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

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

Neuron 1400 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: 0.4451. Min Range: -0.4451

Max Act: 0.4451. Min Act: -0.0001

Data Index: 121861 (C4 (Web Text))

Max Activating Token Index: 652

Text #1

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.3275. Min Act: -0.0001

Data Index: 1371256 (Python Code)

Max Activating Token Index: 2

Text #2

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.3295. Min Act: -0.0001

Data Index: 78879 (C4 (Web Text))

Max Activating Token Index: 322

Text #3

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.3168. Min Act: -0.0001

Data Index: 133615 (C4 (Web Text))

Max Activating Token Index: 270

Text #4

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.3055. Min Act: -0.0001

Data Index: 349009 (C4 (Web Text))

Max Activating Token Index: 12

Text #5

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2560. Min Act: -0.0001

Data Index: 603217 (C4 (Web Text))

Max Activating Token Index: 660

Text #6

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2994. Min Act: -0.0001

Data Index: 903226 (C4 (Web Text))

Max Activating Token Index: 8

Text #7

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2942. Min Act: -0.0001

Data Index: 950068 (C4 (Web Text))

Max Activating Token Index: 818

Text #8

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.3013. Min Act: -0.0001

Data Index: 1260021 (C4 (Web Text))

Max Activating Token Index: 101

Text #9

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2902. Min Act: -0.0001

Data Index: 1477009 (Python Code)

Max Activating Token Index: 11

Text #10

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2893. Min Act: -0.0001

Data Index: 790354 (C4 (Web Text))

Max Activating Token Index: 6

Text #11

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2851. Min Act: -0.0001

Data Index: 106133 (C4 (Web Text))

Max Activating Token Index: 22

Text #12

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2852. Min Act: -0.0001

Data Index: 1063262 (C4 (Web Text))

Max Activating Token Index: 577

Text #13

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2774. Min Act: -0.0001

Data Index: 1281821 (C4 (Web Text))

Max Activating Token Index: 344

Text #14

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2778. Min Act: -0.0001

Data Index: 1523367 (Python Code)

Max Activating Token Index: 17

Text #15

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2551. Min Act: -0.0001

Data Index: 935626 (C4 (Web Text))

Max Activating Token Index: 781

Text #16

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2591. Min Act: -0.0001

Data Index: 980359 (C4 (Web Text))

Max Activating Token Index: 159

Text #17

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2587. Min Act: -0.0001

Data Index: 341955 (C4 (Web Text))

Max Activating Token Index: 232

Text #18

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2583. Min Act: -0.0001

Data Index: 1085760 (C4 (Web Text))

Max Activating Token Index: 5

Text #19

Max Range: 0.4451. Min Range: -0.4451

Max Act: 0.2743. Min Act: -0.0001

Data Index: 818213 (C4 (Web Text))

Max Activating Token Index: 12