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

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

Neuron 438 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: 7.4896. Min Range: -7.4896

Max Act: 7.4896. Min Act: -0.1700

Data Index: 989850 (C4 (Web Text))

Max Activating Token Index: 467

Text #1

Max Range: 7.4896. Min Range: -7.4896

Max Act: 7.2084. Min Act: -0.1700

Data Index: 622000 (C4 (Web Text))

Max Activating Token Index: 601

Text #2

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.9584. Min Act: -0.1700

Data Index: 669640 (C4 (Web Text))

Max Activating Token Index: 651

Text #3

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.9584. Min Act: -0.1700

Data Index: 1171682 (C4 (Web Text))

Max Activating Token Index: 514

Text #4

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.9271. Min Act: -0.1700

Data Index: 736586 (C4 (Web Text))

Max Activating Token Index: 169

Text #5

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.8646. Min Act: -0.1700

Data Index: 1008001 (C4 (Web Text))

Max Activating Token Index: 122

Text #6

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7709. Min Act: -0.1700

Data Index: 1089056 (C4 (Web Text))

Max Activating Token Index: 720

Text #7

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.8021. Min Act: -0.1700

Data Index: 1118460 (C4 (Web Text))

Max Activating Token Index: 760

Text #8

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.8334. Min Act: -0.1700

Data Index: 49777 (C4 (Web Text))

Max Activating Token Index: 806

Text #9

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7709. Min Act: -0.1700

Data Index: 720147 (C4 (Web Text))

Max Activating Token Index: 564

Text #10

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.8021. Min Act: -0.1700

Data Index: 1322767 (C4 (Web Text))

Max Activating Token Index: 765

Text #11

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.8334. Min Act: -0.1700

Data Index: 470924 (C4 (Web Text))

Max Activating Token Index: 652

Text #12

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.8021. Min Act: -0.1700

Data Index: 907645 (C4 (Web Text))

Max Activating Token Index: 315

Text #13

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7709. Min Act: -0.1700

Data Index: 1466474 (Python Code)

Max Activating Token Index: 564

Text #14

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7709. Min Act: -0.1700

Data Index: 1124986 (C4 (Web Text))

Max Activating Token Index: 733

Text #15

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7396. Min Act: -0.1700

Data Index: 215343 (C4 (Web Text))

Max Activating Token Index: 535

Text #16

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7709. Min Act: -0.1700

Data Index: 665857 (C4 (Web Text))

Max Activating Token Index: 915

Text #17

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7084. Min Act: -0.1700

Data Index: 1048328 (C4 (Web Text))

Max Activating Token Index: 741

Text #18

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7084. Min Act: -0.1700

Data Index: 1069479 (C4 (Web Text))

Max Activating Token Index: 177

Text #19

Max Range: 7.4896. Min Range: -7.4896

Max Act: 6.7396. Min Act: -0.1700

Data Index: 74120 (C4 (Web Text))

Max Activating Token Index: 861