Model: GELU Model: 4 Layers, 2048 Neurons per Layer

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

Neuron 576 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')

Text #0

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.7255. Min Act: -0.1696

Data Index: 1624863 (Python Code)

Max Activating Token Index: 692

Text #1

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.4755. Min Act: -0.1700

Data Index: 1241556 (C4 (Web Text))

Max Activating Token Index: 253

Text #2

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.4755. Min Act: -0.1699

Data Index: 1371602 (Python Code)

Max Activating Token Index: 630

Text #3

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.4130. Min Act: -0.1700

Data Index: 796322 (C4 (Web Text))

Max Activating Token Index: 796

Text #4

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.4130. Min Act: -0.1699

Data Index: 1406924 (Python Code)

Max Activating Token Index: 868

Text #5

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.3193. Min Act: -0.1689

Data Index: 1412180 (Python Code)

Max Activating Token Index: 74

Text #6

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.2255. Min Act: -0.1700

Data Index: 565742 (C4 (Web Text))

Max Activating Token Index: 900

Text #7

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1943. Min Act: -0.1700

Data Index: 1405384 (Python Code)

Max Activating Token Index: 506

Text #8

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1630. Min Act: -0.1700

Data Index: 359749 (C4 (Web Text))

Max Activating Token Index: 391

Text #9

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1630. Min Act: -0.1700

Data Index: 1221922 (C4 (Web Text))

Max Activating Token Index: 927

Text #10

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1318. Min Act: -0.1699

Data Index: 1549610 (Python Code)

Max Activating Token Index: 857

Text #11

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1943. Min Act: -0.1699

Data Index: 1404421 (Python Code)

Max Activating Token Index: 121

Text #12

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1943. Min Act: -0.1700

Data Index: 1544541 (Python Code)

Max Activating Token Index: 981

Text #13

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1630. Min Act: -0.1700

Data Index: 197734 (C4 (Web Text))

Max Activating Token Index: 503

Text #14

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1318. Min Act: -0.1700

Data Index: 1518676 (Python Code)

Max Activating Token Index: 643

Text #15

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1005. Min Act: -0.1698

Data Index: 1654733 (Python Code)

Max Activating Token Index: 544

Text #16

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.0692. Min Act: -0.1700

Data Index: 512130 (C4 (Web Text))

Max Activating Token Index: 370

Text #17

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.0380. Min Act: -0.1700

Data Index: 705936 (C4 (Web Text))

Max Activating Token Index: 585

Text #18

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.0380. Min Act: -0.1700

Data Index: 1269598 (C4 (Web Text))

Max Activating Token Index: 1005

Text #19

Max Range: 5.7255. Min Range: -5.7255

Max Act: 5.1318. Min Act: -0.1700

Data Index: 561683 (C4 (Web Text))

Max Activating Token Index: 535