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

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

Neuron 400 in Layer 0

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.1851. Min Act: -0.1700

Data Index: 177193 (C4 (Web Text))

Max Activating Token Index: 31

Text #1

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.1851. Min Act: -0.1700

Data Index: 1017604 (C4 (Web Text))

Max Activating Token Index: 603

Text #2

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.1064. Min Act: -0.1700

Data Index: 1499661 (Python Code)

Max Activating Token Index: 65

Text #3

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.1064. Min Act: -0.1700

Data Index: 519093 (C4 (Web Text))

Max Activating Token Index: 104

Text #4

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0748. Min Act: -0.1700

Data Index: 1286867 (C4 (Web Text))

Max Activating Token Index: 175

Text #5

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0906. Min Act: -0.1700

Data Index: 638682 (C4 (Web Text))

Max Activating Token Index: 633

Text #6

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0906. Min Act: -0.1700

Data Index: 939642 (C4 (Web Text))

Max Activating Token Index: 140

Text #7

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0590. Min Act: -0.1700

Data Index: 689627 (C4 (Web Text))

Max Activating Token Index: 29

Text #8

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0117. Min Act: -0.1700

Data Index: 695384 (C4 (Web Text))

Max Activating Token Index: 541

Text #9

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0275. Min Act: -0.1700

Data Index: 779662 (C4 (Web Text))

Max Activating Token Index: 889

Text #10

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0275. Min Act: -0.1700

Data Index: 376197 (C4 (Web Text))

Max Activating Token Index: 614

Text #11

Max Range: 3.1851. Min Range: -3.1851

Max Act: 2.9959. Min Act: -0.1700

Data Index: 361450 (C4 (Web Text))

Max Activating Token Index: 199

Text #12

Max Range: 3.1851. Min Range: -3.1851

Max Act: 2.9959. Min Act: -0.1700

Data Index: 608564 (C4 (Web Text))

Max Activating Token Index: 108

Text #13

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0117. Min Act: -0.1700

Data Index: 1391566 (Python Code)

Max Activating Token Index: 3

Text #14

Max Range: 3.1851. Min Range: -3.1851

Max Act: 2.9959. Min Act: -0.1700

Data Index: 1635660 (Python Code)

Max Activating Token Index: 24

Text #15

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0117. Min Act: -0.1700

Data Index: 196613 (C4 (Web Text))

Max Activating Token Index: 944

Text #16

Max Range: 3.1851. Min Range: -3.1851

Max Act: 2.9959. Min Act: -0.1700

Data Index: 311770 (C4 (Web Text))

Max Activating Token Index: 947

Text #17

Max Range: 3.1851. Min Range: -3.1851

Max Act: 3.0117. Min Act: -0.1700

Data Index: 554894 (C4 (Web Text))

Max Activating Token Index: 600

Text #18

Max Range: 3.1851. Min Range: -3.1851

Max Act: 2.9325. Min Act: -0.1700

Data Index: 1516330 (Python Code)

Max Activating Token Index: 936

Text #19

Max Range: 3.1851. Min Range: -3.1851

Max Act: 2.9642. Min Act: -0.1700

Data Index: 29277 (C4 (Web Text))

Max Activating Token Index: 275