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

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

Neuron 688 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.9064. Min Act: -0.1700

Data Index: 363652 (C4 (Web Text))

Max Activating Token Index: 831

Text #1

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.9064. Min Act: -0.1700

Data Index: 1245605 (C4 (Web Text))

Max Activating Token Index: 870

Text #2

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.8905. Min Act: -0.1700

Data Index: 365646 (C4 (Web Text))

Max Activating Token Index: 719

Text #3

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.8905. Min Act: -0.1700

Data Index: 517267 (C4 (Web Text))

Max Activating Token Index: 608

Text #4

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.8428. Min Act: -0.1700

Data Index: 557500 (C4 (Web Text))

Max Activating Token Index: 867

Text #5

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.8269. Min Act: -0.1700

Data Index: 1577686 (Python Code)

Max Activating Token Index: 303

Text #6

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.8428. Min Act: -0.1700

Data Index: 1110326 (C4 (Web Text))

Max Activating Token Index: 549

Text #7

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7792. Min Act: -0.1700

Data Index: 725485 (C4 (Web Text))

Max Activating Token Index: 90

Text #8

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7792. Min Act: -0.1700

Data Index: 353313 (C4 (Web Text))

Max Activating Token Index: 161

Text #9

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7632. Min Act: -0.1700

Data Index: 10501 (C4 (Web Text))

Max Activating Token Index: 41

Text #10

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7153. Min Act: -0.1700

Data Index: 1351114 (C4 (Web Text))

Max Activating Token Index: 900

Text #11

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7313. Min Act: -0.1700

Data Index: 41397 (C4 (Web Text))

Max Activating Token Index: 483

Text #12

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7951. Min Act: -0.1700

Data Index: 1188021 (C4 (Web Text))

Max Activating Token Index: 812

Text #13

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7632. Min Act: -0.1700

Data Index: 943275 (C4 (Web Text))

Max Activating Token Index: 155

Text #14

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7313. Min Act: -0.1700

Data Index: 781359 (C4 (Web Text))

Max Activating Token Index: 654

Text #15

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7153. Min Act: -0.1700

Data Index: 739597 (C4 (Web Text))

Max Activating Token Index: 745

Text #16

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7313. Min Act: -0.1700

Data Index: 296631 (C4 (Web Text))

Max Activating Token Index: 722

Text #17

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.6833. Min Act: -0.1700

Data Index: 1042268 (C4 (Web Text))

Max Activating Token Index: 962

Text #18

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7313. Min Act: -0.1700

Data Index: 1071645 (C4 (Web Text))

Max Activating Token Index: 358

Text #19

Max Range: 2.9064. Min Range: -2.9064

Max Act: 2.7153. Min Act: -0.1700

Data Index: 1109867 (C4 (Web Text))

Max Activating Token Index: 458