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

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

Neuron 372 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: 5.8900. Min Range: -5.8900

Max Act: 5.8900. Min Act: -0.1700

Data Index: 1608537 (Python Code)

Max Activating Token Index: 803

Text #1

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7025. Min Act: -0.1700

Data Index: 1199055 (C4 (Web Text))

Max Activating Token Index: 941

Text #2

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7337. Min Act: -0.1700

Data Index: 1474991 (Python Code)

Max Activating Token Index: 997

Text #3

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.8275. Min Act: -0.1700

Data Index: 1596864 (Python Code)

Max Activating Token Index: 246

Text #4

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7025. Min Act: -0.1700

Data Index: 1473936 (Python Code)

Max Activating Token Index: 603

Text #5

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7337. Min Act: -0.1700

Data Index: 1386417 (Python Code)

Max Activating Token Index: 116

Text #6

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7025. Min Act: -0.1700

Data Index: 811506 (C4 (Web Text))

Max Activating Token Index: 751

Text #7

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6712. Min Act: -0.1700

Data Index: 1466009 (Python Code)

Max Activating Token Index: 944

Text #8

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7025. Min Act: -0.1700

Data Index: 1436168 (Python Code)

Max Activating Token Index: 690

Text #9

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.5775. Min Act: -0.1700

Data Index: 1520213 (Python Code)

Max Activating Token Index: 662

Text #10

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6400. Min Act: -0.1700

Data Index: 1295388 (C4 (Web Text))

Max Activating Token Index: 213

Text #11

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.7025. Min Act: -0.1700

Data Index: 1123438 (C4 (Web Text))

Max Activating Token Index: 629

Text #12

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6712. Min Act: -0.1700

Data Index: 1503054 (Python Code)

Max Activating Token Index: 823

Text #13

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6087. Min Act: -0.1700

Data Index: 532533 (C4 (Web Text))

Max Activating Token Index: 1003

Text #14

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6400. Min Act: -0.1700

Data Index: 1170005 (C4 (Web Text))

Max Activating Token Index: 223

Text #15

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.5150. Min Act: -0.1700

Data Index: 781021 (C4 (Web Text))

Max Activating Token Index: 631

Text #16

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6400. Min Act: -0.1700

Data Index: 1068356 (C4 (Web Text))

Max Activating Token Index: 259

Text #17

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6400. Min Act: -0.1700

Data Index: 1415154 (Python Code)

Max Activating Token Index: 250

Text #18

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.6087. Min Act: -0.1700

Data Index: 1498553 (Python Code)

Max Activating Token Index: 643

Text #19

Max Range: 5.8900. Min Range: -5.8900

Max Act: 5.5462. Min Act: -0.1700

Data Index: 47024 (C4 (Web Text))

Max Activating Token Index: 980