v0.6.6

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Support PyTorch 2.0 #645
• Add a Logic class for clousure mode optimization #657
• Enable new PyTorch autocast options #649

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

v0.6.5

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Fix NameError when imported in ipykernel environment #645

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

v0.6.4

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Add interface to trace events in custom runtimes #584
• Support for ONNX 1.13 #639

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

v0.6.3

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Fix circular import in ppe.onnx
• Improve compatibility with PyTorch 1.13 in ppe.onnx

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

v0.6.2

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Support PyTorch 1.13 (#622)
• Several improvements to PPE ONNX exporter PFTO (#602, #605, #607)
• Add ppe.torchscript utility module (#603, #612)

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

v0.6.1

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Fix import failure when user account does not exist (#597)
• Ignore runtime class for torch.save/load (#588)

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

v0.6.0

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Supports PyTorch 1.12
• Adds an extension to easily post to Slack using slack_sdk or webhook URL (#526)
• Allows to export to ONNX modules compiled using torch.jit.ScriptedModule (#553)
• Support for distributed evaluators (#551 thanks @belltailjp!)
• Autoselect the correct extensions for PrintReport and ProgressBar when running in notebooks (#525)
• Multiple corrections to PFTO

Using the Slack and SlackWebhook extensions (#526)

It is now possible to report the training process to Slack. For this, two new extensions are provided; Slack uses the slack_sdk python package and allows to post messages and upload files. SlackWebhook is a simpler version that doesn't have any requirements but can only post simple messages.

An example of how to use these extensions is as follows:

extensions.Slack(
    channel="experiment-progress",
    msg="Epoch #{manager.epoch}: loss = {val/loss}",

    # Surround the username with <> to mention.
    end_msg="{default} \n <@username> Check out the result!",

    # Upload any artifacts generated during the training.
    filenames=['result/statistics.png'],
    # You can specify when to upload these files.
    # e.g., only at the final epoch:
    # upload_trigger=(max_epochs, 'epoch'),

    # Provide a Slack bot token to access the Slack API.
    # If not specified env var `SLACK_BOT_TOKEN` will be used.
    token=slack_token,
)

# Simpler webhook based version
extensions.SlackWebhook(
    url="https://hooks.slack.com/services/Txxxxx.....",
    msg="Epoch #{manager.epoch}: loss = {val/loss}",
    end_msg="{default} \n <@username> Check out the result!",
)

Backward-incompatible changes

• Drop PyTorch 1.8 support.
• Extension.finalize now takes ExtensionsManager as an argument (#556). Users implementing finalize in their own Extension need to be updated.
• ExtensionsManager and Extensions will now be finalized when an error occurs during the Trainer run.
• Remove asynchronous training from Handler (#536)

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

v0.5.8

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Enable to use multiple optimizers in CodeBlock #533 (thanks @mhiroaki!)
• Add documentation for _RuntimeRegistry and CodeBlock #523
• Fix onnx export of torch.cat for pfto #531 (thanks @take-cheeze!)
• Add profiler option to create_trainer to allow using torch.profiler in Trainer #527

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

v0.5.7

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Add a pure python torch.onnx.export implementation #484 (thanks @take-cheeze!)
• Best observation extension #518 (thanks @belltailjp!)
• Add check_grad option to FailOnNonNumber extension #502

Backward-incompatible changes

• Change out_dir to better reflect the fs needs (in order to improve the interoperability with _PosixFileSystem manager) #486

This changes the default value of out_dir to '' (which means the current directory) instead of None when creating the Writer objects. If you create a Writer instance without specifying the out_dir argument and passing it to an Extension, the output directory will no longer be inherited from the Trainer. Please consider specifying out_dir when creating a Writer instance.

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

v0.5.6

Enhancements and bug-fixes

This release includes the following enhancements and bug fixes:

• Support torch 1.10 (#497, #413)
• Add helper function to initialize torch.distributed in OpenMPI environments (#491)

Backward-incompatible changes

• Disable profiler by default in Trainer and ExtensionsManager. Now enable_profile=True kwarg is needed. (#498)

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