v0.5.5

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Fixes a bug that prevented calling on_error in extensions when an exception happened (#490).
• Open snapshot files in binary mode (#483).
• Adds an unstrip_tensor script to improve compatibility with ONNX runtimes (#477)
• Completes the typing support over the codebase.
• Corrects several minor issues with the training API.

See the list of merged pull-requests for the details.

v0.5.4

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Fixes the import error when onnx package is not installed

See the list of merged pull-requests for the details.

v0.5.3

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Adds a CodeBlock API to simplify writing custom Logic classes (#426).
• Fixes bugs related to Writer objects termination and Tensorboard cleanup.
• Improves typing support over the codebase.
• Store ppe.__version__ in snapshot and warn if the version is different.

Backward-incompatible changes

• Mandatory use of ppe.to for Trainer APIs. Now an explicit call ppe.to is needed in the scripts before running the trainer.

See the list of merged pull-requests for the details.

v0.5.2

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Adds create_distributed_subset_indices to efficiently scatter datasets across distributed nodes.
• Adds ppe.nn.Ensure module to assert shape or dtypes in PyTorch code.
• Reduces the snapshot size when using append=True (output logs in JSON Lines or YAML) in LogReport.
• Makes Trainer and Evaluator public APIs so that users can add type annotations to user-defined Logic and Handler classes with these classes.
• Deprecates async option from the trainer and introduces a DeferredResult API instead.
• Stores time_summary in TLS to avoid bugs related to multiprocessing.

Backward-incompatible changes

• Process-wide TimeSummary singleton instance (ppe.profiler.time_summary) has been removed as it was not thread-safe and causing issues on multiprocessing/threading. In this release, you can obtain a thread-local TimeSummary instance using ppe.profiler.get_time_summary().

See the list of merged pull-requests for the details.

v0.5.1

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Adds ppe.onnx.grad API to emit Gradient operator
• Improves tensor stripping in ppe.onnx
• Fixes ProgressBar extension always says inf iters/sec
• Fixes several type annotation issues

See the list of merged pull-requests for the details.

v0.5.0

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Adds a device-agnostic trainer and evaluator. Check the documentation.
• Adds typing support through the codebase.
• Adds online documentation.
• Shows evaluator name in the progress bar.

Backward-incompatible changes

• Drops support for PyTorch 1.7
• Deprecates the LazyLinear and LazyConvNd modules since they have been upstreamed to PyTorch.

See the list of merged pull-requests for the details.

v0.4.5

This release includes the following enhancements and bug fixes:

• Fixes the Profiler API introduced in the v0.4.3 release to avoid initializing multiprocessing contexts and threads on import pytorch_pfn_extras.
• Profiler now reports min/max times.

See the list of merged pull-requests for the details.

v0.4.4

This release includes the following enhancements and bug fixes:

• Restores support of Python 3.6 and 3.7, which was unexpectedly broken in the v0.4.3 release.
• Fixes broken styles on the documentation website.

See the list of merged pull-requests for the details.

v0.4.3

This release includes the following enhancements and bug fixes:

• Adds support to generate documentation using Read The Docs
• Adds logging support using the ppe.logging.get_logger interface.
• Adds a profiler to measure timing in annotated code sections. (ppe.training.extensions.ProfileReport and ppe.profiler.TimeSummary.report)
• Adds type annotations to several modules.

See the list of merged pull-requests for the details.

v0.4.2

This release includes the following enhancements and bug-fixes:

• Adds support for PyTorch 1.9, with this release the minimum supported PyTorch version is 1.7.
• Adds support for Python 3.9.
• Fixes issues with LazyModules when managing the state_dict.
• Fixes issues with temporal backup files for snapshots.

See the list of merged pull-requests for the details.