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

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

Neuron 1414 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: 5.4169. Min Range: -5.4169

Max Act: 5.4169. Min Act: -0.1700

Data Index: 1139035 (C4 (Web Text))

Max Activating Token Index: 922

Text #1

Max Range: 5.4169. Min Range: -5.4169

Max Act: 5.0575. Min Act: -0.1700

Data Index: 1527667 (Python Code)

Max Activating Token Index: 308

Text #2

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.9638. Min Act: -0.1700

Data Index: 615872 (C4 (Web Text))

Max Activating Token Index: 970

Text #3

Max Range: 5.4169. Min Range: -5.4169

Max Act: 5.0419. Min Act: -0.1700

Data Index: 671012 (C4 (Web Text))

Max Activating Token Index: 987

Text #4

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.9950. Min Act: -0.1700

Data Index: 1551842 (Python Code)

Max Activating Token Index: 432

Text #5

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.9481. Min Act: -0.1700

Data Index: 744327 (C4 (Web Text))

Max Activating Token Index: 997

Text #6

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8700. Min Act: -0.1700

Data Index: 1199840 (C4 (Web Text))

Max Activating Token Index: 907

Text #7

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.9013. Min Act: -0.1700

Data Index: 729803 (C4 (Web Text))

Max Activating Token Index: 869

Text #8

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8700. Min Act: -0.1700

Data Index: 1515759 (Python Code)

Max Activating Token Index: 17

Text #9

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8231. Min Act: -0.1700

Data Index: 765214 (C4 (Web Text))

Max Activating Token Index: 220

Text #10

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8388. Min Act: -0.1700

Data Index: 959544 (C4 (Web Text))

Max Activating Token Index: 210

Text #11

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8700. Min Act: -0.1700

Data Index: 1427041 (Python Code)

Max Activating Token Index: 116

Text #12

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.7450. Min Act: -0.1700

Data Index: 1257843 (C4 (Web Text))

Max Activating Token Index: 222

Text #13

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8231. Min Act: -0.1700

Data Index: 835806 (C4 (Web Text))

Max Activating Token Index: 815

Text #14

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.7919. Min Act: -0.1700

Data Index: 72094 (C4 (Web Text))

Max Activating Token Index: 849

Text #15

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.7919. Min Act: -0.1700

Data Index: 190574 (C4 (Web Text))

Max Activating Token Index: 416

Text #16

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8544. Min Act: -0.1700

Data Index: 198190 (C4 (Web Text))

Max Activating Token Index: 421

Text #17

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.7606. Min Act: -0.1700

Data Index: 1583384 (Python Code)

Max Activating Token Index: 992

Text #18

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.8075. Min Act: -0.1700

Data Index: 655315 (C4 (Web Text))

Max Activating Token Index: 897

Text #19

Max Range: 5.4169. Min Range: -5.4169

Max Act: 4.7606. Min Act: -0.1700

Data Index: 976482 (C4 (Web Text))

Max Activating Token Index: 167