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

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

Neuron 3 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: 8.6112. Min Range: -8.6112

Max Act: 8.6112. Min Act: -0.1700

Data Index: 1375545 (Python Code)

Max Activating Token Index: 640

Text #1

Max Range: 8.6112. Min Range: -8.6112

Max Act: 8.2987. Min Act: -0.1700

Data Index: 1489674 (Python Code)

Max Activating Token Index: 520

Text #2

Max Range: 8.6112. Min Range: -8.6112

Max Act: 8.1112. Min Act: -0.1700

Data Index: 1507842 (Python Code)

Max Activating Token Index: 603

Text #3

Max Range: 8.6112. Min Range: -8.6112

Max Act: 7.3612. Min Act: -0.1700

Data Index: 593933 (C4 (Web Text))

Max Activating Token Index: 622

Text #4

Max Range: 8.6112. Min Range: -8.6112

Max Act: 7.0174. Min Act: -0.1700

Data Index: 409683 (C4 (Web Text))

Max Activating Token Index: 615

Text #5

Max Range: 8.6112. Min Range: -8.6112

Max Act: 6.7049. Min Act: -0.1700

Data Index: 1499970 (Python Code)

Max Activating Token Index: 713

Text #6

Max Range: 8.6112. Min Range: -8.6112

Max Act: 6.7674. Min Act: -0.1700

Data Index: 84360 (C4 (Web Text))

Max Activating Token Index: 603

Text #7

Max Range: 8.6112. Min Range: -8.6112

Max Act: 6.5487. Min Act: -0.1700

Data Index: 779993 (C4 (Web Text))

Max Activating Token Index: 715

Text #8

Max Range: 8.6112. Min Range: -8.6112

Max Act: 6.3299. Min Act: -0.1700

Data Index: 1647089 (Python Code)

Max Activating Token Index: 674

Text #9

Max Range: 8.6112. Min Range: -8.6112

Max Act: 6.2987. Min Act: -0.1700

Data Index: 1076712 (C4 (Web Text))

Max Activating Token Index: 760

Text #10

Max Range: 8.6112. Min Range: -8.6112

Max Act: 6.2362. Min Act: -0.1700

Data Index: 1486451 (Python Code)

Max Activating Token Index: 359

Text #11

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.9237. Min Act: -0.1700

Data Index: 1421588 (Python Code)

Max Activating Token Index: 818

Text #12

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.8924. Min Act: -0.1700

Data Index: 1416223 (Python Code)

Max Activating Token Index: 748

Text #13

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.8612. Min Act: -0.1700

Data Index: 1464341 (Python Code)

Max Activating Token Index: 634

Text #14

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.7674. Min Act: -0.1700

Data Index: 1533678 (Python Code)

Max Activating Token Index: 554

Text #15

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.7674. Min Act: -0.1700

Data Index: 472631 (C4 (Web Text))

Max Activating Token Index: 627

Text #16

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.7674. Min Act: -0.1700

Data Index: 1510450 (Python Code)

Max Activating Token Index: 676

Text #17

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.7674. Min Act: -0.1700

Data Index: 1362968 (Python Code)

Max Activating Token Index: 706

Text #18

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.7674. Min Act: -0.1700

Data Index: 492955 (C4 (Web Text))

Max Activating Token Index: 742

Text #19

Max Range: 8.6112. Min Range: -8.6112

Max Act: 5.6424. Min Act: -0.1700

Data Index: 177345 (C4 (Web Text))

Max Activating Token Index: 480