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

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

Neuron 2014 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: 5.5929. Min Range: -5.5929

Max Act: 5.4992. Min Act: -0.1700

Data Index: 1407042 (Python Code)

Max Activating Token Index: 409

Text #1

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4367. Min Act: -0.1700

Data Index: 1360930 (Python Code)

Max Activating Token Index: 848

Text #2

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4992. Min Act: -0.1700

Data Index: 554005 (C4 (Web Text))

Max Activating Token Index: 827

Text #3

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4367. Min Act: -0.1700

Data Index: 1539530 (Python Code)

Max Activating Token Index: 476

Text #4

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.5929. Min Act: -0.1700

Data Index: 610624 (C4 (Web Text))

Max Activating Token Index: 728

Text #5

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4992. Min Act: -0.1700

Data Index: 623979 (C4 (Web Text))

Max Activating Token Index: 133

Text #6

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4679. Min Act: -0.1700

Data Index: 823948 (C4 (Web Text))

Max Activating Token Index: 329

Text #7

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4054. Min Act: -0.1700

Data Index: 118310 (C4 (Web Text))

Max Activating Token Index: 716

Text #8

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4054. Min Act: -0.1700

Data Index: 1574888 (Python Code)

Max Activating Token Index: 258

Text #9

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4367. Min Act: -0.1700

Data Index: 428334 (C4 (Web Text))

Max Activating Token Index: 572

Text #10

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4367. Min Act: -0.1700

Data Index: 913068 (C4 (Web Text))

Max Activating Token Index: 38

Text #11

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.3742. Min Act: -0.1700

Data Index: 228336 (C4 (Web Text))

Max Activating Token Index: 1016

Text #12

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4054. Min Act: -0.1700

Data Index: 1079336 (C4 (Web Text))

Max Activating Token Index: 426

Text #13

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.4054. Min Act: -0.1700

Data Index: 1525957 (Python Code)

Max Activating Token Index: 573

Text #14

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.3742. Min Act: -0.1700

Data Index: 1073364 (C4 (Web Text))

Max Activating Token Index: 792

Text #15

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.3117. Min Act: -0.1700

Data Index: 1068285 (C4 (Web Text))

Max Activating Token Index: 125

Text #16

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.3742. Min Act: -0.1700

Data Index: 459402 (C4 (Web Text))

Max Activating Token Index: 293

Text #17

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.3742. Min Act: -0.1700

Data Index: 99371 (C4 (Web Text))

Max Activating Token Index: 607

Text #18

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.3742. Min Act: -0.1700

Data Index: 693607 (C4 (Web Text))

Max Activating Token Index: 990

Text #19

Max Range: 5.5929. Min Range: -5.5929

Max Act: 5.1554. Min Act: -0.1700

Data Index: 564593 (C4 (Web Text))

Max Activating Token Index: 810