Learning Rate: 0.05:10, 0.001:35, 0.005:1000, 0.001:2500, 0.0005:5000, 0.0001
Gradient Clip: norm, Clip: 0.1
Batch size: 1, Gradient Accumulation steps 4
14 training images
3 vectors per token

Average vector strength: 0.1579
Average vector magnitude: 2.7207