v0.9.12

Important

This release is identical to v0.9.11, except for a change made to fix a flakey test that prevented our CI/CD workflow from completing.

What's Changed

• Allow URLPath files to be missing when cleaning up inputs by @mattt in #1763
• updated notebook to lab ... by @gschian0 in #846
• Add test for using Torch & Tensorflow together by @bfirsh in #1023
• fix: 🐛 a better fix for #1635 by @EduarteXD in #1656
• Fix integration test deadlocks / timeouts by @mattt in #1767
• Make cog local runner match /tmp/shm size on Replicate by @nelsonjchen in #1540
• Restructure CI workflow (again) by @mattt in #1768
• Add passing test that includes Torch and TensorFlow together by @mattt in #1123
• add Etelis as a contributor for code by @allcontributors in #1770
• add gschian0 as a contributor for doc by @allcontributors in #1771
• Support minor torch version by @andreasjansson in #1769
• Warn if cog.yaml or predictor are missing by @asingh9530 in #1470
• Fix possible data corruption due to race condition in StreamRedirector by @mattt in #1773
• Relax linter requirements for CI workflow, resolve new Ruff warnings by @mattt in #1775
• Fix cog build --use-cog-base-image with minor and patch versions of Python and Torch by @mattt in #1774
• Rewrite HTTP API documentation by @mattt in #1761
• Reduce size of output to reduce likelihood of OOM in unit test by @mattt in #1776
• Handle predictors with deferred annotations by @mattt in #1772

New Contributors

• @gschian0 made their first contribution in #846
• @asingh9530 made their first contribution in #1470

Full Changelog: v0.9.10...v0.9.12

v0.9.11

Important

The tagged commit was unable to complete its CI/CD workflow due to a flakey test. These changes are available in Cog v0.9.12.

v0.9.10

What's Changed

• Set a minimum CUDA version for Tensorflow compatibility by @marcusmartins in #1715
• Validate minor version for CUDA and Python by @marcusmartins in #1721
• Document Secret type in Python API docs by @mattt in #1724
• Update default python version to 3.12 by @marcusmartins in #1722
• Improve cog base image handling by @marcusmartins in #1699
• Unhide the use-cog-base-image-flag by @marcusmartins in #1725
• Update environment variables documentation by @mattt in #1749
• Modify internal base image struct by @marcusmartins in #1712
• Set 10 minute timeout for test-integration CI job by @mattt in #1755
• add necessary headers to curl by @Kettukaa in #1751
• Change config's type to TypedDict instead of Dict. by @JacekDuszenko in #1527
• Bump to latest supported python & torch in README.md by @nevillelyh in #1655
• Use timestamps for tmp build directories by @mattt in #1756
• Install Script by @chigozienri in #1558
• Update install script instructions by @mattt in #1757
• Document how Cog raises CancelationException in predict when a prediction is canceled by @mattt in #1754
• Implement pre-build repository check for Replicate images by @Etelis in #1733
• Improve CUDA compatibility checks by @nevillelyh in #1690
• Add label for base image's last layer's diff id to cog image by @jessie-young in #1747
• Reduce unnecessary work in CI workflow by @mattt in #1760
• Update HTTP server to make host configurable by @mattt in #1748
• Document --host option by @mattt in #1762

New Contributors

• @Kettukaa made their first contribution in #1751
• @nevillelyh made their first contribution in #1655
• @chigozienri made their first contribution in #1558
• @Etelis made their first contribution in #1733
• @jessie-young made their first contribution in #1747

Full Changelog: v0.9.9...v0.9.10

v0.10.0-alpha13

What's Changed

• add batch_size metric by @technillogue in #1750

Full Changelog: v0.10.0-alpha12...v0.10.0-alpha13

v0.10.0-alpha12

What's Changed

• log traceback and set error properly by @technillogue in #1734

Full Changelog: v0.10.0-alpha11...v0.10.0-alpha12

v0.10.0-alpha11

What's Changed

• Fix upload logging by @mattt in #1707
• fix upload redirect handling by @technillogue in #1714

Full Changelog: v0.10.0-alpha10...v0.10.0-alpha11

v0.9.9

What's Changed

• Include prediction id upload request by @aron in #1667
• update cog init to include a GitHub Actions workflow file by @zeke in #1671
• document how cog.yaml image works with cog push by @zeke in #1679
• Fix linter errors by @mattt in #1691
• fixing install logic to reenable no cuda base image by @daanelson in #1688
• Increase Hypothesis test deadlines by @marcusmartins in #1700
• Quick fix for Apple Silicon M1 support in run by @timfallmk in #1704
• Set missing Env argument in call to NewPredictor on retry by @mattt in #1709
• Add top-level concurrency setting to cog.yaml schema by @mattt in #1703
• Add --env/-e flags to cog train subcommand by @mattt in #1710
• Propagate trace context to webhook and upload requests by @aron in #1698

New Contributors

• @aron made their first contribution in #1667
• @daanelson made their first contribution in #1688
• @marcusmartins made their first contribution in #1700
• @timfallmk made their first contribution in #1704

Full Changelog: v0.9.8...v0.9.9

v0.10.0-alpha10

What's Changed

• Backport Secret type to async branch by @mattt in #1706

Full Changelog: v0.10.0-alpha9...v0.10.0-alpha10

v0.9.9-alpha2

What's Changed

• update cog init to include a GitHub Actions workflow file by @zeke in #1671
• document how cog.yaml image works with cog push by @zeke in #1679
• Fix linter errors by @mattt in #1691
• fixing install logic to reenable no cuda base image by @daanelson in #1688

New Contributors

• @daanelson made their first contribution in #1688

Full Changelog: v0.9.9-alpha...v0.9.9-alpha2

v0.10.0-alpha9

This release fixes being able to push concurrent models with cog.