Model: SoLU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 369 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-3l')

Text #0

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0110. Min Act: -0.0006

Data Index: 1366015 (Python Code)

Max Activating Token Index: 959

Text #1

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0100. Min Act: -0.0006

Data Index: 901013 (C4 (Web Text))

Max Activating Token Index: 903

Text #2

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0099. Min Act: -0.0005

Data Index: 846729 (C4 (Web Text))

Max Activating Token Index: 544

Text #3

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0099. Min Act: -0.0006

Data Index: 1485155 (Python Code)

Max Activating Token Index: 535

Text #4

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0102. Min Act: -0.0005

Data Index: 1011002 (C4 (Web Text))

Max Activating Token Index: 101

Text #5

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0097. Min Act: -0.0005

Data Index: 632804 (C4 (Web Text))

Max Activating Token Index: 55

Text #6

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0098. Min Act: -0.0005

Data Index: 599724 (C4 (Web Text))

Max Activating Token Index: 553

Text #7

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0098. Min Act: -0.0006

Data Index: 1309755 (C4 (Web Text))

Max Activating Token Index: 597

Text #8

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0097. Min Act: -0.0006

Data Index: 48456 (C4 (Web Text))

Max Activating Token Index: 85

Text #9

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0099. Min Act: -0.0006

Data Index: 1385077 (Python Code)

Max Activating Token Index: 735

Text #10

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0096. Min Act: -0.0005

Data Index: 401918 (C4 (Web Text))

Max Activating Token Index: 321

Text #11

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0096. Min Act: -0.0005

Data Index: 6577 (C4 (Web Text))

Max Activating Token Index: 916

Text #12

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0092. Min Act: -0.0006

Data Index: 1368655 (Python Code)

Max Activating Token Index: 300

Text #13

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0097. Min Act: -0.0005

Data Index: 201974 (C4 (Web Text))

Max Activating Token Index: 259

Text #14

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0099. Min Act: -0.0005

Data Index: 1317319 (C4 (Web Text))

Max Activating Token Index: 157

Text #15

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0096. Min Act: -0.0005

Data Index: 91069 (C4 (Web Text))

Max Activating Token Index: 130

Text #16

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0094. Min Act: -0.0006

Data Index: 1214944 (C4 (Web Text))

Max Activating Token Index: 752

Text #17

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0091. Min Act: -0.0005

Data Index: 640230 (C4 (Web Text))

Max Activating Token Index: 719

Text #18

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0093. Min Act: -0.0005

Data Index: 287567 (C4 (Web Text))

Max Activating Token Index: 331

Text #19

Max Range: 0.0110. Min Range: -0.0110

Max Act: 0.0088. Min Act: -0.0006

Data Index: 55974 (C4 (Web Text))

Max Activating Token Index: 767