Model: SoLU Model: 6 Layers, 3072 Neurons per Layer

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

Neuron 643 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-6l')

Text #0

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2962. Min Act: -0.0001

Data Index: 575426 (C4 (Web Text))

Max Activating Token Index: 810

Text #1

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2697. Min Act: -0.0001

Data Index: 118048 (C4 (Web Text))

Max Activating Token Index: 861

Text #2

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2636. Min Act: -0.0001

Data Index: 1064856 (C4 (Web Text))

Max Activating Token Index: 293

Text #3

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2312. Min Act: -0.0001

Data Index: 105127 (C4 (Web Text))

Max Activating Token Index: 816

Text #4

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2329. Min Act: -0.0001

Data Index: 636926 (C4 (Web Text))

Max Activating Token Index: 836

Text #5

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2201. Min Act: -0.0001

Data Index: 235125 (C4 (Web Text))

Max Activating Token Index: 46

Text #6

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2232. Min Act: -0.0001

Data Index: 1273127 (C4 (Web Text))

Max Activating Token Index: 37

Text #7

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2229. Min Act: -0.0001

Data Index: 502964 (C4 (Web Text))

Max Activating Token Index: 74

Text #8

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.1860. Min Act: -0.0001

Data Index: 736587 (C4 (Web Text))

Max Activating Token Index: 303

Text #9

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.1998. Min Act: -0.0001

Data Index: 183284 (C4 (Web Text))

Max Activating Token Index: 838

Text #10

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2220. Min Act: -0.0001

Data Index: 341159 (C4 (Web Text))

Max Activating Token Index: 208

Text #11

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2261. Min Act: -0.0001

Data Index: 566743 (C4 (Web Text))

Max Activating Token Index: 727

Text #12

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2173. Min Act: -0.0001

Data Index: 1183318 (C4 (Web Text))

Max Activating Token Index: 160

Text #13

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2185. Min Act: -0.0001

Data Index: 1623079 (Python Code)

Max Activating Token Index: 909

Text #14

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2082. Min Act: -0.0001

Data Index: 382333 (C4 (Web Text))

Max Activating Token Index: 59

Text #15

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2210. Min Act: -0.0001

Data Index: 1268044 (C4 (Web Text))

Max Activating Token Index: 943

Text #16

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.1840. Min Act: -0.0001

Data Index: 196104 (C4 (Web Text))

Max Activating Token Index: 218

Text #17

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2140. Min Act: -0.0001

Data Index: 1039071 (C4 (Web Text))

Max Activating Token Index: 524

Text #18

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.1898. Min Act: -0.0001

Data Index: 193413 (C4 (Web Text))

Max Activating Token Index: 43

Text #19

Max Range: 0.2962. Min Range: -0.2962

Max Act: 0.2009. Min Act: -0.0001

Data Index: 1352273 (C4 (Web Text))

Max Activating Token Index: 925