* ci: Bump all GitHub Actions to latest stable versions

checkout@v3 -> v4
cache@v3 -> v4
cache/restore@v3 -> v4
upload-artifact@v4 -> v7
download-artifact@v4 -> v8
setup-uv@v6 -> v8

* ci: Add workflow_dispatch trigger to checks.yml

* fix: Use setup-uv@v7 (v8 major tag not published)

* fix: Use setup-uv@v7 in release.yml too

* fix: Truncate floats to 3 decimal places in notebook comparisons

* fix: Scale-aware float truncation for notebook comparisons

* fix: Significant figures (3) for small-number notebook comparisons

* Resolution for #644 and #341

* Started activation cache improvement

* Full resolution for 210 + a demo notebook

* Resolution for #644 and #341

* Started activation cache improvement

* Full resolution for 210 + a demo notebook

* Resolution for #796, Factored Matrix memory leak

* Resolved #453, underlying issue

* Resolution for Issue #385, added notes about forced eager, added a test to check for future drift

* Added hook introspection mixin for #297

* Made improvements to booting training revisions

* Resolution for #644 and #341

* Started activation cache improvement

* Full resolution for 210 + a demo notebook

* Resolution for #796, Factored Matrix memory leak

* Resolved #453, underlying issue

* Resolution for Issue #385, added notes about forced eager, added a test to check for future drift

* Added hook introspection mixin for #297

* Made improvements to booting training revisions

* Adapter test improvements

* format cleanup

* Resolution for #644 and #341

* Started activation cache improvement

* Full resolution for 210 + a demo notebook

* Resolution for #796, Factored Matrix memory leak

* Resolved #453, underlying issue

* Resolution for Issue #385, added notes about forced eager, added a test to check for future drift

* Added hook introspection mixin for #297

* Made improvements to booting training revisions

* Adapter test improvements

* format cleanup

* Adding a way to display logit vector for #112 and add type hinting for #830

* Type issue resolution – Resolving import confusion between TransformerBridgeConfig module and class, which were separate entities sharing the same name. Files renamed to properly differentiate

* Format and type fixes

* Removed unnecssary assertion

* Fix type of HookedTransformerConfig.device (#1230)

* Fix type of HookedTransformerConfig.device

This is typed as `Optional[str]` but sometimes returns `torch.device`.
Updated the code to just return the `str` instead of wrapping with a
device.

I'm not confident that every function which takes a device will
always be passed a string, so I didn't change functions like
warn_if_mps.

Found while working on #1219

* more cleanup

* 3.0 CI Bugs (#1261)

* Fixing `utils` imports

* skip gated notebooks on PR from forks

* Updating notebooks

* Ensure LLaMA only runs when HF_TOKEN is available

---------

Co-authored-by: jlarson4 <jonahalarson@comcast.net>

* test: add OPT architecture adapter coverage

* chore: rerun CI

---------

Co-authored-by: Brendan Long <self@brendanlong.com>
Co-authored-by: jlarson4 <jonahalarson@comcast.net>

* add Gpt2 MOdel Bridge tests

* removing unused params

* adding missing bos and attn checks

* updating FatoryRegistration test

* formatting via black

* adding custom head GPT2 tests

* Fix type of HookedTransformerConfig.device (#1230)

* Fix type of HookedTransformerConfig.device

This is typed as `Optional[str]` but sometimes returns `torch.device`.
Updated the code to just return the `str` instead of wrapping with a
device.

I'm not confident that every function which takes a device will
always be passed a string, so I didn't change functions like
warn_if_mps.

Found while working on #1219

* more cleanup

* 3.0 CI Bugs (#1261)

* Fixing `utils` imports

* skip gated notebooks on PR from forks

* Updating notebooks

* Ensure LLaMA only runs when HF_TOKEN is available

---------

Co-authored-by: jlarson4 <jonahalarson@comcast.net>

* Add Qwen3.5 support and improve adapter validation

- Document Qwen3.5 text-only model usage in special_cases.md
- Update pyproject.toml to include transformers dependency for Qwen3.5
- Enhance unit tests for Qwen3.5 architecture detection and dependency handling
- Modify transformers.py to use prepared model config
- Implement stricter validation in Qwen3_5ArchitectureAdapter for model compatibility

* update lock file

* Declare packaging for Qwen3.5 extra

* Fix Qwen3.5 format and Mamba cache typing

* Fix bridge component compatibility checks

---------

Co-authored-by: Brendan Long <self@brendanlong.com>
Co-authored-by: jlarson4 <jonahalarson@comcast.net>

* feat: External architecture adapter registration and entry-point discovery

* docs: External adapter registration guide with examples

* test: ArchitectureAdapterFactory registration, selection, and entry-point tests

* fix: Clarify matching requirement and add warnings for failed discovery

* fix: Make register_adapter doctest self-contained with inline adapter class

* fix: Address Copilot review - alias, loop isolation, test state leak

* fix: Apply black formatting to factory file

* fix: use select_architecture_adapter in register_adapter doctest

* fix: guard native adapter override in discover_entry_points

* test: drop TestSupportedArchitectures

* test: add entry-point discovery tests

* docs: document native adapter override prevention

* fix: apply black formatting to test file

* fix: handle None ep.dist in entry-point guard warning

* fix: apply black formatting to factory file

* Fixed double scaling of gemma, bump transformers pin to a 5.x version

* testing and formatting