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

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

Neuron 3615 in Layer 5

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.9510. Min Act: -0.0001

Data Index: 48624 (C4 (Web Text))

Max Activating Token Index: 941

Text #1

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.8619. Min Act: -0.0001

Data Index: 881461 (C4 (Web Text))

Max Activating Token Index: 981

Text #2

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.8734. Min Act: -0.0001

Data Index: 430320 (C4 (Web Text))

Max Activating Token Index: 40

Text #3

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.8384. Min Act: -0.0001

Data Index: 590748 (C4 (Web Text))

Max Activating Token Index: 801

Text #4

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.7583. Min Act: -0.0000

Data Index: 330996 (C4 (Web Text))

Max Activating Token Index: 896

Text #5

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6918. Min Act: -0.0000

Data Index: 1143761 (C4 (Web Text))

Max Activating Token Index: 826

Text #6

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.7193. Min Act: -0.0000

Data Index: 43226 (C4 (Web Text))

Max Activating Token Index: 601

Text #7

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.7003. Min Act: -0.0000

Data Index: 254103 (C4 (Web Text))

Max Activating Token Index: 965

Text #8

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6283. Min Act: -0.0000

Data Index: 82108 (C4 (Web Text))

Max Activating Token Index: 245

Text #9

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6504. Min Act: -0.0000

Data Index: 1221003 (C4 (Web Text))

Max Activating Token Index: 1011

Text #10

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6206. Min Act: -0.0000

Data Index: 1458477 (Python Code)

Max Activating Token Index: 55

Text #11

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6176. Min Act: -0.0000

Data Index: 1476341 (Python Code)

Max Activating Token Index: 322

Text #12

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6153. Min Act: -0.0000

Data Index: 1535744 (Python Code)

Max Activating Token Index: 114

Text #13

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6163. Min Act: -0.0000

Data Index: 1366668 (Python Code)

Max Activating Token Index: 505

Text #14

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6149. Min Act: -0.0000

Data Index: 1194572 (C4 (Web Text))

Max Activating Token Index: 430

Text #15

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6355. Min Act: -0.0000

Data Index: 1525729 (Python Code)

Max Activating Token Index: 127

Text #16

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6079. Min Act: -0.0000

Data Index: 1509981 (Python Code)

Max Activating Token Index: 79

Text #17

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6005. Min Act: -0.0000

Data Index: 930712 (C4 (Web Text))

Max Activating Token Index: 177

Text #18

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.6006. Min Act: -0.0000

Data Index: 676303 (C4 (Web Text))

Max Activating Token Index: 46

Text #19

Max Range: 0.9510. Min Range: -0.9510

Max Act: 0.5645. Min Act: -0.0000

Data Index: 1283319 (C4 (Web Text))

Max Activating Token Index: 391