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

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

Neuron 3791 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1307. Min Act: -0.0000

Data Index: 303058 (C4 (Web Text))

Max Activating Token Index: 806

Text #1

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1183. Min Act: -0.0001

Data Index: 1014473 (C4 (Web Text))

Max Activating Token Index: 727

Text #2

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1062. Min Act: -0.0001

Data Index: 866258 (C4 (Web Text))

Max Activating Token Index: 863

Text #3

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1114. Min Act: -0.0000

Data Index: 396474 (C4 (Web Text))

Max Activating Token Index: 879

Text #4

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1075. Min Act: -0.0001

Data Index: 411940 (C4 (Web Text))

Max Activating Token Index: 553

Text #5

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1038. Min Act: -0.0001

Data Index: 760144 (C4 (Web Text))

Max Activating Token Index: 145

Text #6

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1026. Min Act: -0.0001

Data Index: 944507 (C4 (Web Text))

Max Activating Token Index: 538

Text #7

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1072. Min Act: -0.0001

Data Index: 1044486 (C4 (Web Text))

Max Activating Token Index: 18

Text #8

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1063. Min Act: -0.0001

Data Index: 1024172 (C4 (Web Text))

Max Activating Token Index: 594

Text #9

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1105. Min Act: -0.0001

Data Index: 9180 (C4 (Web Text))

Max Activating Token Index: 838

Text #10

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1036. Min Act: -0.0001

Data Index: 959209 (C4 (Web Text))

Max Activating Token Index: 757

Text #11

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1026. Min Act: -0.0000

Data Index: 380127 (C4 (Web Text))

Max Activating Token Index: 661

Text #12

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.0988. Min Act: -0.0001

Data Index: 323538 (C4 (Web Text))

Max Activating Token Index: 662

Text #13

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1003. Min Act: -0.0001

Data Index: 681103 (C4 (Web Text))

Max Activating Token Index: 329

Text #14

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.1042. Min Act: -0.0001

Data Index: 657113 (C4 (Web Text))

Max Activating Token Index: 955

Text #15

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.0964. Min Act: -0.0001

Data Index: 1420373 (Python Code)

Max Activating Token Index: 584

Text #16

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.0930. Min Act: -0.0001

Data Index: 586398 (C4 (Web Text))

Max Activating Token Index: 290

Text #17

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.0948. Min Act: -0.0001

Data Index: 1084759 (C4 (Web Text))

Max Activating Token Index: 477

Text #18

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.0915. Min Act: -0.0001

Data Index: 77637 (C4 (Web Text))

Max Activating Token Index: 561

Text #19

Max Range: 0.1307. Min Range: -0.1307

Max Act: 0.0905. Min Act: -0.0001

Data Index: 1174522 (C4 (Web Text))

Max Activating Token Index: 390