Model: GELU Model: 2 Layers, 2048 Neurons per Layer

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

Neuron 18 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-2l')

Text #0

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.4000. Min Act: -0.1700

Data Index: 1108476 (C4 (Web Text))

Max Activating Token Index: 30

Text #1

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.1187. Min Act: -0.1700

Data Index: 492587 (C4 (Web Text))

Max Activating Token Index: 114

Text #2

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.1500. Min Act: -0.1700

Data Index: 832488 (C4 (Web Text))

Max Activating Token Index: 31

Text #3

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.1187. Min Act: -0.1700

Data Index: 991034 (C4 (Web Text))

Max Activating Token Index: 59

Text #4

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.0406. Min Act: -0.1700

Data Index: 1092876 (C4 (Web Text))

Max Activating Token Index: 1017

Text #5

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.0562. Min Act: -0.1700

Data Index: 1350458 (C4 (Web Text))

Max Activating Token Index: 29

Text #6

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.0250. Min Act: -0.1700

Data Index: 172376 (C4 (Web Text))

Max Activating Token Index: 221

Text #7

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9625. Min Act: -0.1700

Data Index: 1244008 (C4 (Web Text))

Max Activating Token Index: 320

Text #8

Max Range: 5.4000. Min Range: -5.4000

Max Act: 5.0406. Min Act: -0.1700

Data Index: 1523392 (Python Code)

Max Activating Token Index: 73

Text #9

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9937. Min Act: -0.1699

Data Index: 1068285 (C4 (Web Text))

Max Activating Token Index: 653

Text #10

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9781. Min Act: -0.1700

Data Index: 968863 (C4 (Web Text))

Max Activating Token Index: 416

Text #11

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9937. Min Act: -0.1700

Data Index: 954769 (C4 (Web Text))

Max Activating Token Index: 304

Text #12

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9625. Min Act: -0.1699

Data Index: 540026 (C4 (Web Text))

Max Activating Token Index: 115

Text #13

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9937. Min Act: -0.1700

Data Index: 432860 (C4 (Web Text))

Max Activating Token Index: 617

Text #14

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9156. Min Act: -0.1700

Data Index: 1222103 (C4 (Web Text))

Max Activating Token Index: 318

Text #15

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9156. Min Act: -0.1700

Data Index: 241905 (C4 (Web Text))

Max Activating Token Index: 854

Text #16

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9781. Min Act: -0.1700

Data Index: 981226 (C4 (Web Text))

Max Activating Token Index: 797

Text #17

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9781. Min Act: -0.1700

Data Index: 1652317 (Python Code)

Max Activating Token Index: 41

Text #18

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9781. Min Act: -0.1700

Data Index: 350014 (C4 (Web Text))

Max Activating Token Index: 324

Text #19

Max Range: 5.4000. Min Range: -5.4000

Max Act: 4.9156. Min Act: -0.1700

Data Index: 695815 (C4 (Web Text))

Max Activating Token Index: 84