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

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

Neuron 66 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 4.0559. Min Range: -4.0559

Max Act: 4.0559. Min Act: -0.1700

Data Index: 1015866 (C4 (Web Text))

Max Activating Token Index: 642

Text #1

Max Range: 4.0559. Min Range: -4.0559

Max Act: 4.0559. Min Act: -0.1700

Data Index: 983288 (C4 (Web Text))

Max Activating Token Index: 84

Text #2

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.9934. Min Act: -0.1700

Data Index: 1271001 (C4 (Web Text))

Max Activating Token Index: 808

Text #3

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.9308. Min Act: -0.1700

Data Index: 712335 (C4 (Web Text))

Max Activating Token Index: 997

Text #4

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8370. Min Act: -0.1700

Data Index: 116376 (C4 (Web Text))

Max Activating Token Index: 719

Text #5

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8683. Min Act: -0.1700

Data Index: 1360958 (Python Code)

Max Activating Token Index: 801

Text #6

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8995. Min Act: -0.1700

Data Index: 330824 (C4 (Web Text))

Max Activating Token Index: 764

Text #7

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8370. Min Act: -0.1700

Data Index: 357251 (C4 (Web Text))

Max Activating Token Index: 151

Text #8

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8995. Min Act: -0.1699

Data Index: 582144 (C4 (Web Text))

Max Activating Token Index: 411

Text #9

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8057. Min Act: -0.1700

Data Index: 603338 (C4 (Web Text))

Max Activating Token Index: 824

Text #10

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8370. Min Act: -0.1700

Data Index: 234362 (C4 (Web Text))

Max Activating Token Index: 88

Text #11

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8057. Min Act: -0.1700

Data Index: 780023 (C4 (Web Text))

Max Activating Token Index: 108

Text #12

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8370. Min Act: -0.1700

Data Index: 861215 (C4 (Web Text))

Max Activating Token Index: 826

Text #13

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8057. Min Act: -0.1700

Data Index: 1579700 (Python Code)

Max Activating Token Index: 654

Text #14

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.7744. Min Act: -0.1700

Data Index: 47838 (C4 (Web Text))

Max Activating Token Index: 143

Text #15

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8057. Min Act: -0.1700

Data Index: 226100 (C4 (Web Text))

Max Activating Token Index: 676

Text #16

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8370. Min Act: -0.1700

Data Index: 921847 (C4 (Web Text))

Max Activating Token Index: 482

Text #17

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8057. Min Act: -0.1700

Data Index: 381881 (C4 (Web Text))

Max Activating Token Index: 582

Text #18

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.8057. Min Act: -0.1700

Data Index: 1649711 (Python Code)

Max Activating Token Index: 773

Text #19

Max Range: 4.0559. Min Range: -4.0559

Max Act: 3.7744. Min Act: -0.1700

Data Index: 1361823 (Python Code)

Max Activating Token Index: 51