Commit message (Collapse) | Author | Age | Files | Lines | ||
---|---|---|---|---|---|---|
... | ||||||
* | Use context manager for EMA, on_train/eval hooks | Volpeon | 2023-01-06 | 1 | -54/+66 | |
| | ||||||
* | Add contextmanager to EMAModel to apply weights temporarily | Volpeon | 2023-01-06 | 1 | -33/+24 | |
| | ||||||
* | Log EMA decay | Volpeon | 2023-01-05 | 1 | -1/+2 | |
| | ||||||
* | Added EMA to TI | Volpeon | 2023-01-05 | 1 | -1/+58 | |
| | ||||||
* | Update | Volpeon | 2023-01-05 | 1 | -7/+7 | |
| | ||||||
* | Fix | Volpeon | 2023-01-05 | 1 | -8/+11 | |
| | ||||||
* | Various cleanups | Volpeon | 2023-01-05 | 1 | -65/+21 | |
| | ||||||
* | Update | Volpeon | 2023-01-04 | 1 | -1/+1 | |
| | ||||||
* | Various updates | Volpeon | 2023-01-04 | 1 | -0/+8 | |
| | ||||||
* | Better eval generator | Volpeon | 2023-01-04 | 1 | -6/+6 | |
| | ||||||
* | Fixed reproducibility, more consistant validation | Volpeon | 2023-01-04 | 1 | -5/+19 | |
| | ||||||
* | Don't use vector_dropout by default | Volpeon | 2023-01-03 | 1 | -1/+1 | |
| | ||||||
* | Added vector dropout | Volpeon | 2023-01-03 | 1 | -5/+19 | |
| | ||||||
* | Fixed LR finder | Volpeon | 2023-01-02 | 1 | -3/+3 | |
| | ||||||
* | Update | Volpeon | 2023-01-02 | 1 | -3/+12 | |
| | ||||||
* | Fix | Volpeon | 2023-01-02 | 1 | -4/+2 | |
| | ||||||
* | Save args before training, too | Volpeon | 2023-01-02 | 1 | -0/+4 | |
| | ||||||
* | Fix | Volpeon | 2023-01-02 | 1 | -2/+1 | |
| | ||||||
* | Update | Volpeon | 2023-01-02 | 1 | -6/+6 | |
| | ||||||
* | Improved one cycle scheduler | Volpeon | 2023-01-02 | 1 | -1/+1 | |
| | ||||||
* | Update | Volpeon | 2023-01-01 | 1 | -1/+1 | |
| | ||||||
* | Updates | Volpeon | 2023-01-01 | 1 | -26/+25 | |
| | ||||||
* | Fixed accuracy calc, other improvements | Volpeon | 2023-01-01 | 1 | -3/+33 | |
| | ||||||
* | Update | Volpeon | 2022-12-31 | 1 | -0/+1 | |
| | ||||||
* | Simplified multi-vector embedding code | Volpeon | 2022-12-31 | 1 | -3/+2 | |
| | ||||||
* | Fixes | Volpeon | 2022-12-31 | 1 | -2/+1 | |
| | ||||||
* | Added multi-vector embeddings | Volpeon | 2022-12-31 | 1 | -44/+44 | |
| | ||||||
* | Misc improvements | Volpeon | 2022-12-30 | 1 | -29/+18 | |
| | ||||||
* | Training script improvements | Volpeon | 2022-12-30 | 1 | -5/+38 | |
| | ||||||
* | Training improvements | Volpeon | 2022-12-29 | 1 | -6/+15 | |
| | ||||||
* | Integrated updates from diffusers | Volpeon | 2022-12-28 | 1 | -8/+13 | |
| | ||||||
* | Improved learning rate finder | Volpeon | 2022-12-27 | 1 | -7/+3 | |
| | ||||||
* | Added validation phase to learn rate finder | Volpeon | 2022-12-27 | 1 | -4/+8 | |
| | ||||||
* | Added learning rate finder | Volpeon | 2022-12-27 | 1 | -87/+87 | |
| | ||||||
* | Set default dimensions to 768; add config inheritance | Volpeon | 2022-12-26 | 1 | -6/+5 | |
| | ||||||
* | Code simplifications, avoid autocast | Volpeon | 2022-12-25 | 1 | -1/+7 | |
| | ||||||
* | Fixed Textual Inversion | Volpeon | 2022-12-24 | 1 | -2/+8 | |
| | ||||||
* | Training update | Volpeon | 2022-12-24 | 1 | -6/+4 | |
| | ||||||
* | Update | Volpeon | 2022-12-24 | 1 | -1/+9 | |
| | ||||||
* | Better dataset prompt handling | Volpeon | 2022-12-23 | 1 | -1/+1 | |
| | ||||||
* | num_class_images is now class images per train image | Volpeon | 2022-12-23 | 1 | -1/+1 | |
| | ||||||
* | Improved class prompt handling | Volpeon | 2022-12-23 | 1 | -20/+9 | |
| | ||||||
* | Fix | Volpeon | 2022-12-23 | 1 | -3/+1 | |
| | ||||||
* | Fixed Textual Inversion | Volpeon | 2022-12-22 | 1 | -5/+7 | |
| | ||||||
* | Improved Textual Inversion: Completely exclude untrained embeddings from ↵ | Volpeon | 2022-12-22 | 1 | -18/+6 | |
| | | | | training | |||||
* | Moved common training code into separate module | Volpeon | 2022-12-21 | 1 | -120/+55 | |
| | ||||||
* | Fix training | Volpeon | 2022-12-20 | 1 | -3/+2 | |
| | ||||||
* | Fix Textual Inversion dataset filtering | Volpeon | 2022-12-20 | 1 | -1/+5 | |
| | ||||||
* | Improved dataset prompt handling, fixed | Volpeon | 2022-12-19 | 1 | -0/+1032 | |