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

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

Neuron 1556 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 8.3245. Min Range: -8.3245

Max Act: 8.3245. Min Act: -0.1699

Data Index: 919685 (C4 (Web Text))

Max Activating Token Index: 583

Text #1

Max Range: 8.3245. Min Range: -8.3245

Max Act: 8.1057. Min Act: -0.1699

Data Index: 423127 (C4 (Web Text))

Max Activating Token Index: 987

Text #2

Max Range: 8.3245. Min Range: -8.3245

Max Act: 8.1370. Min Act: -0.1699

Data Index: 1636714 (Python Code)

Max Activating Token Index: 522

Text #3

Max Range: 8.3245. Min Range: -8.3245

Max Act: 8.1057. Min Act: -0.1699

Data Index: 254915 (C4 (Web Text))

Max Activating Token Index: 938

Text #4

Max Range: 8.3245. Min Range: -8.3245

Max Act: 8.1057. Min Act: -0.1699

Data Index: 466450 (C4 (Web Text))

Max Activating Token Index: 607

Text #5

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.9182. Min Act: -0.1699

Data Index: 1427306 (Python Code)

Max Activating Token Index: 829

Text #6

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.9495. Min Act: -0.1699

Data Index: 861575 (C4 (Web Text))

Max Activating Token Index: 853

Text #7

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.9182. Min Act: -0.1699

Data Index: 97611 (C4 (Web Text))

Max Activating Token Index: 629

Text #8

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.9807. Min Act: -0.1699

Data Index: 349259 (C4 (Web Text))

Max Activating Token Index: 370

Text #9

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.8557. Min Act: -0.1699

Data Index: 1029398 (C4 (Web Text))

Max Activating Token Index: 1019

Text #10

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.8557. Min Act: -0.1699

Data Index: 1615407 (Python Code)

Max Activating Token Index: 816

Text #11

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.8870. Min Act: -0.1699

Data Index: 1405629 (Python Code)

Max Activating Token Index: 977

Text #12

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.7307. Min Act: -0.1699

Data Index: 285789 (C4 (Web Text))

Max Activating Token Index: 687

Text #13

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.6682. Min Act: -0.1699

Data Index: 1596976 (Python Code)

Max Activating Token Index: 646

Text #14

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.7620. Min Act: -0.1699

Data Index: 1188517 (C4 (Web Text))

Max Activating Token Index: 933

Text #15

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.6682. Min Act: -0.1699

Data Index: 1240945 (C4 (Web Text))

Max Activating Token Index: 85

Text #16

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.7307. Min Act: -0.1699

Data Index: 1038193 (C4 (Web Text))

Max Activating Token Index: 864

Text #17

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.6057. Min Act: -0.1699

Data Index: 578098 (C4 (Web Text))

Max Activating Token Index: 613

Text #18

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.6370. Min Act: -0.1699

Data Index: 1236070 (C4 (Web Text))

Max Activating Token Index: 953

Text #19

Max Range: 8.3245. Min Range: -8.3245

Max Act: 7.6370. Min Act: -0.1699

Data Index: 906212 (C4 (Web Text))

Max Activating Token Index: 453