summaryrefslogtreecommitdiffstats
path: root/training
Commit message (Expand)AuthorAgeFilesLines
...
* More training adjustmentsVolpeon2023-01-163-8/+9
* Handle empty validation datasetVolpeon2023-01-163-47/+58
* Extended Dreambooth: Train TI tokens separatelyVolpeon2023-01-161-0/+1
* Implemented extended Dreambooth trainingVolpeon2023-01-161-3/+4
* Added Dreambooth strategyVolpeon2023-01-151-0/+183
* Restored functional trainerVolpeon2023-01-152-27/+83
* UpdateVolpeon2023-01-153-119/+64
* Removed unused code, put training callbacks in dataclassVolpeon2023-01-152-44/+39
* Added functional TI strategyVolpeon2023-01-152-0/+282
* Added functional trainerVolpeon2023-01-151-1/+74
* UpdateVolpeon2023-01-142-122/+24
* UpdateVolpeon2023-01-141-10/+9
* WIP: Modularization ("free(): invalid pointer" my ass)Volpeon2023-01-143-220/+130
* TI: Prepare UNet with Accelerate as wellVolpeon2023-01-142-27/+26
* FixVolpeon2023-01-141-1/+1
* CleanupVolpeon2023-01-142-60/+63
* Reverted modularization mostlyVolpeon2023-01-135-531/+70
* More modularizationVolpeon2023-01-136-32/+541
* Simplified step calculationsVolpeon2023-01-131-20/+22
* Removed PromptProcessor, modularized training loopVolpeon2023-01-132-10/+208
* Code deduplicationVolpeon2023-01-131-0/+55
* UpdateVolpeon2023-01-121-4/+7
* FixVolpeon2023-01-111-2/+2
* TI: Use grad clipping from LoRA #104Volpeon2023-01-111-1/+1
* Added arg to disable tag shufflingVolpeon2023-01-101-10/+10
* Fixed aspect ratio bucketing; allow passing token IDs to pipelineVolpeon2023-01-081-6/+8
* Improved aspect ratio bucketingVolpeon2023-01-081-1/+1
* CleanupVolpeon2023-01-071-0/+54
* Made aspect ratio bucketing configurableVolpeon2023-01-071-7/+2
* Added progressive aspect ratio bucketingVolpeon2023-01-071-2/+2
* UpdateVolpeon2023-01-061-6/+6
* Use context manager for EMA, on_train/eval hooksVolpeon2023-01-062-27/+26
* Add contextmanager to EMAModel to apply weights temporarilyVolpeon2023-01-061-0/+12
* Added EMA to TIVolpeon2023-01-051-5/+95
* Fix LR finderVolpeon2023-01-051-7/+23
* UpdateVolpeon2023-01-051-1/+4
* Various cleanupsVolpeon2023-01-051-0/+75
* UpdateVolpeon2023-01-041-1/+6
* Various updatesVolpeon2023-01-041-4/+0
* Better eval generatorVolpeon2023-01-041-3/+3
* Fixed reproducibility, more consistant validationVolpeon2023-01-042-30/+68
* Added vector dropoutVolpeon2023-01-031-3/+6
* Fixed LR finderVolpeon2023-01-021-17/+21
* UpdateVolpeon2023-01-021-2/+31
* FixVolpeon2023-01-021-1/+1
* Improved one cycle schedulerVolpeon2023-01-021-30/+62
* UpdateVolpeon2023-01-011-3/+3
* UpdatesVolpeon2023-01-012-1/+7
* Fixed accuracy calc, other improvementsVolpeon2023-01-011-48/+0
* Added multi-vector embeddingsVolpeon2022-12-311-4/+56