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

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

Neuron 297 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: 7.1218. Min Range: -7.1218

Max Act: 7.1218. Min Act: -0.1700

Data Index: 1455998 (Python Code)

Max Activating Token Index: 322

Text #1

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.9343. Min Act: -0.1700

Data Index: 943271 (C4 (Web Text))

Max Activating Token Index: 761

Text #2

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.7780. Min Act: -0.1700

Data Index: 241895 (C4 (Web Text))

Max Activating Token Index: 907

Text #3

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.6999. Min Act: -0.1700

Data Index: 419067 (C4 (Web Text))

Max Activating Token Index: 972

Text #4

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.7155. Min Act: -0.1700

Data Index: 1615915 (Python Code)

Max Activating Token Index: 48

Text #5

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.6843. Min Act: -0.1700

Data Index: 724115 (C4 (Web Text))

Max Activating Token Index: 920

Text #6

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.6687. Min Act: -0.1700

Data Index: 1462375 (Python Code)

Max Activating Token Index: 61

Text #7

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.6218. Min Act: -0.1700

Data Index: 932389 (C4 (Web Text))

Max Activating Token Index: 610

Text #8

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.6062. Min Act: -0.1699

Data Index: 462932 (C4 (Web Text))

Max Activating Token Index: 807

Text #9

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.5749. Min Act: -0.1700

Data Index: 353376 (C4 (Web Text))

Max Activating Token Index: 961

Text #10

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.5905. Min Act: -0.1700

Data Index: 1398532 (Python Code)

Max Activating Token Index: 992

Text #11

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.5437. Min Act: -0.1700

Data Index: 692606 (C4 (Web Text))

Max Activating Token Index: 710

Text #12

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.5280. Min Act: -0.1700

Data Index: 2514 (C4 (Web Text))

Max Activating Token Index: 685

Text #13

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.4655. Min Act: -0.1700

Data Index: 1423921 (Python Code)

Max Activating Token Index: 287

Text #14

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.4812. Min Act: -0.1699

Data Index: 367892 (C4 (Web Text))

Max Activating Token Index: 932

Text #15

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.4968. Min Act: -0.1700

Data Index: 613775 (C4 (Web Text))

Max Activating Token Index: 266

Text #16

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.4655. Min Act: -0.1700

Data Index: 1341990 (C4 (Web Text))

Max Activating Token Index: 507

Text #17

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.5280. Min Act: -0.1700

Data Index: 1622879 (Python Code)

Max Activating Token Index: 552

Text #18

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.4343. Min Act: -0.1700

Data Index: 1179855 (C4 (Web Text))

Max Activating Token Index: 698

Text #19

Max Range: 7.1218. Min Range: -7.1218

Max Act: 6.4343. Min Act: -0.1700

Data Index: 1296584 (C4 (Web Text))

Max Activating Token Index: 738