summaryrefslogtreecommitdiffstats
path: root/textual_inversion.py
Commit message (Expand)AuthorAgeFilesLines
* Improved dataset prompt handling, fixedVolpeon2022-12-191-1034/+0
* Avoid increased VRAM usage on validationVolpeon2022-12-151-3/+8
* More generic datset filterVolpeon2022-12-141-3/+10
* Unified loading of TI embeddingsVolpeon2022-12-131-12/+11
* Optimized Textual Inversion training by filtering dataset by existence of add...Volpeon2022-12-131-3/+10
* Fixed sample/checkpoint frequencyVolpeon2022-12-131-11/+9
* Add support for resume in Textual InversionVolpeon2022-12-131-65/+54
* Dreambooth: Support loading Textual Inversion embeddingsVolpeon2022-12-121-14/+23
* Support attention_mask of text encoderVolpeon2022-12-101-4/+5
* UpdateVolpeon2022-12-041-1/+3
* UpdateVolpeon2022-12-011-3/+3
* UpdateVolpeon2022-12-011-4/+5
* UpdateVolpeon2022-12-011-0/+1
* Adjusted training to upstreamVolpeon2022-11-281-23/+26
* Fixed noise calculation for v-predictionVolpeon2022-11-281-9/+20
* Fixed and improved Textual Inversion trainingVolpeon2022-11-271-44/+68
* FixVolpeon2022-11-271-1/+0
* Generic loading of scheduler (training)Volpeon2022-11-271-14/+9
* RefactoringVolpeon2022-11-141-4/+4
* UpdateVolpeon2022-11-141-4/+9
* UpdateVolpeon2022-11-031-3/+11
* Added CLI arg to set dataloader worker num; improved text encoder handling wi...Volpeon2022-10-271-0/+10
* New Euler_a schedulerVolpeon2022-10-261-3/+3
* Advanced datasetsVolpeon2022-10-241-24/+34
* Training updateVolpeon2022-10-221-1/+0
* Add optional TI functionality to DreamboothVolpeon2022-10-211-7/+8
* Adapted other scripts for new prompt processingVolpeon2022-10-181-7/+10
* Upstream updates; better handling of textual embeddingVolpeon2022-10-171-17/+24
* UpdateVolpeon2022-10-161-2/+4
* UpdateVolpeon2022-10-161-11/+17
* UpdateVolpeon2022-10-151-12/+21
* Removed aesthetic gradients; training improvementsVolpeon2022-10-141-18/+14
* Added support for Aesthetic GradientsVolpeon2022-10-131-10/+0
* Added TI+Dreambooth trainingVolpeon2022-10-131-3/+8
* Various updatesVolpeon2022-10-121-60/+14
* Added EMA support to Textual InversionVolpeon2022-10-101-6/+48
* TI: Use different lr schedulerVolpeon2022-10-101-1/+1
* Updated default paramsVolpeon2022-10-101-3/+3
* UpdateVolpeon2022-10-091-6/+4
* UpdateVolpeon2022-10-081-29/+28
* Fix small detailsVolpeon2022-10-071-35/+35
* Training: Create multiple class images per training imageVolpeon2022-10-071-13/+35
* UpdateVolpeon2022-10-061-11/+101
* Add Textual Inversion with class dataset (a la Dreambooth)Volpeon2022-10-041-7/+6
* Default sample steps 30 -> 40Volpeon2022-10-041-1/+1
* Dreambooth: Generate specialized class images from input promptsVolpeon2022-10-031-4/+2
* Small perf improvementsVolpeon2022-10-031-2/+4
* Added negative prompt support for training scriptsVolpeon2022-10-031-60/+23
* Fixed euler_a generator argumentVolpeon2022-10-031-8/+0
* Assign unused images in validation dataset to train datasetVolpeon2022-10-031-3/+9
generated by cgit v1.2.3-54-g00ecf (git 2.44.0) at 2024-07-01 11:04:27 +0000