Upload a good LoRA this time!

Trained on immiscible noise.

network_args = ["rank_dropout=0.1", "module_dropout=0.1", "context_mod_dim=0", "context_mlp_dim=0", "x_mod_dim=0", "x_mlp_dim=0", "train_block_indices=4-9"]
I think what made this lora good was: low gradient clipping (0.1), rms learning rate adjustment and clipping (FARMSCropv2 optimizer), and only training attn blocks 4-9.

mspaint-style-sd3-FinallyGoodEdition.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e4030519405a65942294ef7851fef72410551d4aad37a8912f6e5a91cacb711
+size 10636120