Sae-lens

Fix

* fix: fixing bug with cfg loading for fine-tuning (241) ([`5a88d2c`](https://github.com/jbloomAus/SAELens/commit/5a88d2c2f3086393a35e2b8cc5d356e161494e9e))

Unknown

* Update deploy_docs.yml

Removed the Debug Info step that was causing issues. ([`71fd509`](https://github.com/jbloomAus/SAELens/commit/71fd509bf719481b2e89718cca53a39f9a113c80))

* More tests for the negative case ([`a0b0f54`](https://github.com/jbloomAus/SAELens/commit/a0b0f547df51e17568f92d15390c2356f05beed5))

* Upped version ([`845d5d7`](https://github.com/jbloomAus/SAELens/commit/845d5d75f3033ce325f32e2d3c4c9421e15f847f))

* Added tests ([`e5ff793`](https://github.com/jbloomAus/SAELens/commit/e5ff793f32c1f3d6644c97127754fbc64ec631b8))

Fix

* fix: Trainer eval config will now respect trainer config params (238)

* Trainer eval config will now respect trainer config params

* Corrected toml version ([`5375505`](https://github.com/jbloomAus/SAELens/commit/5375505f86785c96157719f6540f5e0dba541fe6))

Unknown

* Neuronpedia Autointerp/Explanation Improvements (239)

* Neuronpedia autointerp API improvements: new API, new flags for save to disk and test key, fix bug with scoring disabled

* Ignore C901 ([`ba7d218`](https://github.com/jbloomAus/SAELens/commit/ba7d21815c6429e1b92b36ebc030278acdea8c5e))

* Fixed toml file ([`8211cac`](https://github.com/jbloomAus/SAELens/commit/8211cac570a7c8baf0dbcc61e6cf72d4635c05a3))

* Ensured that even detatched SAEs are returned to former state ([`90ac661`](https://github.com/jbloomAus/SAELens/commit/90ac66179a3bfec9324161bdd86f4f6245aa3791))

* Added use_error_term functionality to run_with_x functions ([`1531c1f`](https://github.com/jbloomAus/SAELens/commit/1531c1f72d21b8068ad78ae60ec08af9da1d3751))

* Added use_error_term to hooked sae transformer ([`d172e79`](https://github.com/jbloomAus/SAELens/commit/d172e792441bb8c233e50cf5c7af74b8c37b865f))

* Trainer will now fold and log estimated norm scaling factor (229)

* Trainer will now fold and log estimated norm scaling factor after doing fit

* Updated tutorials to use SAEDashboard

* fix: sae hook location (235)

* 3.12.2

Automatically generated by python-semantic-release

* fix: sae to method (236)

* 3.12.3

Automatically generated by python-semantic-release

* Trainer will now fold and log estimated norm scaling factor after doing fit

* Added functionality to load and fold in precomputed scaling factors from the YAML directory

* Fixed toml

---------

Co-authored-by: Joseph Bloom <69127271+jbloomAususers.noreply.github.com>
Co-authored-by: github-actions <github-actionsgithub.com> ([`8d38d96`](https://github.com/jbloomAus/SAELens/commit/8d38d96056802a8025c60d0296624878777a9159))

* Update README.md ([`172ad6a`](https://github.com/jbloomAus/SAELens/commit/172ad6a099504acc90c4c3d5c29a3084367f560f))

Fix

* fix: sae to method (236) ([`4df78ea`](https://github.com/jbloomAus/SAELens/commit/4df78ea6b44affe0ee0b3f033f4d1836a7f4a3e9))

Fix

* fix: sae hook location (235) ([`94ba11c`](https://github.com/jbloomAus/SAELens/commit/94ba11c13bdf96740e6217d8c87af47237f89ca5))

Unknown

* Updated tutorials to use SAEDashboard ([`db89dbc`](https://github.com/jbloomAus/SAELens/commit/db89dbc813e5a64a7609165474e3c66e2b537925))

* Merge branch &39;JoshEngels-Evals&39; ([`fe66285`](https://github.com/jbloomAus/SAELens/commit/fe66285d75bf7b0617a8d0aedef010e74b972677))

* Removed redundant lines ([`29e3aa2`](https://github.com/jbloomAus/SAELens/commit/29e3aa2b1a5b808dc32c5fc8187f5d7be808c7e3))

* Merge branch &39;Evals&39; of https://github.com/JoshEngels/SAELens into JoshEngels-Evals ([`7b84053`](https://github.com/jbloomAus/SAELens/commit/7b84053027f9113ac3cf61946c5acc0367298e05))

Fix

* fix: force release of dtype_fix ([`bfe7feb`](https://github.com/jbloomAus/SAELens/commit/bfe7feb10d987f39a9e0d00f5595869a64de2b5b))

Unknown

* Merge pull request 225 from jbloomAus/dtype_fix

fix: load_from_pretrained should not require a dtype nor default to float32 ([`71d9da8`](https://github.com/jbloomAus/SAELens/commit/71d9da80a0e48a56f8d2394aecb0383a5f2a0cc6))

* TrainingSAE should: 1) respect device override and 2) not default to float32 dtype, and instead default to the SAE&39;s dtype ([`a4a1c46`](https://github.com/jbloomAus/SAELens/commit/a4a1c469fff7e3b0685ced4c0632808fc3690359))

* load_from_pretrained should not require a dtype nor default to float32 ([`a485dc0`](https://github.com/jbloomAus/SAELens/commit/a485dc00d45bd3a69add64ccdf0cb5697d8b77ad))

* Fix SAE failing to upload to wandb due to artifact name. (224)

* Fix SAE artifact name.

* format

---------

Co-authored-by: Joseph Bloom <jbloomausgmail.com> ([`6ae4849`](https://github.com/jbloomAus/SAELens/commit/6ae4849bb1947b6c83fd438b9dd00a8172f1f4b8))

Feature

* feat: use TransformerLens 2 (214)

* Updated pyproject.toml to use TL ^2.0, and to use fork of sae-vis that also uses TL ^2.0

* Removed reliance on sae-vis

* Removed neuronpedia tutorial

* Added error handling for view operation

* Corrected formatting ([`526e736`](https://github.com/jbloomAus/SAELens/commit/526e736b937f95333969c33c83d2500dacab43d7))

Unknown

* Fix/allow device override (221)

* Forced load_from_pretrained to respect device and dtype params

* Removed test file ([`697dd5f`](https://github.com/jbloomAus/SAELens/commit/697dd5f0911d145d6a7c956e29bbcf28cf9fee38))

* Fixed hooks for single head SAEs (219)

* included zero-ablation-hook for single-head SAEs

* fixed a typo in single_head_replacement_hook ([`3bb4f73`](https://github.com/jbloomAus/SAELens/commit/3bb4f73933984278bd4fb0446ead630f6edad600))