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

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

Neuron 3189 in Layer 8

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.8926. Min Act: -0.0000

Data Index: 766014 (C4 (Web Text))

Max Activating Token Index: 219

Text #1

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.9095. Min Act: -0.0000

Data Index: 366577 (C4 (Web Text))

Max Activating Token Index: 794

Text #2

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.7971. Min Act: -0.0000

Data Index: 1200263 (C4 (Web Text))

Max Activating Token Index: 224

Text #3

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.7679. Min Act: -0.0000

Data Index: 1461774 (Python Code)

Max Activating Token Index: 817

Text #4

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.7276. Min Act: -0.0000

Data Index: 1107267 (C4 (Web Text))

Max Activating Token Index: 98

Text #5

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.7030. Min Act: -0.0000

Data Index: 319511 (C4 (Web Text))

Max Activating Token Index: 157

Text #6

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.7239. Min Act: -0.0000

Data Index: 230699 (C4 (Web Text))

Max Activating Token Index: 775

Text #7

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.7537. Min Act: -0.0000

Data Index: 73541 (C4 (Web Text))

Max Activating Token Index: 54

Text #8

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.6542. Min Act: -0.0000

Data Index: 1455041 (Python Code)

Max Activating Token Index: 304

Text #9

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.6468. Min Act: -0.0000

Data Index: 449816 (C4 (Web Text))

Max Activating Token Index: 57

Text #10

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.6354. Min Act: -0.0000

Data Index: 848420 (C4 (Web Text))

Max Activating Token Index: 911

Text #11

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.6227. Min Act: -0.0000

Data Index: 1643318 (Python Code)

Max Activating Token Index: 104

Text #12

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5746. Min Act: -0.0000

Data Index: 904922 (C4 (Web Text))

Max Activating Token Index: 460

Text #13

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.6331. Min Act: -0.0000

Data Index: 136861 (C4 (Web Text))

Max Activating Token Index: 688

Text #14

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5925. Min Act: -0.0000

Data Index: 1246980 (C4 (Web Text))

Max Activating Token Index: 441

Text #15

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5906. Min Act: -0.0000

Data Index: 1475904 (Python Code)

Max Activating Token Index: 805

Text #16

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5253. Min Act: -0.0000

Data Index: 1428176 (Python Code)

Max Activating Token Index: 323

Text #17

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5996. Min Act: -0.0000

Data Index: 789205 (C4 (Web Text))

Max Activating Token Index: 124

Text #18

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5564. Min Act: -0.0000

Data Index: 649055 (C4 (Web Text))

Max Activating Token Index: 552

Text #19

Max Range: 0.9095. Min Range: -0.9095

Max Act: 0.5662. Min Act: -0.0000

Data Index: 947930 (C4 (Web Text))

Max Activating Token Index: 197