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

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

Neuron 51 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1600. Min Act: -0.1700

Data Index: 101858 (C4 (Web Text))

Max Activating Token Index: 492

Text #1

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1600. Min Act: -0.1700

Data Index: 1468958 (Python Code)

Max Activating Token Index: 572

Text #2

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1128. Min Act: -0.1700

Data Index: 1551827 (Python Code)

Max Activating Token Index: 780

Text #3

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1600. Min Act: -0.1700

Data Index: 382147 (C4 (Web Text))

Max Activating Token Index: 45

Text #4

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1600. Min Act: -0.1700

Data Index: 1023228 (C4 (Web Text))

Max Activating Token Index: 632

Text #5

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0970. Min Act: -0.1700

Data Index: 848423 (C4 (Web Text))

Max Activating Token Index: 604

Text #6

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1285. Min Act: -0.1700

Data Index: 1557831 (Python Code)

Max Activating Token Index: 595

Text #7

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1285. Min Act: -0.1700

Data Index: 1416325 (Python Code)

Max Activating Token Index: 778

Text #8

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1285. Min Act: -0.1700

Data Index: 1433085 (Python Code)

Max Activating Token Index: 673

Text #9

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1758. Min Act: -0.1700

Data Index: 490333 (C4 (Web Text))

Max Activating Token Index: 571

Text #10

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1128. Min Act: -0.1700

Data Index: 1493934 (Python Code)

Max Activating Token Index: 761

Text #11

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1285. Min Act: -0.1700

Data Index: 1508317 (Python Code)

Max Activating Token Index: 80

Text #12

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1285. Min Act: -0.1700

Data Index: 834504 (C4 (Web Text))

Max Activating Token Index: 588

Text #13

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.1128. Min Act: -0.1700

Data Index: 5810 (C4 (Web Text))

Max Activating Token Index: 567

Text #14

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0654. Min Act: -0.1700

Data Index: 1637262 (Python Code)

Max Activating Token Index: 449

Text #15

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0970. Min Act: -0.1700

Data Index: 1370109 (Python Code)

Max Activating Token Index: 74

Text #16

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0812. Min Act: -0.1700

Data Index: 1620151 (Python Code)

Max Activating Token Index: 929

Text #17

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0970. Min Act: -0.1700

Data Index: 1496069 (Python Code)

Max Activating Token Index: 149

Text #18

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0812. Min Act: -0.1700

Data Index: 1362782 (Python Code)

Max Activating Token Index: 701

Text #19

Max Range: 3.1758. Min Range: -3.1758

Max Act: 3.0812. Min Act: -0.1700

Data Index: 749112 (C4 (Web Text))

Max Activating Token Index: 341