mirror of
https://github.com/p-e-w/heretic.git
synced 2026-10-14 20:33:44 +00:00
fix: replace tqdm progress bars with Rich progress bars
ci.yml #47 -Commit
96c7a7d98a
pushed by
vrr
fix: prevent div-by-zero in evaluator when base_refusals is 0 (#225)
ci.yml #44 -Commit
515a7b9eb5
pushed by
vrr
fix: display all abliterable components across layers (#215)
ci.yml #43 -Commit
e26da5e0e6
pushed by
vrr
fix: recheck prefix after inserting predefined (#194)
ci.yml #41 -Commit
303ba9d978
pushed by
vrr
fix: report VRAM usage across all GPUs instead of only the default device (#169)
ci.yml #39 -Commit
4c80c4beb9
pushed by
vrr
fix: produce card for local models with existing readme (#157)
ci.yml #38 -Commit
3a115e280c
pushed by
vrr
fix: use compatible release specifiers for non-dev dependencies
ci.yml #35 -Commit
6017bcd347
pushed by
vrr
ci: upgrade GitHub Actions to latest versions (#137)
ci.yml #33 -Commit
745b582414
pushed by
vrr
Implement Magnitude-Preserving Orthogonal Ablation (#52)
ci.yml #31 -Commit
3525b1ac22
pushed by
vrr
fix: Use file instead of symlink lock (for windows) (#116)
ci.yml #30 -Commit
42f5a9b553
pushed by
vrr
feat: Allow study progress to be saved & resumed (#106)
ci.yml #29 -Commit
ebc22c299e
pushed by
vrr
feat: Refactor save machinery and always allow user to save LoRA (#110)
ci.yml #28 -Commit
c86f49035e
pushed by
vrr
fix: Include kernels (allows MXFP4 to be loaded in MXFP4 instead of upcasting) (#107)
ci.yml #27 -Commit
85a6ec5ecb
pushed by
vrr
feat: add continuous optimization option(latest changes updated) (#76)
ci.yml #24 -Commit
8d44b65670
pushed by
vrr
feat: enumerate all available GPUs on startup (#86)
ci.yml #23 -Commit
92d0c0d551
pushed by
vrr
feat: Add 4-bit loading + LoRA support for low VRAM optimization (#60)
ci.yml #22 -Commit
243f821d93
pushed by
vrr
feat: add max_memory parameter to limit memory usage (#83)
ci.yml #21 -Commit
740aab61ba
pushed by
vrr
fix: skip common response prefix for thinking models
ci.yml #18 -Commit
15781a8a0c
pushed by
vrr