Merge pull request #11 from qslug/patch-1

Update SD 1.5 training example to use --ed1_mode
This commit is contained in:
Victor Hall 2023-01-12 13:50:43 -08:00 committed by GitHub
commit 10c9ba4b20
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 2 additions and 1 deletions

View File

@ -41,7 +41,8 @@ Resuming from a checkpoint, 50 epochs, 6 batch size, 3e-6 learning rate, cosine
--sample_steps 200 ^ --sample_steps 200 ^
--lr 3e-6 ^ --lr 3e-6 ^
--ckpt_every_n_minutes 10 ^ --ckpt_every_n_minutes 10 ^
--useadam8bit --useadam8bit ^
--ed1_mode
Training from SD2 512 base model, 18 epochs, 4 batch size, 1.2e-6 learning rate, constant LR, generate samples evern 100 steps, 30 minute checkpoint interval, adam8bit, using imagesin the x:\mydata folder, training at resolution class of 640: Training from SD2 512 base model, 18 epochs, 4 batch size, 1.2e-6 learning rate, constant LR, generate samples evern 100 steps, 30 minute checkpoint interval, adam8bit, using imagesin the x:\mydata folder, training at resolution class of 640: