Skip to content

Pull requests: p-e-w/heretic

New pull request New
34 Open 179 Closed
34 Open 179 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

fix: stop retrying non-recoverable dtype errors
#435 opened Aug 27, 2026 by xuyua9 Loading…
1
fix: improve dtype fallback error reporting
#434 opened Aug 27, 2026 by xuyua9 Loading…
1
feat: preserve provenance for local datasets
#433 opened Aug 24, 2026 by aim9sour Loading…
1
feat: TPU support via PyTorch/XLA (auto-detection, FSDP, XLA-safe merge/export)
#431 opened Aug 22, 2026 by instax-dutta Loading…
1
feat: A better logic for detecting the thinking prefix in the response
#423 opened Aug 6, 2026 by Vinay-Umrethe Collaborator Loading…
11
feat(ara): add DiffusionGemma block-diffusion model support
#421 opened Aug 3, 2026 by umran666 Contributor • Draft
1
feat/fix(ara): Save-time model tensor integrity check
#420 opened Jul 29, 2026 by hpnyaggerman Loading…
9
wip: binary llm refusal judge
#419 opened Jul 23, 2026 by red40maxxer Contributor • Draft
2
fix: reject empty scorer prompt sets
#408 opened Jul 9, 2026 by Bortlesboat Contributor Loading…
5
fix: handle positional model before CLI options
#406 opened Jul 9, 2026 by Bortlesboat Contributor Loading…
13
narrow batch-size benchmark except to (RuntimeError, ValueError, TypeError, OSError)
#396 opened Jun 23, 2026 by HrachShah Loading…
3
fix: SVD convergence crash via CPU LAPACK fallback
#393 opened Jun 23, 2026 by umran666 Contributor Loading…
1
8
feat: add support for Qwen3Next shared experts
#391 opened Jun 18, 2026 by umran666 Contributor Loading…
1
Use adaptive per-method batch sizing, drop startup benchmark
#390 opened Jun 17, 2026 by kali113 Contributor Loading…
1
17
feat: generate model card on local save (#258)
#388 opened Jun 17, 2026 by umran666 Contributor Loading…
2
fix: capture residual hidden states via forward hooks instead of output_hidden_states
#384 opened Jun 16, 2026 by kali113 Contributor Loading…
1
9
feat: diffusion LLM abliteration support (DiffusionGemma)
#378 opened Jun 13, 2026 by anurag12-webster • Draft
1
21
fix: preserve Hugging Face token in session to avoid repeated prompt (#142)
#366 opened Jun 10, 2026 by umran666 Contributor Loading…
7
binary search for batch size
#362 opened Jun 7, 2026 by NoahOksuz Loading…
5
feat: add config.nosyco.toml
#349 opened Jun 4, 2026 by UnstableLlama Contributor Loading…
22
docs: add contributing guide
#347 opened Jun 3, 2026 by zichen0116 Loading…
2
feat: Write README.md for LoRA adapter as well
#314 opened Apr 28, 2026 by anrp Contributor Loading…
5
perf(ara): cache original weights to skip disk reload between trials
#313 opened Apr 26, 2026 by ghecko Loading…
6
feat: Support local weight paths and improve memory management for new architectures
#307 opened Apr 21, 2026 by antonvice Loading…
5
feat: add LLM-based judge for refusal classification
#244 opened Mar 21, 2026 by hpnyaggerman • Draft
6 tasks done
16
ProTip! Follow long discussions with comments:>50.