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

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

Neuron 32 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.1028. Min Act: -0.0003

Data Index: 459060 (C4 (Web Text))

Max Activating Token Index: 135

Text #1

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0941. Min Act: -0.0004

Data Index: 1040443 (C4 (Web Text))

Max Activating Token Index: 190

Text #2

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0995. Min Act: -0.0003

Data Index: 929359 (C4 (Web Text))

Max Activating Token Index: 684

Text #3

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0920. Min Act: -0.0004

Data Index: 1304889 (C4 (Web Text))

Max Activating Token Index: 678

Text #4

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0926. Min Act: -0.0003

Data Index: 1229581 (C4 (Web Text))

Max Activating Token Index: 158

Text #5

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0802. Min Act: -0.0003

Data Index: 745143 (C4 (Web Text))

Max Activating Token Index: 697

Text #6

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0891. Min Act: -0.0003

Data Index: 284568 (C4 (Web Text))

Max Activating Token Index: 272

Text #7

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0813. Min Act: -0.0004

Data Index: 444412 (C4 (Web Text))

Max Activating Token Index: 212

Text #8

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0818. Min Act: -0.0003

Data Index: 1634687 (Python Code)

Max Activating Token Index: 185

Text #9

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0839. Min Act: -0.0003

Data Index: 213291 (C4 (Web Text))

Max Activating Token Index: 799

Text #10

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0792. Min Act: -0.0003

Data Index: 1541521 (Python Code)

Max Activating Token Index: 704

Text #11

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0802. Min Act: -0.0003

Data Index: 900615 (C4 (Web Text))

Max Activating Token Index: 246

Text #12

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0781. Min Act: -0.0003

Data Index: 1524769 (Python Code)

Max Activating Token Index: 545

Text #13

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0789. Min Act: -0.0004

Data Index: 207711 (C4 (Web Text))

Max Activating Token Index: 140

Text #14

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0812. Min Act: -0.0003

Data Index: 1129562 (C4 (Web Text))

Max Activating Token Index: 266

Text #15

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0769. Min Act: -0.0004

Data Index: 124206 (C4 (Web Text))

Max Activating Token Index: 482

Text #16

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0767. Min Act: -0.0003

Data Index: 361654 (C4 (Web Text))

Max Activating Token Index: 919

Text #17

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0758. Min Act: -0.0003

Data Index: 1141389 (C4 (Web Text))

Max Activating Token Index: 266

Text #18

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0776. Min Act: -0.0003

Data Index: 1516094 (Python Code)

Max Activating Token Index: 943

Text #19

Max Range: 0.1028. Min Range: -0.1028

Max Act: 0.0688. Min Act: -0.0003

Data Index: 516047 (C4 (Web Text))

Max Activating Token Index: 567