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

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

Neuron 1036 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: 5.3240. Min Range: -5.3240

Max Act: 5.3240. Min Act: -0.1700

Data Index: 573402 (C4 (Web Text))

Max Activating Token Index: 800

Text #1

Max Range: 5.3240. Min Range: -5.3240

Max Act: 5.2302. Min Act: -0.1700

Data Index: 6166 (C4 (Web Text))

Max Activating Token Index: 32

Text #2

Max Range: 5.3240. Min Range: -5.3240

Max Act: 5.0427. Min Act: -0.1700

Data Index: 453991 (C4 (Web Text))

Max Activating Token Index: 420

Text #3

Max Range: 5.3240. Min Range: -5.3240

Max Act: 5.0115. Min Act: -0.1700

Data Index: 96563 (C4 (Web Text))

Max Activating Token Index: 471

Text #4

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.9802. Min Act: -0.1700

Data Index: 1124978 (C4 (Web Text))

Max Activating Token Index: 779

Text #5

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.9490. Min Act: -0.1700

Data Index: 751055 (C4 (Web Text))

Max Activating Token Index: 357

Text #6

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.9490. Min Act: -0.1700

Data Index: 692513 (C4 (Web Text))

Max Activating Token Index: 78

Text #7

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.9490. Min Act: -0.1700

Data Index: 459983 (C4 (Web Text))

Max Activating Token Index: 207

Text #8

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.8865. Min Act: -0.1700

Data Index: 794415 (C4 (Web Text))

Max Activating Token Index: 940

Text #9

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.8865. Min Act: -0.1700

Data Index: 1145319 (C4 (Web Text))

Max Activating Token Index: 747

Text #10

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.8240. Min Act: -0.1700

Data Index: 620870 (C4 (Web Text))

Max Activating Token Index: 1007

Text #11

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7927. Min Act: -0.1700

Data Index: 728506 (C4 (Web Text))

Max Activating Token Index: 1002

Text #12

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7927. Min Act: -0.1700

Data Index: 1155085 (C4 (Web Text))

Max Activating Token Index: 145

Text #13

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7615. Min Act: -0.1700

Data Index: 91738 (C4 (Web Text))

Max Activating Token Index: 31

Text #14

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7615. Min Act: -0.1700

Data Index: 706166 (C4 (Web Text))

Max Activating Token Index: 858

Text #15

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7615. Min Act: -0.1699

Data Index: 107523 (C4 (Web Text))

Max Activating Token Index: 155

Text #16

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7615. Min Act: -0.1700

Data Index: 283711 (C4 (Web Text))

Max Activating Token Index: 605

Text #17

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7302. Min Act: -0.1700

Data Index: 1554702 (Python Code)

Max Activating Token Index: 893

Text #18

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7302. Min Act: -0.1700

Data Index: 494610 (C4 (Web Text))

Max Activating Token Index: 965

Text #19

Max Range: 5.3240. Min Range: -5.3240

Max Act: 4.7302. Min Act: -0.1700

Data Index: 136472 (C4 (Web Text))

Max Activating Token Index: 167