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

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

Neuron 19 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: 9.3934. Min Range: -9.3934

Max Act: 9.3934. Min Act: -0.1700

Data Index: 995947 (C4 (Web Text))

Max Activating Token Index: 519

Text #1

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.9247. Min Act: -0.1700

Data Index: 818224 (C4 (Web Text))

Max Activating Token Index: 859

Text #2

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.8309. Min Act: -0.1700

Data Index: 70989 (C4 (Web Text))

Max Activating Token Index: 459

Text #3

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.6434. Min Act: -0.1700

Data Index: 1589462 (Python Code)

Max Activating Token Index: 324

Text #4

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.5184. Min Act: -0.1700

Data Index: 1650926 (Python Code)

Max Activating Token Index: 861

Text #5

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.5497. Min Act: -0.1700

Data Index: 1279883 (C4 (Web Text))

Max Activating Token Index: 673

Text #6

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.6122. Min Act: -0.1700

Data Index: 1541715 (Python Code)

Max Activating Token Index: 236

Text #7

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.5184. Min Act: -0.1700

Data Index: 871323 (C4 (Web Text))

Max Activating Token Index: 350

Text #8

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.5184. Min Act: -0.1700

Data Index: 570045 (C4 (Web Text))

Max Activating Token Index: 970

Text #9

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.5184. Min Act: -0.1700

Data Index: 1543590 (Python Code)

Max Activating Token Index: 366

Text #10

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.4247. Min Act: -0.1700

Data Index: 491930 (C4 (Web Text))

Max Activating Token Index: 461

Text #11

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.3309. Min Act: -0.1700

Data Index: 1574001 (Python Code)

Max Activating Token Index: 626

Text #12

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.3309. Min Act: -0.1700

Data Index: 1293615 (C4 (Web Text))

Max Activating Token Index: 866

Text #13

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.3309. Min Act: -0.1700

Data Index: 1547385 (Python Code)

Max Activating Token Index: 924

Text #14

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.3622. Min Act: -0.1700

Data Index: 442956 (C4 (Web Text))

Max Activating Token Index: 70

Text #15

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.2684. Min Act: -0.1700

Data Index: 866953 (C4 (Web Text))

Max Activating Token Index: 970

Text #16

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.2059. Min Act: -0.1700

Data Index: 1413428 (Python Code)

Max Activating Token Index: 238

Text #17

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.2372. Min Act: -0.1700

Data Index: 572680 (C4 (Web Text))

Max Activating Token Index: 1013

Text #18

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.2684. Min Act: -0.1700

Data Index: 154825 (C4 (Web Text))

Max Activating Token Index: 396

Text #19

Max Range: 9.3934. Min Range: -9.3934

Max Act: 8.1747. Min Act: -0.1700

Data Index: 1630804 (Python Code)

Max Activating Token Index: 604