Commit message (Expand) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Update | Volpeon | 2023-06-24 | 2 | -5/+10 |
* | Remove convnext | Volpeon | 2023-06-22 | 4 | -458/+0 |
* | Update | Volpeon | 2023-06-21 | 8 | -92/+176 |
* | Avoid model recompilation due to varying prompt lengths | Volpeon | 2023-05-16 | 1 | -8/+15 |
* | Support LoRA training for token embeddings | Volpeon | 2023-05-16 | 1 | -0/+3 |
* | Fix for latest PEFT | Volpeon | 2023-05-12 | 1 | -145/+0 |
* | Fix | Volpeon | 2023-04-16 | 1 | -1/+1 |
* | Fix | Volpeon | 2023-04-16 | 3 | -71/+157 |
* | Fix | Volpeon | 2023-04-16 | 1 | -16/+18 |
* | Fix | Volpeon | 2023-04-16 | 2 | -23/+21 |
* | Update | Volpeon | 2023-04-16 | 2 | -24/+38 |
* | Fix | Volpeon | 2023-04-15 | 1 | -1/+1 |
* | Fix | Volpeon | 2023-04-15 | 2 | -5/+5 |
* | TI via LoRA | Volpeon | 2023-04-15 | 3 | -116/+157 |
* | Update | Volpeon | 2023-04-13 | 2 | -8/+2 |
* | Experimental convnext discriminator support | Volpeon | 2023-04-11 | 1 | -0/+35 |
* | Update | Volpeon | 2023-04-09 | 1 | -1/+4 |
* | Update | Volpeon | 2023-04-09 | 2 | -6/+14 |
* | Update | Volpeon | 2023-04-08 | 2 | -2/+2 |
* | TI: Bring back old embedding decay | Volpeon | 2023-04-04 | 2 | -14/+15 |
* | Improved sparse embeddings | Volpeon | 2023-04-03 | 2 | -31/+78 |
* | TI: Delta learning | Volpeon | 2023-04-03 | 1 | -17/+33 |
* | Revert | Volpeon | 2023-04-01 | 1 | -38/+4 |
* | Combined TI with embedding and LoRA | Volpeon | 2023-04-01 | 1 | -4/+15 |
* | Experimental: TI via LoRA | Volpeon | 2023-04-01 | 1 | -15/+38 |
* | Add support for Adafactor, add TI initializer noise | Volpeon | 2023-04-01 | 1 | -1/+9 |
* | Update | Volpeon | 2023-03-31 | 1 | -24/+9 |
* | Fix TI | Volpeon | 2023-03-27 | 1 | -9/+25 |
* | Fix TI | Volpeon | 2023-03-27 | 1 | -25/+9 |
* | Revert to regular embeddings | Volpeon | 2023-03-27 | 1 | -19/+15 |
* | Sparse TI embeddings without sparse tensors | Volpeon | 2023-03-27 | 1 | -17/+23 |
* | Fix TI embeddings init | Volpeon | 2023-03-26 | 1 | -1/+1 |
* | Improved TI embeddings | Volpeon | 2023-03-26 | 1 | -7/+23 |
* | Update | Volpeon | 2023-03-23 | 1 | -3/+3 |
* | Update | Volpeon | 2023-03-01 | 1 | -1/+1 |
* | Embedding normalization: Ignore tensors with grad = 0 | Volpeon | 2023-02-21 | 1 | -8/+0 |
* | Optimized embedding normalization | Volpeon | 2023-01-17 | 1 | -5/+2 |
* | Update | Volpeon | 2023-01-17 | 1 | -0/+3 |
* | Cleanup | Volpeon | 2023-01-14 | 1 | -1/+4 |
* | More modularization | Volpeon | 2023-01-13 | 1 | -5/+1 |
* | Removed PromptProcessor, modularized training loop | Volpeon | 2023-01-13 | 3 | -39/+39 |
* | Code deduplication | Volpeon | 2023-01-13 | 1 | -2/+2 |
* | Fixed TI decay | Volpeon | 2023-01-12 | 1 | -1/+9 |
* | Update | Volpeon | 2023-01-07 | 1 | -5/+0 |
* | Update | Volpeon | 2023-01-06 | 1 | -1/+1 |
* | Added EMA to TI | Volpeon | 2023-01-05 | 1 | -3/+3 |
* | Update | Volpeon | 2023-01-05 | 1 | -4/+6 |
* | Various cleanups | Volpeon | 2023-01-05 | 2 | -7/+7 |
* | Update | Volpeon | 2023-01-04 | 2 | -2/+4 |
* | Added vector dropout | Volpeon | 2023-01-03 | 1 | -2/+25 |