summaryrefslogtreecommitdiffstats
Commit message (Expand)AuthorAgeFilesLines
* UpdateVolpeon2023-06-222-17/+9
* Remove training guidance_scaleVolpeon2023-06-225-42/+1
* Remove convnextVolpeon2023-06-227-482/+0
* UpdateVolpeon2023-06-215-36/+11
* UpdateVolpeon2023-06-2124-989/+1873
* Fix LoRA training with DAdanVolpeon2023-05-165-43/+42
* Avoid model recompilation due to varying prompt lengthsVolpeon2023-05-165-65/+43
* Patch xformers to cast dtypesVolpeon2023-05-162-7/+17
* Update depsVolpeon2023-05-161-3/+3
* LoRA: Apply to out layers as wellVolpeon2023-05-161-2/+4
* Support LoRA training for token embeddingsVolpeon2023-05-165-25/+36
* Fix for latest PEFTVolpeon2023-05-123-156/+50
* UpdateVolpeon2023-05-114-13/+45
* UpdateVolpeon2023-05-065-13/+205
* UpdateVolpeon2023-05-055-32/+71
* torch.compile won't work yet, keep code preparedVolpeon2023-04-294-4/+52
* Optional xformersVolpeon2023-04-291-3/+10
* Fixed model compilationVolpeon2023-04-292-3/+9
* Support torch.compileVolpeon2023-04-283-9/+27
* Fixed loss/acc loggingVolpeon2023-04-282-17/+22
* FixVolpeon2023-04-271-8/+19
* UpdateVolpeon2023-04-277-41/+119
* Fix cycle loopVolpeon2023-04-232-6/+10
* UpdateVolpeon2023-04-214-8/+12
* Fix PTIVolpeon2023-04-201-2/+4
* UpdateVolpeon2023-04-203-34/+126
* FixVolpeon2023-04-173-7/+6
* FixVolpeon2023-04-161-1/+1
* Improved automation capsVolpeon2023-04-166-54/+100
* Added option to use constant LR on cycles > 1Volpeon2023-04-162-2/+20
* FixVolpeon2023-04-161-1/+1
* FixVolpeon2023-04-166-90/+177
* FixVolpeon2023-04-161-16/+18
* FixVolpeon2023-04-162-23/+21
* UpdateVolpeon2023-04-166-32/+69
* FixVolpeon2023-04-151-1/+1
* FixVolpeon2023-04-153-24/+5
* TI via LoRAVolpeon2023-04-159-140/+212
* Added cycle LR decayVolpeon2023-04-133-9/+16
* UpdateVolpeon2023-04-138-67/+80
* Experimental convnext discriminator supportVolpeon2023-04-114-25/+109
* Store sample images in Tensorboard as wellVolpeon2023-04-111-13/+12
* UpdateVolpeon2023-04-114-27/+43
* Randomize dataset across cyclesVolpeon2023-04-103-12/+12
* UpdateVolpeon2023-04-104-27/+30
* Fix sample gen: models sometimes weren't in eval modeVolpeon2023-04-104-28/+29
* UpdateVolpeon2023-04-106-55/+89
* UpdateVolpeon2023-04-094-12/+33
* FixVolpeon2023-04-091-1/+1
* UpdateVolpeon2023-04-091-6/+1