Model: GELU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 1038 in Layer 0

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-3l')

Text #0

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.6494. Min Act: -0.1700

Data Index: 1029611 (C4 (Web Text))

Max Activating Token Index: 968

Text #1

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.5555. Min Act: -0.1700

Data Index: 362874 (C4 (Web Text))

Max Activating Token Index: 958

Text #2

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.5868. Min Act: -0.1700

Data Index: 883245 (C4 (Web Text))

Max Activating Token Index: 115

Text #3

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.5711. Min Act: -0.1700

Data Index: 1254981 (C4 (Web Text))

Max Activating Token Index: 936

Text #4

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4928. Min Act: -0.1700

Data Index: 286263 (C4 (Web Text))

Max Activating Token Index: 354

Text #5

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.5085. Min Act: -0.1700

Data Index: 109001 (C4 (Web Text))

Max Activating Token Index: 474

Text #6

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4928. Min Act: -0.1700

Data Index: 104344 (C4 (Web Text))

Max Activating Token Index: 965

Text #7

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4928. Min Act: -0.1700

Data Index: 1314766 (C4 (Web Text))

Max Activating Token Index: 854

Text #8

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4771. Min Act: -0.1700

Data Index: 515607 (C4 (Web Text))

Max Activating Token Index: 116

Text #9

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4771. Min Act: -0.1700

Data Index: 1446941 (Python Code)

Max Activating Token Index: 680

Text #10

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4458. Min Act: -0.1700

Data Index: 1091072 (C4 (Web Text))

Max Activating Token Index: 763

Text #11

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4144. Min Act: -0.1700

Data Index: 1218334 (C4 (Web Text))

Max Activating Token Index: 739

Text #12

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3987. Min Act: -0.1700

Data Index: 245450 (C4 (Web Text))

Max Activating Token Index: 295

Text #13

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.4301. Min Act: -0.1700

Data Index: 1623995 (Python Code)

Max Activating Token Index: 373

Text #14

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3987. Min Act: -0.1700

Data Index: 996520 (C4 (Web Text))

Max Activating Token Index: 820

Text #15

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3987. Min Act: -0.1700

Data Index: 530588 (C4 (Web Text))

Max Activating Token Index: 250

Text #16

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3517. Min Act: -0.1700

Data Index: 35045 (C4 (Web Text))

Max Activating Token Index: 933

Text #17

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3674. Min Act: -0.1700

Data Index: 980220 (C4 (Web Text))

Max Activating Token Index: 860

Text #18

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3674. Min Act: -0.1700

Data Index: 825804 (C4 (Web Text))

Max Activating Token Index: 9

Text #19

Max Range: 3.6494. Min Range: -3.6494

Max Act: 3.3674. Min Act: -0.1700

Data Index: 1462323 (Python Code)

Max Activating Token Index: 590