summaryrefslogtreecommitdiffstats
path: root/training/util.py
Commit message (Expand)AuthorAgeFilesLines
* Fixed loss/acc loggingVolpeon2023-04-281-1/+6
* UpdateVolpeon2023-04-271-8/+14
* Implemented different noise offsetVolpeon2023-03-031-1/+0
* Restored LR finderVolpeon2023-01-201-143/+3
* UpdateVolpeon2023-01-171-8/+3
* Restored functional trainerVolpeon2023-01-151-3/+5
* UpdateVolpeon2023-01-141-112/+0
* WIP: Modularization ("free(): invalid pointer" my ass)Volpeon2023-01-141-96/+118
* TI: Prepare UNet with Accelerate as wellVolpeon2023-01-141-8/+8
* CleanupVolpeon2023-01-141-15/+11
* More modularizationVolpeon2023-01-131-6/+9
* Removed PromptProcessor, modularized training loopVolpeon2023-01-131-1/+12
* Fixed aspect ratio bucketing; allow passing token IDs to pipelineVolpeon2023-01-081-6/+8
* Improved aspect ratio bucketingVolpeon2023-01-081-1/+1
* 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 hooksVolpeon2023-01-061-1/+1
* Add contextmanager to EMAModel to apply weights temporarilyVolpeon2023-01-061-0/+12
* Added EMA to TIVolpeon2023-01-051-5/+95
* Added multi-vector embeddingsVolpeon2022-12-311-4/+56
* Integrated updates from diffusersVolpeon2022-12-281-5/+0
* Code simplifications, avoid autocastVolpeon2022-12-251-57/+49
* FixVolpeon2022-12-241-1/+1
* Fix againVolpeon2022-12-241-1/+1
* FixVolpeon2022-12-241-1/+4
* Improved class prompt handlingVolpeon2022-12-231-3/+1
* Moved common training code into separate moduleVolpeon2022-12-211-0/+131