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

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

Neuron 5005 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.3384. Min Act: -0.0001

Data Index: 72460 (C4 (Web Text))

Max Activating Token Index: 998

Text #1

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.3317. Min Act: -0.0001

Data Index: 252256 (C4 (Web Text))

Max Activating Token Index: 708

Text #2

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.3113. Min Act: -0.0001

Data Index: 955697 (C4 (Web Text))

Max Activating Token Index: 25

Text #3

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.3208. Min Act: -0.0001

Data Index: 1205893 (C4 (Web Text))

Max Activating Token Index: 37

Text #4

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.3035. Min Act: -0.0001

Data Index: 199822 (C4 (Web Text))

Max Activating Token Index: 13

Text #5

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2967. Min Act: -0.0001

Data Index: 1075667 (C4 (Web Text))

Max Activating Token Index: 37

Text #6

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.3035. Min Act: -0.0001

Data Index: 844747 (C4 (Web Text))

Max Activating Token Index: 40

Text #7

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2996. Min Act: -0.0001

Data Index: 993491 (C4 (Web Text))

Max Activating Token Index: 680

Text #8

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2706. Min Act: -0.0001

Data Index: 179507 (C4 (Web Text))

Max Activating Token Index: 40

Text #9

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2714. Min Act: -0.0001

Data Index: 131096 (C4 (Web Text))

Max Activating Token Index: 249

Text #10

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2623. Min Act: -0.0001

Data Index: 1093789 (C4 (Web Text))

Max Activating Token Index: 37

Text #11

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2781. Min Act: -0.0001

Data Index: 1643021 (Python Code)

Max Activating Token Index: 5

Text #12

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2888. Min Act: -0.0001

Data Index: 829295 (C4 (Web Text))

Max Activating Token Index: 522

Text #13

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2688. Min Act: -0.0001

Data Index: 989527 (C4 (Web Text))

Max Activating Token Index: 49

Text #14

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2665. Min Act: -0.0001

Data Index: 533228 (C4 (Web Text))

Max Activating Token Index: 20

Text #15

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2727. Min Act: -0.0001

Data Index: 111351 (C4 (Web Text))

Max Activating Token Index: 34

Text #16

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2741. Min Act: -0.0001

Data Index: 283307 (C4 (Web Text))

Max Activating Token Index: 980

Text #17

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2708. Min Act: -0.0001

Data Index: 260807 (C4 (Web Text))

Max Activating Token Index: 201

Text #18

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2782. Min Act: -0.0001

Data Index: 1621639 (Python Code)

Max Activating Token Index: 255

Text #19

Max Range: 0.3384. Min Range: -0.3384

Max Act: 0.2775. Min Act: -0.0001

Data Index: 276444 (C4 (Web Text))

Max Activating Token Index: 846