Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

Dataset: 80% C4 (Web Text) and 20% Python Code

Neuron 4500 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.9000. Min Act: -0.0001

Data Index: 189953 (C4 (Web Text))

Max Activating Token Index: 311

Text #1

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.7693. Min Act: -0.0001

Data Index: 404398 (C4 (Web Text))

Max Activating Token Index: 536

Text #2

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.7343. Min Act: -0.0001

Data Index: 804887 (C4 (Web Text))

Max Activating Token Index: 466

Text #3

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5789. Min Act: -0.0001

Data Index: 385849 (C4 (Web Text))

Max Activating Token Index: 176

Text #4

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5881. Min Act: -0.0001

Data Index: 642385 (C4 (Web Text))

Max Activating Token Index: 162

Text #5

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.6716. Min Act: -0.0001

Data Index: 356478 (C4 (Web Text))

Max Activating Token Index: 888

Text #6

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.6271. Min Act: -0.0001

Data Index: 976219 (C4 (Web Text))

Max Activating Token Index: 797

Text #7

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.6647. Min Act: -0.0001

Data Index: 998410 (C4 (Web Text))

Max Activating Token Index: 133

Text #8

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.6107. Min Act: -0.0001

Data Index: 318040 (C4 (Web Text))

Max Activating Token Index: 866

Text #9

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5837. Min Act: -0.0001

Data Index: 801820 (C4 (Web Text))

Max Activating Token Index: 332

Text #10

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5160. Min Act: -0.0001

Data Index: 890958 (C4 (Web Text))

Max Activating Token Index: 678

Text #11

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.6050. Min Act: -0.0001

Data Index: 1190629 (C4 (Web Text))

Max Activating Token Index: 900

Text #12

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5557. Min Act: -0.0001

Data Index: 899568 (C4 (Web Text))

Max Activating Token Index: 13

Text #13

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5321. Min Act: -0.0001

Data Index: 1089950 (C4 (Web Text))

Max Activating Token Index: 431

Text #14

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.6139. Min Act: -0.0001

Data Index: 583657 (C4 (Web Text))

Max Activating Token Index: 118

Text #15

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5360. Min Act: -0.0001

Data Index: 648669 (C4 (Web Text))

Max Activating Token Index: 49

Text #16

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5703. Min Act: -0.0001

Data Index: 58041 (C4 (Web Text))

Max Activating Token Index: 131

Text #17

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5315. Min Act: -0.0001

Data Index: 1515483 (Python Code)

Max Activating Token Index: 109

Text #18

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5814. Min Act: -0.0001

Data Index: 1295302 (C4 (Web Text))

Max Activating Token Index: 757

Text #19

Max Range: 0.9000. Min Range: -0.9000

Max Act: 0.5086. Min Act: -0.0001

Data Index: 475544 (C4 (Web Text))

Max Activating Token Index: 907