v5.0a<pre><code>
 "unetLR": 0.0005,
 "clipSkip": 2,
 "networkDim": 128,
 "numRepeats": 10,
 "resolution": 1024,
 "lrScheduler": "cosine_with_restarts",
 "noiseOffset": 0.1,
 "networkAlpha": 64,
 "optimizerType": "AdamW8Bit",
 "textEncoderLR": 0.00005,

bigger dataset </code></pre>v3.0b:Text Encoder learning rate:0.000005Unet learning rate:0.0001LR Scheduler:cosineOptimizer:AdamW8bitNetwork Dim:64Network Alpha:32v3.0a:Bigger training dataset0.8 have best result1 look weird for meText Encoder learning rate:0.00001Unet learning rate:0.0001LR Scheduler:constantOptimizer:AdamW8bitNetwork Dim:32Network Alpha:16

Brown Dust 2 style

bds3.0b.safetensors

v3.0b

v5.0a

Files with SHA256: 5bf49e0a0de9307f64fbfe4a30de8b422111fdabb325fc30987b7a188b36f01d