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

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

Neuron 1910 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')

Text #0

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.1631. Min Act: -0.1700

Data Index: 1502336 (Python Code)

Max Activating Token Index: 13

Text #1

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.1473. Min Act: -0.1700

Data Index: 977150 (C4 (Web Text))

Max Activating Token Index: 952

Text #2

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0527. Min Act: -0.1700

Data Index: 1653942 (Python Code)

Max Activating Token Index: 387

Text #3

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0685. Min Act: -0.1700

Data Index: 710922 (C4 (Web Text))

Max Activating Token Index: 856

Text #4

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.1158. Min Act: -0.1700

Data Index: 293991 (C4 (Web Text))

Max Activating Token Index: 851

Text #5

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.1000. Min Act: -0.1700

Data Index: 639602 (C4 (Web Text))

Max Activating Token Index: 329

Text #6

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0527. Min Act: -0.1700

Data Index: 1454253 (Python Code)

Max Activating Token Index: 849

Text #7

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0211. Min Act: -0.1700

Data Index: 1328369 (C4 (Web Text))

Max Activating Token Index: 774

Text #8

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9895. Min Act: -0.1700

Data Index: 1522052 (Python Code)

Max Activating Token Index: 706

Text #9

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0369. Min Act: -0.1700

Data Index: 34623 (C4 (Web Text))

Max Activating Token Index: 918

Text #10

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0369. Min Act: -0.1700

Data Index: 1541340 (Python Code)

Max Activating Token Index: 647

Text #11

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0211. Min Act: -0.1700

Data Index: 1463384 (Python Code)

Max Activating Token Index: 548

Text #12

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9737. Min Act: -0.1700

Data Index: 1623543 (Python Code)

Max Activating Token Index: 713

Text #13

Max Range: 3.1631. Min Range: -3.1631

Max Act: 3.0211. Min Act: -0.1700

Data Index: 85601 (C4 (Web Text))

Max Activating Token Index: 412

Text #14

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9737. Min Act: -0.1700

Data Index: 1650244 (Python Code)

Max Activating Token Index: 290

Text #15

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9737. Min Act: -0.1700

Data Index: 49586 (C4 (Web Text))

Max Activating Token Index: 691

Text #16

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9895. Min Act: -0.1700

Data Index: 535587 (C4 (Web Text))

Max Activating Token Index: 1017

Text #17

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9895. Min Act: -0.1700

Data Index: 1613989 (Python Code)

Max Activating Token Index: 11

Text #18

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9895. Min Act: -0.1700

Data Index: 1448505 (Python Code)

Max Activating Token Index: 815

Text #19

Max Range: 3.1631. Min Range: -3.1631

Max Act: 2.9420. Min Act: -0.1700

Data Index: 1623046 (Python Code)

Max Activating Token Index: 861