aboutsummaryrefslogtreecommitdiff
path: root/extensions-builtin/Lora/network_oft.py
AgeCommit message (Collapse)Author
2024-02-26Merge pull request #14973 from AUTOMATIC1111/Fix-new-oft-boftAUTOMATIC1111
Fix the OFT/BOFT bugs when using new LyCORIS implementation
2024-02-19Merge pull request #14871 from v0xie/boftAUTOMATIC1111
Support inference with LyCORIS BOFT networks
2024-01-22fix: wrong devices for eye and constraintv0xie
2024-01-05Fix dtype casting for OFT moduleKohaku-Blueleaf
2023-12-14better naming and correct order for device.Kohaku-Blueleaf
2023-12-14remove debug printKohaku-Blueleaf
2023-12-14Use self.scale instead of custom finalizeKohaku-Blueleaf
2023-12-14Fix network_oftKohaku-Blueleaf
2023-11-15feat: LyCORIS/kohya OFT network supportv0xie
2023-11-15fix: ignore calc_scale() for COFT which has very small alphav0xie
2023-11-04refactor: remove unused functionv0xie
2023-11-04refactor: use same updown for both kohya OFT and LyCORIS diag-oftv0xie
2023-11-03refactor: move factorization to lyco_helpers, separate calc_updown for kohya ↵v0xie
and kb
2023-11-03skip multihead attn for nowv0xie
2023-11-02no idea what i'm doing, trying to support both type of OFT, kblueleaf ↵v0xie
diag_oft has MultiheadAttn which kohya's doesn't?, attempt create new module based off network_lora.py, errors about tensor dim mismatch
2023-11-01test implementation based on kohaku diag-oft implementationv0xie
2023-10-22style: conform stylev0xie
2023-10-22fix: multiplier applied twice in finalize_updownv0xie
2023-10-22refactor: remove used OFT functionsv0xie
2023-10-21fix: use merge_weight to cache valuev0xie
2023-10-21style: cleanup oftv0xie
2023-10-21fix: support multiplier, no forward pass hookv0xie
2023-10-21fix: return orig weights during updown, merge weights before forwardv0xie
2023-10-21refactor: use forward hook instead of custom forwardv0xie
2023-10-19style: fix ambiguous variable namev0xie
2023-10-19style: formattingv0xie
2023-10-19refactor: fix constraint, re-use get_weightv0xie
2023-10-18faster by calculating R in updown and using cached R in forwardv0xie
2023-10-18faster by using cached R in forwardv0xie
2023-10-18inference working but SLOWv0xie
2023-10-17wip incorrect OFT implementationv0xie
generated by cgit v1.2.1 (git 2.18.0) at 2025-05-12 19:23:02 +0000