[0.1.24] TensorNeko Release

Features:

• The codes which do not require PyTorch and PyTorch_Lightning have been moved to an independent library tensorneko_util, but the previous API is not changed.
  • You can use this new library with pip install tensorneko_util

Fix:

• tensorneko.util.identity function now can process the single argument.
• tensorneko.io.write.video.to now requires argument audio_codec to correctly save video if the argument audio is provided.

[0.1.23] TensorNeko Release

Features

• Add pickle friendly identity function tensorneko.util.identity
• Add wrappers for matplotlib plot and imshow

Fix:

• The matplotlib and seaborn requirements now are optional
• Align the text reader default call to of

[0.1.22] TensorNeko Release

Features:

• Several defined colors for visualization, and a seaborn palette with these colors.
• Add fast implementation based on Pytorch to speed up the resize_video.

Fix:

• Fix the tooltip line break in LineChart.

[0.1.21] TensorNeko Release

Features:

• Now the program will warn when the signature of a dispatch function is overridden.
• View can get the components byView[name]
• Add @dispatch.of(*types) for manually registering types
• Add ref for Ref type wrapper for primitive types
• Variable and ProgressBar can bind to Ref.
• LineChart for web watcher visualization

Fix:

• @dispatch now can support class methods

[0.1.19] TensorNeko Release

Features:

• Add View.add_all() which can add all previously defined components to the view.
• @dispatch for multi-dispatch functions.

[0.1.18] TensorNeko Release

Features:

• View.add uses vararg can add multiple components
• write.text.to_json now supports the object annotated by json_data

Fix:

• Fix the web watcher components constructors.
• Fix iou_1d return type annotation.

[0.1.17] TensorNeko Release

Features:

• Implement iou_1d in tensorneko.evaluation.metrics to calculate 1D IOU score for multiple proposals to multiple labels.

Fix:

• Fix subtitles in web watcher loggers
• Fix the shape input in video writer

[0.1.15] TensorNeko Release

Features:

• Add unpack support for audio_data and video_data
• Add default call for readers
• Add Conv1d, Conv3d, Aggregation and Stack
• Add padding_video, padding_audio
• Add Server for tensorboard. Add with keyword support for servers.

[0.1.13] TensorNeko Release

Features:

• Add Image class for web watcher display images.
• Add @json_data for parsing JSON to class. Use cls argument in tensorneko.io.read.text.of_json.
• Add Seed class for manually setting seeds to ensure reproducibility.

[0.1.11] TensorNeko Release

Fix:

• Variable watcher feature should work fine now.
• Refactor the implementation of web variable watcher.
• Change module name web to watcher.