Skip to content

Pull requests: decoderesearch/SAELens

New pull request New
25 Open 521 Closed
25 Open 521 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

feat: add quadratic sparcity loss from GemmaScope 2 for JumpReluSAE
#732 opened Sep 3, 2026 by Estomeee Loading…
10 tasks done
fix: load trainer state onto the configured device
#731 opened Aug 31, 2026 by SuryanshSS1011 Contributor Loading…
8 tasks done
1
feat(training): add held-out evaluation data
#730 opened Aug 30, 2026 by janmenjayap • Draft
7 of 16 tasks
1
Add covariance whitening normalization option (#647)
#697 opened Jun 7, 2026 by azrabano23 Loading…
4
Fix incorrect buffer size calculations in ActivationsStore
#692 opened Jun 6, 2026 by danra Contributor Loading…
7 of 8 tasks
7
fix: _validate_seqpos rejects valid step_size of 1 in seqpos_slice
#681 opened May 29, 2026 by robbiebusinessacc Contributor Loading…
feat: extending cache activation runner options for tokenization, batching, and shuffling
#612 opened Jan 3, 2026 by chanind Collaborator Loading…
14
feat: Add Swanlab logging backend
#605 opened Dec 24, 2025 by wantbook-book Loading…
3 of 15 tasks
1
5
fix: top 10 activations tutorial
#582 opened Nov 5, 2025 by rosepolle Loading…
2 tasks done
docs: example — Cross-SAE alignment via FeatureMatch (cosine + top-k)
#572 opened Oct 21, 2025 by BentleyRolling Loading…
3 tasks done
Willz/smm optimization [DRAFT]
#558 opened Oct 6, 2025 by chanind Collaborator • Draft
1
opt: Replace COO with CSR sparse tensors for bfloat16/float16 support & faster speedups [IN PROGRESS]
#557 opened Oct 5, 2025 by wz-ml Contributor Loading…
9 of 13 tasks
fix: GatedSAE needs to prevent the model from cheating L0 over steps
#546 opened Sep 8, 2025 by jasonrichdarmawan Loading…
4 of 15 tasks
1
1
fix: detach grads from gated SAE for aux loss
#545 opened Sep 4, 2025 by chanind Collaborator Loading…
6 of 11 tasks
1
1
adds dictionary_learning converters
#477 opened May 10, 2025 by anthonyduong9 Collaborator • Draft
15 tasks
1
feat: Add support for acausal crosscoders.
#475 opened May 7, 2025 by mkbehr Contributor • Draft
5 of 9 tasks
Move from manual sharding to HF dataset builder.
#391 opened Nov 25, 2024 by tom-pollak Contributor • Draft
8 tasks done
2
fix: Added hf_revision to runner config
#384 opened Nov 24, 2024 by tom-pollak Contributor • Draft
7 of 9 tasks
feat: Add linear probe trainer
#356 opened Oct 31, 2024 by tom-pollak Contributor Loading…
2 of 9 tasks
3
extracts repeated code to load_pretrained_saes_yaml()
#338 opened Oct 18, 2024 by anthonyduong9 Collaborator • Draft
15 tasks
Encode with slice
#334 opened Oct 16, 2024 by callummcdougall Contributor Loading…
14
Support seqpos slicing
#294 opened Sep 18, 2024 by callummcdougall Contributor • Draft
6
extend quick list creation options
#286 opened Sep 12, 2024 by TDulka Loading…
5 of 11 tasks
14
Memmap shuffling
#216 opened Jul 7, 2024 by Lewington-pitsos Contributor Loading…
9 of 13 tasks
11
[WIP] feat: add mlp transcoders
#183 opened Jun 15, 2024 by dtch1997 Contributor • Draft
2 of 13 tasks
1
6
ProTip! What’s not been updated in a month: updated:<2026-08-08.