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

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

Neuron 218 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: 5.5366. Min Range: -5.5366

Max Act: 5.5366. Min Act: -0.1700

Data Index: 1387314 (Python Code)

Max Activating Token Index: 680

Text #1

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.3804. Min Act: -0.1700

Data Index: 9695 (C4 (Web Text))

Max Activating Token Index: 291

Text #2

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.2866. Min Act: -0.1700

Data Index: 1309931 (C4 (Web Text))

Max Activating Token Index: 85

Text #3

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.2241. Min Act: -0.1700

Data Index: 134820 (C4 (Web Text))

Max Activating Token Index: 432

Text #4

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.1616. Min Act: -0.1700

Data Index: 1524334 (Python Code)

Max Activating Token Index: 922

Text #5

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.1616. Min Act: -0.1700

Data Index: 1075831 (C4 (Web Text))

Max Activating Token Index: 824

Text #6

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.1616. Min Act: -0.1699

Data Index: 1431770 (Python Code)

Max Activating Token Index: 510

Text #7

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0991. Min Act: -0.1700

Data Index: 215139 (C4 (Web Text))

Max Activating Token Index: 792

Text #8

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0991. Min Act: -0.1700

Data Index: 267245 (C4 (Web Text))

Max Activating Token Index: 857

Text #9

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0991. Min Act: -0.1700

Data Index: 1652457 (Python Code)

Max Activating Token Index: 115

Text #10

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0679. Min Act: -0.1699

Data Index: 1612116 (Python Code)

Max Activating Token Index: 425

Text #11

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0679. Min Act: -0.1700

Data Index: 303407 (C4 (Web Text))

Max Activating Token Index: 922

Text #12

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0679. Min Act: -0.1700

Data Index: 1095670 (C4 (Web Text))

Max Activating Token Index: 819

Text #13

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0054. Min Act: -0.1700

Data Index: 1118548 (C4 (Web Text))

Max Activating Token Index: 688

Text #14

Max Range: 5.5366. Min Range: -5.5366

Max Act: 4.9429. Min Act: -0.1700

Data Index: 1447196 (Python Code)

Max Activating Token Index: 940

Text #15

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0054. Min Act: -0.1700

Data Index: 1448941 (Python Code)

Max Activating Token Index: 303

Text #16

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0366. Min Act: -0.1700

Data Index: 597425 (C4 (Web Text))

Max Activating Token Index: 208

Text #17

Max Range: 5.5366. Min Range: -5.5366

Max Act: 4.9741. Min Act: -0.1700

Data Index: 995498 (C4 (Web Text))

Max Activating Token Index: 889

Text #18

Max Range: 5.5366. Min Range: -5.5366

Max Act: 4.9741. Min Act: -0.1700

Data Index: 306705 (C4 (Web Text))

Max Activating Token Index: 664

Text #19

Max Range: 5.5366. Min Range: -5.5366

Max Act: 5.0054. Min Act: -0.1700

Data Index: 683044 (C4 (Web Text))

Max Activating Token Index: 865