External Issues

External Issues

In this doc, we will summarize the issues we filed to external projects. Since ElasticDL uses some frameworks such as TensorFlow 2.0, it may have some dependency on these projects to implement some features. What's more, we also have some ideas to make improvement on external projects. We will list the issues of all the categories here.

TensorFlow

Memory Leak

Issue 35010: Memory leak with tf.py_function in eager mode using TF 2.0

Issue 35152: Memory leaks when using tf.strings.split in map_func for tf.data.Dataset.map with eager execution.

Issue 35804: Memory leak when using py_function inside tf.data.Dataset

Issue 35044: Memory leaks when using tf.keras.metrics update_states in multi-threads

Feature Column

Issue 34862: tf.feature_column.shared_embeddings supports eager mode

Others

Issue 34618: Support exporting the transform graph together with the model graph defined using Keras to SavedModel in TensorFlow 2.0.

Issue 34793: Sequence pad support for Ragged Tensor when call to_tensor

Issue 33880: Embedding layer for sparse inputs in tf.keras.layers