Janouk Kelderman: v1.0: janoukKelderman.safetensors

PICS: 80

EPOCHS: 12

REG PICS: none

BLIP: 5 - 75

steps: 8% 137/1644 [03:14<35:37, 1.42s/it, loss=0.129]

steps: 17% 274/1644 [06:36<33:03, 1.45s/it, loss=0.117]

steps: 25% 411/1644 [09:58<29:55, 1.46s/it, loss=0.116]

steps: 33% 548/1644 [13:20<26:40, 1.46s/it, loss=0.12]

steps: 42% 685/1644 [16:42<23:22, 1.46s/it, loss=0.121]

steps: 58% 959/1644 [23:26<16:44, 1.47s/it, loss=0.122]

steps: 67% 1096/1644 [26:48<13:24, 1.47s/it, loss=0.118]

steps: 75% 1233/1644 [30:09<10:03, 1.47s/it, loss=0.115]

steps: 83% 1370/1644 [33:31<06:42, 1.47s/it, loss=0.123]

steps: 92% 1507/1644 [36:53<03:21, 1.47s/it, loss=0.13]

steps: 100% 1644/1644 [40:25<00:00, 1.48s/it, loss=0.111]

I think v2 is quite an improvement. Some noise_offset to improve color, higher network_dim to improve likeness, and more high quality images for robustness.

I always try to put my settings in the "About this version" section, so check that out.