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

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

Neuron 507 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.2087. Min Act: -0.1700

Data Index: 647044 (C4 (Web Text))

Max Activating Token Index: 389

Text #1

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.0720. Min Act: -0.1700

Data Index: 919559 (C4 (Web Text))

Max Activating Token Index: 24

Text #2

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.0212. Min Act: -0.1700

Data Index: 174544 (C4 (Web Text))

Max Activating Token Index: 152

Text #3

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.0094. Min Act: -0.1699

Data Index: 609415 (C4 (Web Text))

Max Activating Token Index: 49

Text #4

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.0329. Min Act: -0.1700

Data Index: 869072 (C4 (Web Text))

Max Activating Token Index: 939

Text #5

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.0055. Min Act: -0.1700

Data Index: 1326765 (C4 (Web Text))

Max Activating Token Index: 20

Text #6

Max Range: 4.2087. Min Range: -4.2087

Max Act: 4.0016. Min Act: -0.1700

Data Index: 222246 (C4 (Web Text))

Max Activating Token Index: 903

Text #7

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9586. Min Act: -0.1700

Data Index: 271202 (C4 (Web Text))

Max Activating Token Index: 36

Text #8

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9703. Min Act: -0.1700

Data Index: 1395728 (Python Code)

Max Activating Token Index: 359

Text #9

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9352. Min Act: -0.1700

Data Index: 1142227 (C4 (Web Text))

Max Activating Token Index: 775

Text #10

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9508. Min Act: -0.1700

Data Index: 295009 (C4 (Web Text))

Max Activating Token Index: 566

Text #11

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9899. Min Act: -0.1700

Data Index: 115544 (C4 (Web Text))

Max Activating Token Index: 51

Text #12

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9625. Min Act: -0.1700

Data Index: 1275722 (C4 (Web Text))

Max Activating Token Index: 143

Text #13

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9664. Min Act: -0.1700

Data Index: 1418110 (Python Code)

Max Activating Token Index: 380

Text #14

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9508. Min Act: -0.1700

Data Index: 28613 (C4 (Web Text))

Max Activating Token Index: 69

Text #15

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9117. Min Act: -0.1700

Data Index: 241826 (C4 (Web Text))

Max Activating Token Index: 886

Text #16

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9508. Min Act: -0.1700

Data Index: 1653552 (Python Code)

Max Activating Token Index: 505

Text #17

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9391. Min Act: -0.1700

Data Index: 962969 (C4 (Web Text))

Max Activating Token Index: 88

Text #18

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9430. Min Act: -0.1700

Data Index: 626211 (C4 (Web Text))

Max Activating Token Index: 554

Text #19

Max Range: 4.2087. Min Range: -4.2087

Max Act: 3.9782. Min Act: -0.1700

Data Index: 279377 (C4 (Web Text))

Max Activating Token Index: 700