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

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

Neuron 1721 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: 5.4297. Min Range: -5.4297

Max Act: 5.4297. Min Act: -0.1700

Data Index: 1551002 (Python Code)

Max Activating Token Index: 999

Text #1

Max Range: 5.4297. Min Range: -5.4297

Max Act: 5.1172. Min Act: -0.1700

Data Index: 331532 (C4 (Web Text))

Max Activating Token Index: 859

Text #2

Max Range: 5.4297. Min Range: -5.4297

Max Act: 5.0547. Min Act: -0.1699

Data Index: 1318207 (C4 (Web Text))

Max Activating Token Index: 284

Text #3

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.9609. Min Act: -0.1700

Data Index: 552775 (C4 (Web Text))

Max Activating Token Index: 270

Text #4

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.9297. Min Act: -0.1700

Data Index: 1484993 (Python Code)

Max Activating Token Index: 574

Text #5

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8984. Min Act: -0.1700

Data Index: 1512561 (Python Code)

Max Activating Token Index: 312

Text #6

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8047. Min Act: -0.1700

Data Index: 951439 (C4 (Web Text))

Max Activating Token Index: 944

Text #7

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8359. Min Act: -0.1700

Data Index: 1445426 (Python Code)

Max Activating Token Index: 515

Text #8

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8047. Min Act: -0.1700

Data Index: 695711 (C4 (Web Text))

Max Activating Token Index: 636

Text #9

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7734. Min Act: -0.1700

Data Index: 523288 (C4 (Web Text))

Max Activating Token Index: 133

Text #10

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8047. Min Act: -0.1700

Data Index: 1513784 (Python Code)

Max Activating Token Index: 186

Text #11

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7734. Min Act: -0.1700

Data Index: 1543824 (Python Code)

Max Activating Token Index: 43

Text #12

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7109. Min Act: -0.1700

Data Index: 1565393 (Python Code)

Max Activating Token Index: 891

Text #13

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7109. Min Act: -0.1700

Data Index: 1569189 (Python Code)

Max Activating Token Index: 917

Text #14

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7422. Min Act: -0.1700

Data Index: 253407 (C4 (Web Text))

Max Activating Token Index: 353

Text #15

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7734. Min Act: -0.1700

Data Index: 352685 (C4 (Web Text))

Max Activating Token Index: 794

Text #16

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8359. Min Act: -0.1700

Data Index: 754170 (C4 (Web Text))

Max Activating Token Index: 614

Text #17

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7734. Min Act: -0.1700

Data Index: 199490 (C4 (Web Text))

Max Activating Token Index: 441

Text #18

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.8047. Min Act: -0.1700

Data Index: 1034729 (C4 (Web Text))

Max Activating Token Index: 138

Text #19

Max Range: 5.4297. Min Range: -5.4297

Max Act: 4.7422. Min Act: -0.1700

Data Index: 1418591 (Python Code)

Max Activating Token Index: 1010