perceiver-io 0.11.1

What's Changed

• Fix activation checkpointing by @cstub in #57

Full Changelog: 0.11.0...0.11.1

perceiver-io 0.11.0

What's Changed

• ci: flake8 -> ruff by @Borda in #48 (includes #46 and #47)
• Refactor redundancies in causal sequence modeling code by @krasserm in #49
• Key value cache and support for contrastive search by @krasserm in #51

Full Changelog: 0.10.0...0.11.0

perceiver-io 0.10.0

What's Changed

• Symbolic audio generation using Perceiver AR by @cstub in #45

Full Changelog: 0.9.0...0.10.0

perceiver-io 0.9.0

Highlights

• All official models and training checkpoint now available on the 🤗 Hub
• Inference notebook updated to use models from 🤗 Hub

What's Changed

• Upgrade to PyTorch 2.0 and PyTorch Lightning 2.0 by @krasserm in #43
• Hugging Face interface for inference by @krasserm in #44

Full Changelog: 0.8.2...0.9.0

perceiver-io 0.8.2

• Replace example-wise with batch-wise random-truncation
• Optional bias term for TiedTextOutputAdapter
• Optional absolute position embedding for Perceiver AR

perceiver-io 0.8.1

• Support tokenizers that do not have a pad_token configured (#37)
• Parameterize generate method with top_k directly instead of threshold (#38)

perceiver-io 0.8.0

What's Changed

• Perceiver AR enhancements by @krasserm in #36
• Support key padding masks for Perceiver AR by @krasserm in #25
• Implement streaming c4 dataset by @cstub in #29

Changelog: 0.7.0...0.8.0

See also milestone 0.8.0 for a complete list of pull requests and closed tickets.

perceiver-io 0.7.0

This release adds a Perceiver IO for predicting the optical flow between two images. It also adds utilities for producing an optical flow video from an input video (see inference notebook, for a demo). Thanks to @cstub for this great contribution. See milestone 0.7.0 for a list of closed tickets.

perceiver-io 0.7b1

Data preprocessing and documentation enhancements, major refactorings

Functional enhancements:

• Support for static word masking in addition to dynamic word masking.
• Support for individual token masking in addition to whole word masking.
• Task-specific data preprocessing for all supported text datasets.
• Constant learning rate scheduler with warmup now used by default.

Documentation enhancements:

• All training examples now provided as command line and Python script.
• Better overview of official models and example training checkpoints.
• Example training checkpoints can now be downloaded individually.
• Minor enhancements to all other documentation sections.

Refactorings and breaking changes:

• Rename image package to vision.
• TextDataModule base class now implements complete preprocessing logic.
• TextDataModule subclasses only convert source dataset to a common structure.
• Abstraction over cross-attention query creation (QueryProvider).
• Decouple OutputAdapter interface from trainable cross-attention query.
• Implement learned positions encodings as nn.Embedding.
• Move adapters to separate perceiver.model.core.adapter module.
• Rename PerceiverConfig to PerceiverIOConfig
• Rename LitModel base class to LitPerceiverIO.
• LitClassifier.forward now behaves like the wrapped model's forward.
• Object-oriented design of conversion from Hugging Face Perceiver models.
• Major refactoring of PerceiverAR and CausalLanguageModel.
• Move FourierPositionEncoding to perceiver.model.core.position` module.

perceiver-io 0.6.0

Implementation of Perceiver AR including training and inference examples (#20).