Model: GELU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 853 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-3l')

Text #0

Max Range: 5.6080. Min Range: -5.6080

Max Act: 5.6080. Min Act: -0.1700

Data Index: 902464 (C4 (Web Text))

Max Activating Token Index: 984

Text #1

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.9830. Min Act: -0.1700

Data Index: 300735 (C4 (Web Text))

Max Activating Token Index: 612

Text #2

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.8892. Min Act: -0.1698

Data Index: 1236729 (C4 (Web Text))

Max Activating Token Index: 196

Text #3

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.7955. Min Act: -0.1700

Data Index: 723129 (C4 (Web Text))

Max Activating Token Index: 401

Text #4

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.7642. Min Act: -0.1700

Data Index: 989238 (C4 (Web Text))

Max Activating Token Index: 716

Text #5

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.7330. Min Act: -0.1700

Data Index: 1044178 (C4 (Web Text))

Max Activating Token Index: 288

Text #6

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.7017. Min Act: -0.1698

Data Index: 511616 (C4 (Web Text))

Max Activating Token Index: 467

Text #7

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.7330. Min Act: -0.1699

Data Index: 789972 (C4 (Web Text))

Max Activating Token Index: 57

Text #8

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.6080. Min Act: -0.1700

Data Index: 835208 (C4 (Web Text))

Max Activating Token Index: 568

Text #9

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.6080. Min Act: -0.1700

Data Index: 1626931 (Python Code)

Max Activating Token Index: 560

Text #10

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5767. Min Act: -0.1700

Data Index: 1454073 (Python Code)

Max Activating Token Index: 61

Text #11

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5767. Min Act: -0.1700

Data Index: 1515653 (Python Code)

Max Activating Token Index: 59

Text #12

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5455. Min Act: -0.1700

Data Index: 874978 (C4 (Web Text))

Max Activating Token Index: 504

Text #13

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5455. Min Act: -0.1700

Data Index: 1417178 (Python Code)

Max Activating Token Index: 35

Text #14

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5767. Min Act: -0.1699

Data Index: 1461987 (Python Code)

Max Activating Token Index: 119

Text #15

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5142. Min Act: -0.1700

Data Index: 1468529 (Python Code)

Max Activating Token Index: 158

Text #16

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5142. Min Act: -0.1700

Data Index: 1543722 (Python Code)

Max Activating Token Index: 268

Text #17

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.5142. Min Act: -0.1700

Data Index: 1565121 (Python Code)

Max Activating Token Index: 501

Text #18

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.4830. Min Act: -0.1699

Data Index: 210193 (C4 (Web Text))

Max Activating Token Index: 467

Text #19

Max Range: 5.6080. Min Range: -5.6080

Max Act: 4.4830. Min Act: -0.1700

Data Index: 1569272 (Python Code)

Max Activating Token Index: 145