Update API docs

Author: Lifann

docs/api_docs/tfra/dynamic_embedding.md

@@ -27,18 +27,32 @@ Export dynamic_embedding APIs.
 
 ## Modules
 
+[`keras`](../tfra/dynamic_embedding/keras.md) module
+
 [`math`](../tfra/dynamic_embedding/math.md) module: math operations.
 
+[`shadow_ops`](../tfra/dynamic_embedding/shadow_ops.md) module: Dynamic Embedding is designed for Large-scale Sparse Weights Training.
+
 ## Classes
 
 [`class CuckooHashTable`](../tfra/dynamic_embedding/CuckooHashTable.md): A generic mutable hash table implementation.
 
+[`class CuckooHashTableConfig`](../tfra/dynamic_embedding/CuckooHashTableConfig.md)
+
+[`class CuckooHashTableCreator`](../tfra/dynamic_embedding/CuckooHashTableCreator.md):   A generic KV table creator.
+
 [`class FrequencyRestrictPolicy`](../tfra/dynamic_embedding/FrequencyRestrictPolicy.md): A derived policy to eliminate features in variable follow the
 
-[`class GraphKeys`](../tfra/dynamic_embedding/GraphKeys.md): Extended standard names related to `dynamic_embedding_ops.Variable` to use
+[`class GraphKeys`](../tfra/dynamic_embedding/GraphKeys.md): (Deprecated) extended standard names related to
 
 [`class ModelMode`](../tfra/dynamic_embedding/ModelMode.md): The global config of model modes.
 
+[`class RedisTable`](../tfra/dynamic_embedding/RedisTable.md): A generic mutable hash table implementation.
+
+[`class RedisTableConfig`](../tfra/dynamic_embedding/RedisTableConfig.md): RedisTableConfig config json file for connecting Redis service and 
+
+[`class RedisTableCreator`](../tfra/dynamic_embedding/RedisTableCreator.md):   RedisTableCreator will create a object to pass itself to the others classes
+
 [`class RestrictPolicy`](../tfra/dynamic_embedding/RestrictPolicy.md): Base class of restrict policies. Never use this class directly, but
 
 [`class TimestampRestrictPolicy`](../tfra/dynamic_embedding/TimestampRestrictPolicy.md): A derived policy to eliminate features in variable follow the

docs/api_docs/tfra/dynamic_embedding/CuckooHashTable.md

@@ -7,6 +7,8 @@
 <meta itemprop="property" content="value_dtype"/>
 <meta itemprop="property" content="__getitem__"/>
 <meta itemprop="property" content="__init__"/>
+<meta itemprop="property" content="accum"/>
+<meta itemprop="property" content="clear"/>
 <meta itemprop="property" content="export"/>
 <meta itemprop="property" content="insert"/>
 <meta itemprop="property" content="lookup"/>
@@ -67,8 +69,9 @@ __init__(
     value_dtype,
     default_value,
     name='CuckooHashTable',
-    checkpoint=True,
-    init_size=0
+    checkpoint=(True),
+    init_size=0,
+    config=None
 )
 ```
 
@@ -87,7 +90,7 @@ and value_dtype, respectively.
 * <b>`checkpoint`</b>: if True, the contents of the table are saved to and restored
   from checkpoints. If `shared_name` is empty for a checkpointed table, it
   is shared using the table node name.
-* <b>`init_size`</b>: initial size for the Variable and initial size of each hash 
+* <b>`init_size`</b>: initial size for the Variable and initial size of each hash
   tables will be int(init_size / N), N is the number of the devices.
 
 
@@ -139,6 +142,68 @@ __getitem__(keys)
 Looks up `keys` in a table, outputs the corresponding values.
 
 
+<h3 id="accum"><code>accum</code></h3>
+
+<a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/cuckoo_hashtable_ops.py">View source</a>
+
+``` python
+accum(
+    keys,
+    values_or_deltas,
+    exists,
+    name=None
+)
+```
+
+Associates `keys` with `values`.
+
+
+#### Args:
+
+
+* <b>`keys`</b>: Keys to accmulate. Can be a tensor of any shape.
+  Must match the table's key type.
+* <b>`values_or_deltas`</b>: values to be associated with keys. Must be a tensor of
+  the same shape as `keys` and match the table's value type.
+* <b>`exists`</b>: A bool type tensor indicates if keys already exist or not.
+  Must be a tensor of the same shape as `keys`.
+* <b>`name`</b>: A name for the operation (optional).
+
+
+#### Returns:
+
+The created Operation.
+
+
+
+#### Raises:
+
+
+* <b>`TypeError`</b>: when `keys` or `values` doesn't match the table data
+  types.
+
+<h3 id="clear"><code>clear</code></h3>
+
+<a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/cuckoo_hashtable_ops.py">View source</a>
+
+``` python
+clear(name=None)
+```
+
+clear all keys and values in the table.
+
+
+#### Args:
+
+
+* <b>`name`</b>: A name for the operation (optional).
+
+
+#### Returns:
+
+The created Operation.
+
+
 <h3 id="export"><code>export</code></h3>
 
 <a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/cuckoo_hashtable_ops.py">View source</a>
@@ -207,6 +272,7 @@ The created Operation.
 lookup(
     keys,
     dynamic_default_values=None,
+    return_exists=(False),
     name=None
 )
 ```
@@ -223,6 +289,8 @@ The `default_value` is used for keys not present in the table.
 * <b>`dynamic_default_values`</b>: The values to use if a key is missing in the
   table. If None (by default), the static default_value
   `self._default_value` will be used.
+* <b>`return_exists`</b>: if True, will return a additional Tensor which indicates
+  if or not keys are existing in the table.
 * <b>`name`</b>: A name for the operation (optional).
 
 
@@ -231,6 +299,9 @@ The `default_value` is used for keys not present in the table.
 A tensor containing the values in the same shape as `keys` using the
   table's value type.
 
+* <b>`exists`</b>:   A bool type Tensor of the same shape as `keys` which indicates
+    if keys are existing in the table.
+    Only provided if `return_exists` is True.
 
 
 #### Raises:

docs/api_docs/tfra/dynamic_embedding/CuckooHashTableConfig.md

@@ -0,0 +1,47 @@
+<div itemscope itemtype="http://developers.google.com/ReferenceObject">
+<meta itemprop="name" content="tfra.dynamic_embedding.CuckooHashTableConfig" />
+<meta itemprop="path" content="Stable" />
+<meta itemprop="property" content="__init__"/>
+</div>
+
+# tfra.dynamic_embedding.CuckooHashTableConfig
+
+<!-- Insert buttons and diff -->
+
+<table class="tfo-notebook-buttons tfo-api" align="left">
+
+<td>
+  <a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/dynamic_embedding_creator.py">
+    <img src="https://www.tensorflow.org/images/GitHub-Mark-32px.png" />
+    View source on GitHub
+  </a>
+</td></table>
+<br/>
+<br/>
+<br/>
+<br/>
+
+
+
+## Class `CuckooHashTableConfig`
+
+
+
+
+
+<!-- Placeholder for "Used in" -->
+
+
+<h2 id="__init__"><code>__init__</code></h2>
+
+<a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/dynamic_embedding_creator.py">View source</a>
+
+``` python
+__init__()
+```
+
+CuckooHashTableConfig include nothing for parameter default satisfied.
+    
+
+
+

docs/api_docs/tfra/dynamic_embedding/CuckooHashTableCreator.md

@@ -0,0 +1,99 @@
+<div itemscope itemtype="http://developers.google.com/ReferenceObject">
+<meta itemprop="name" content="tfra.dynamic_embedding.CuckooHashTableCreator" />
+<meta itemprop="path" content="Stable" />
+<meta itemprop="property" content="__init__"/>
+<meta itemprop="property" content="create"/>
+<meta itemprop="property" content="get_config"/>
+</div>
+
+# tfra.dynamic_embedding.CuckooHashTableCreator
+
+<!-- Insert buttons and diff -->
+
+<table class="tfo-notebook-buttons tfo-api" align="left">
+
+<td>
+  <a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/dynamic_embedding_creator.py">
+    <img src="https://www.tensorflow.org/images/GitHub-Mark-32px.png" />
+    View source on GitHub
+  </a>
+</td></table>
+<br/>
+<br/>
+<br/>
+<br/>
+
+
+
+## Class `CuckooHashTableCreator`
+
+  A generic KV table creator.
+
+
+
+<!-- Placeholder for "Used in" -->
+
+  KV table instance will be created by the create function with config.
+And also a config class for specific table instance backend should be
+inited before callling the creator function.
+  And then, the KVCreator class instance will be passed to the Variable
+class for creating the real KV table backend(TF resource).
+
+#### Example usage:
+
+
+
+```python
+redis_config1=tfra.dynamic_embedding.RedisTableConfig(
+  redis_config_abs_dir="xx/yy.json"
+)
+redis_creator1=tfra.dynamic_embedding.RedisTableCreator(redis_config1)
+```
+
+<h2 id="__init__"><code>__init__</code></h2>
+
+<a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/dynamic_embedding_creator.py">View source</a>
+
+``` python
+__init__(config=None)
+```
+
+Initialize self.  See help(type(self)) for accurate signature.
+
+
+
+
+## Methods
+
+<h3 id="create"><code>create</code></h3>
+
+<a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/dynamic_embedding_creator.py">View source</a>
+
+``` python
+create(
+    key_dtype=None,
+    value_dtype=None,
+    default_value=None,
+    name=None,
+    checkpoint=None,
+    init_size=None,
+    config=None
+)
+```
+
+
+
+
+<h3 id="get_config"><code>get_config</code></h3>
+
+<a target="_blank" href="https://github.com/tensorflow/recommenders-addons/tree/master/tensorflow_recommenders_addons/dynamic_embedding/python/ops/dynamic_embedding_creator.py">View source</a>
+
+``` python
+get_config()
+```
+
+
+
+
+
+

docs/api_docs/tfra/dynamic_embedding/DynamicEmbeddingOptimizer.md

@@ -25,7 +25,10 @@
 An optimizer wrapper to make any TensorFlow optimizer capable of training
 
 ``` python
-tfra.dynamic_embedding.DynamicEmbeddingOptimizer()
+tfra.dynamic_embedding.DynamicEmbeddingOptimizer(
+    bp_v2=(False),
+    synchronous=(False)
+)
 ```
 
 
@@ -37,6 +40,11 @@ Dynamic Embeddding Variables.
 
 
 * <b>`self`</b>: a TensorFlow optimizer.
+* <b>`bp_v2`</b>: If True, updating parameters will use updating instead of setting, which solves
+  the race condition problem among workers during back-propagation in large-scale
+  distributed asynchronous training. Reference: https://www.usenix.org/system/files/osdi20-jiang.pdf
+* <b>`synchronous`</b>: If True, we will use horovod's all-reduce method to merge the dense grad of model parameter, 
+  the default reduce method is SUM. For TrainableWrapper's grad, keep same with before.
 
 
 #### Example usage:

docs/api_docs/tfra/dynamic_embedding/GraphKeys.md

@@ -26,15 +26,13 @@
 
 ## Class `GraphKeys`
 
-Extended standard names related to `dynamic_embedding_ops.Variable` to use
+(Deprecated) extended standard names related to
 
 
 
 <!-- Placeholder for "Used in" -->
-for graph collections.
-
+`dynamic_embedding_ops.Variable` to use for graph collections.
 The following standard keys are defined:
-
 * `DYNAMIC_EMBEDDING_VARIABLES`: the default collection of
   all `dynamic_embedding_ops.Variable` objects.
 * `TRAINABLE_DYNAMIC_EMBEDDING_VARIABLES`: the subset of