Skip to main content

Multiverse for Deep Learning Developers without Pitfall

Project description

Multiverse for Deep Learning Developers without Pitfall

Release Note 0.6.0

New Features:

  1. new TextSequenceDataset and TextSequenceDataProvider for prepare dense/ sparse text tensor.

  2. rnn/ lstm/ gru is supporting now ( in pytorch backend).

  3. reinforcement learning is supporting now ( in pytorch backend).

  4. normalization in sequence supporting (automatic permute)

  5. bounding box inplace operation. (no need to split first 4 element)

  6. New optimizer: RangerLars, and gradient centralization is now supporting.

  7. New method ‘trigger_when’ in Model class , make add new callbacks or new event-listening much more easy.

Bug fix: #. min, max, reduce_min, reduce_max bad arguments issue.

0.5.9 #. fix the Model argments ordinal bug. #. fix the deprecated addcdiv_ functions

0.5.8 #. fix the deprecated divide ops and add_ functions

0.5.7 #. Support AMP (automatic mixed precision) with pytorch 1.6. #. attach lots of functions in ops on the tensor and use @numpy_compatible to share same syntax within numpy array or tensor. #. New attribute ‘sequence_rank’ comes to all conv_blocks familay, if sequence_rank=’cna’ means the order is ‘Convolution-Normalization-Activation’. #. Adding EvoNorm, SIREN ..

Project details

Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

tridentx-0.6.0.tar.gz (408.2 kB view hashes)

Uploaded Source

Built Distribution

tridentx-0.6.0-py3-none-any.whl (500.9 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page