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

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

Neuron 197 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: 5.1987. Min Range: -5.1987

Max Act: 5.1987. Min Act: -0.1700

Data Index: 763204 (C4 (Web Text))

Max Activating Token Index: 662

Text #1

Max Range: 5.1987. Min Range: -5.1987

Max Act: 5.1674. Min Act: -0.1699

Data Index: 236847 (C4 (Web Text))

Max Activating Token Index: 961

Text #2

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.9799. Min Act: -0.1700

Data Index: 1380757 (Python Code)

Max Activating Token Index: 407

Text #3

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8862. Min Act: -0.1700

Data Index: 1370596 (Python Code)

Max Activating Token Index: 779

Text #4

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.9174. Min Act: -0.1700

Data Index: 168669 (C4 (Web Text))

Max Activating Token Index: 586

Text #5

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8862. Min Act: -0.1700

Data Index: 246969 (C4 (Web Text))

Max Activating Token Index: 822

Text #6

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8862. Min Act: -0.1700

Data Index: 1488911 (Python Code)

Max Activating Token Index: 426

Text #7

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8862. Min Act: -0.1700

Data Index: 152856 (C4 (Web Text))

Max Activating Token Index: 472

Text #8

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8862. Min Act: -0.1700

Data Index: 853525 (C4 (Web Text))

Max Activating Token Index: 1020

Text #9

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8862. Min Act: -0.1700

Data Index: 1329076 (C4 (Web Text))

Max Activating Token Index: 294

Text #10

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.9174. Min Act: -0.1700

Data Index: 662199 (C4 (Web Text))

Max Activating Token Index: 725

Text #11

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8237. Min Act: -0.1696

Data Index: 904495 (C4 (Web Text))

Max Activating Token Index: 713

Text #12

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8237. Min Act: -0.1700

Data Index: 1588671 (Python Code)

Max Activating Token Index: 1005

Text #13

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8549. Min Act: -0.1699

Data Index: 1285584 (C4 (Web Text))

Max Activating Token Index: 538

Text #14

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.8237. Min Act: -0.1700

Data Index: 476838 (C4 (Web Text))

Max Activating Token Index: 524

Text #15

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.7924. Min Act: -0.1700

Data Index: 1410958 (Python Code)

Max Activating Token Index: 715

Text #16

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.7924. Min Act: -0.1700

Data Index: 1607783 (Python Code)

Max Activating Token Index: 808

Text #17

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.7612. Min Act: -0.1700

Data Index: 222125 (C4 (Web Text))

Max Activating Token Index: 475

Text #18

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.7924. Min Act: -0.1700

Data Index: 738979 (C4 (Web Text))

Max Activating Token Index: 687

Text #19

Max Range: 5.1987. Min Range: -5.1987

Max Act: 4.7299. Min Act: -0.1700

Data Index: 574855 (C4 (Web Text))

Max Activating Token Index: 405