+     callbacks.on_epoch_end(epoch, epoch_logs)
+   final_logs=...
+   callbacks.on_train_end(final_logs)
+</code></pre>
 <h2 id="attributes">Attributes</h2>
 <dl>
 <dt><strong><code>params</code></strong></dt>
@@ -211,19 +211,19 @@ <h3>Methods</h3>
 </code></dt>
 <dd>
 <div class="desc"><p>Called at the end of an epoch.</p>
-<p>Subclasses should override for any actions to run. This function should only
-be called during TRAIN mode.</p>
+<p>Subclasses should override for any actions to run. This function should
+only be called during TRAIN mode.</p>
 <h2 id="args">Args</h2>
 <dl>
 <dt><strong><code>epoch</code></strong></dt>
 <dd>Integer, index of epoch.</dd>
 <dt><strong><code>logs</code></strong></dt>
 <dd>Dict, metric results for this training epoch, and for the
-validation epoch if validation is performed. Validation result keys
-are prefixed with <code>val_</code>. For training epoch, the values of the</dd>
-</dl>
-<p><code>Model</code>'s metrics are returned. Example : <code>{'loss': 0.2, 'accuracy':
-0.7}</code>.</p></div>
+validation epoch if validation is performed. Validation result
+keys are prefixed with <code>val_</code>. For training epoch, the values of
+the <code>Model</code>'s metrics are returned. Example:
+<code>{'loss': 0.2, 'accuracy': 0.7}</code>.</dd>
+</dl></div>
 <details class="source">
 <summary>
 <span>Expand source code</span>

docs/text/ner/anago/layers.html

@@ -1786,8 +1786,8 @@ <h3>Ancestors</h3>
 <ul class="hlist">
 <li>keras.engine.base_layer.Layer</li>
 <li>tensorflow.python.module.module.Module</li>
-<li>tensorflow.python.training.tracking.autotrackable.AutoTrackable</li>
-<li>tensorflow.python.training.tracking.base.Trackable</li>
+<li>tensorflow.python.trackable.autotrackable.AutoTrackable</li>
+<li>tensorflow.python.trackable.base.Trackable</li>
 <li>keras.utils.version_utils.LayerVersionSelector</li>
 </ul>
 <h3>Static methods</h3>
@@ -2020,10 +2020,13 @@ <h2 id="args">Args</h2>
 </code></dt>
 <dd>
 <div class="desc"><p>This is where the layer's logic lives.</p>
-<p>The <code>call()</code> method may not create state (except in its first invocation,
-wrapping the creation of variables or other resources in <code>tf.init_scope()</code>).
-It is recommended to create state in <code>__init__()</code>, or the <code>build()</code> method
-that is called automatically before <code>call()</code> executes the first time.</p>
+<p>The <code>call()</code> method may not create state (except in its first
+invocation, wrapping the creation of variables or other resources in
+<code>tf.init_scope()</code>).
+It is recommended to create state, including
+<code>tf.Variable</code> instances and nested <code>Layer</code> instances,
+in <code>__init__()</code>, or in the <code>build()</code> method that is
+called automatically before <code>call()</code> executes for the first time.</p>
 <h2 id="args">Args</h2>
 <dl>
 <dt><strong><code>inputs</code></strong></dt>
@@ -2032,15 +2035,17 @@ <h2 id="args">Args</h2>
 - <code>inputs</code> must be explicitly passed. A layer cannot have zero
 arguments, and <code>inputs</code> cannot be provided via the default value
 of a keyword argument.
-- NumPy array or Python scalar values in <code>inputs</code> get cast as tensors.
+- NumPy array or Python scalar values in <code>inputs</code> get cast as
+tensors.
 - Keras mask metadata is only collected from <code>inputs</code>.
 - Layers are built (<code>build(input_shape)</code> method)
 using shape info from <code>inputs</code> only.
 - <code>input_spec</code> compatibility is only checked against <code>inputs</code>.
 - Mixed precision input casting is only applied to <code>inputs</code>.
 If a layer has tensor arguments in <code>*args</code> or <code>**kwargs</code>, their
 casting behavior in mixed precision should be handled manually.
-- The SavedModel input specification is generated using <code>inputs</code> only.
+- The SavedModel input specification is generated using <code>inputs</code>
+only.
 - Integration with various ecosystem packages like TFMOT, TFLite,
 TF.js, etc is only supported for <code>inputs</code> and not for tensors in
 positional and keyword arguments.</dd>
@@ -2054,10 +2059,10 @@ <h2 id="args">Args</h2>
 - <code>training</code>: Boolean scalar tensor of Python boolean indicating
 whether the <code>call</code> is meant for training or inference.
 - <code>mask</code>: Boolean input mask. If the layer's <code>call()</code> method takes a
-<code>mask</code> argument, its default value will be set to the mask generated
-for <code>inputs</code> by the previous layer (if <code>input</code> did come from a layer
-that generated a corresponding mask, i.e. if it came from a Keras
-layer with masking support).</dd>
+<code>mask</code> argument, its default value will be set to the mask
+generated for <code>inputs</code> by the previous layer (if <code>input</code> did come
+from a layer that generated a corresponding mask, i.e. if it came
+from a Keras layer with masking support).</dd>
 </dl>
 <h2 id="returns">Returns</h2>
 <p>A tensor or list/tuple of tensors.</p></div>
@@ -2123,13 +2128,15 @@ <h2 id="returns">Returns</h2>
 <h2 id="args">Args</h2>
 <dl>
 <dt><strong><code>input_shape</code></strong></dt>
-<dd>Shape tuple (tuple of integers)
-or list of shape tuples (one per output tensor of the layer).
+<dd>Shape tuple (tuple of integers) or <code>tf.TensorShape</code>,
+or structure of shape tuples / <code>tf.TensorShape</code> instances
+(one per output tensor of the layer).
 Shape tuples can include None for free dimensions,
 instead of an integer.</dd>
 </dl>
 <h2 id="returns">Returns</h2>
-<p>An input shape tuple.</p></div>
+<p>A <code>tf.TensorShape</code> instance
+or structure of <code>tf.TensorShape</code> instances.</p></div>
 <details class="source">
 <summary>
 <span>Expand source code</span>
@@ -2163,9 +2170,9 @@ <h2 id="returns">Returns</h2>
 <p>The config of a layer does not include connectivity
 information, nor the layer class name. These are handled
 by <code>Network</code> (one layer of abstraction above).</p>
-<p>Note that <code>get_config()</code> does not guarantee to return a fresh copy of dict
-every time it is called. The callers should make a copy of the returned dict
-if they want to modify it.</p>
+<p>Note that <code>get_config()</code> does not guarantee to return a fresh copy of
+dict every time it is called. The callers should make a copy of the
+returned dict if they want to modify it.</p>
 <h2 id="returns">Returns</h2>
 <p>Python dictionary.</p></div>
 <details class="source">