Skip to content

v1.10.2
Compare
Choose a tag to compare
@mart-r mart-r released this 28 Feb 16:47
· 16 commits to production since this release
v1.10.2
e311d36
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.

This patch release mostly addresses another issue with the DeID model(s).

The underlying RoBERTa models have a token limit (512) and because of that later parts of larger documents would fail to de-identify.
This Release (or more specifically, PR #405) fixes that issue by allowing the user to specify the overlapping tokens (and defaults to 5).

What's Changed

  • CU-8693v3tt6 SOMED opcs refset selection by @mart-r in #402
  • CU-8693v6epd: Move typing imports away from pydantic by @mart-r in #403
  • CU-8693qx9yp Deid chunking - hugging face pipeline approach by @shubham-s-agarwal in #405

New Contributors

Full Changelog: v1.10.1...v1.10.2

Contributors

mart-r and shubham-s-agarwal
Assets 2