An A6000 from Massed Compute can take Rank 7 training, I hope? I think they provide ample storage sp
An A6000 from Massed Compute can take Rank 7 training, I hope? I think they provide ample storage space.
Fsingle A6000 can do batch size 7
Fand it cant do that at lora 
Fmulti gpu kohya arrived to windows as well
Flast experiment almost completed
F
DHaha, well I'm just downloading stuff on the massed compute, readying it up for when you post BS7 fine-tune config
Fyou can already use included bs7 config
Fby the way xformers a little bit better alike
Fso use it on linux
Fon windows i am waiting torch 2.5 update
DOh, is that so? I thought you were going to update it to ver.6 or something 
Fcurrent config didnt change
Fbut i will add info
DI'll begin right away then, thanks!
Fyou are welcome
Tthen it will make me look old?
F maybe. you can add age too to prompt DWhat's the expected speed for BS7 config?
DThe file here says 'Batch_Size_7_48GB_GPU_46250MB_28.9_second_it_Tier_1', does that mean 28.9 s/it 
Fbut 1 it = 7 it of batch size 1
DDo I need to adjust the epochs accordingly?
Fno epoch same
Fit handles it
Fif your nujmber of images not divisible
Fyou will get lesser epoch
Fbut same step count total
DI see.
DUnderstood!
F256 images lora testing
Fbatch size 7
F
Fgood quality
F
Fbut fine tuning way better i think we will see grids
F
F
F
Fupgraded my internet to 1 gbit
Ffaster speed really helps to reinstall apps and verify they are broken or working
DThis just keeps getting better and better 
Fye really amazing results
Fbest checkpoints will clash
Fi am generating grids right now
DI've got a feeling I'm going to run out of space on massed compute lol. Can I move the completed checkpoints to drive and delete them while the training is ongoing?
F
DHow things started
