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

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

Neuron 2027 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: 5.1818. Min Range: -5.1818

Max Act: 5.1505. Min Act: -0.1700

Data Index: 401290 (C4 (Web Text))

Max Activating Token Index: 452

Text #1

Max Range: 5.1818. Min Range: -5.1818

Max Act: 5.1818. Min Act: -0.1700

Data Index: 1288628 (C4 (Web Text))

Max Activating Token Index: 458

Text #2

Max Range: 5.1818. Min Range: -5.1818

Max Act: 5.0568. Min Act: -0.1700

Data Index: 650519 (C4 (Web Text))

Max Activating Token Index: 743

Text #3

Max Range: 5.1818. Min Range: -5.1818

Max Act: 5.0568. Min Act: -0.1699

Data Index: 1106000 (C4 (Web Text))

Max Activating Token Index: 663

Text #4

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.9943. Min Act: -0.1700

Data Index: 556896 (C4 (Web Text))

Max Activating Token Index: 976

Text #5

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.9005. Min Act: -0.1700

Data Index: 86649 (C4 (Web Text))

Max Activating Token Index: 634

Text #6

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.8380. Min Act: -0.1700

Data Index: 605938 (C4 (Web Text))

Max Activating Token Index: 74

Text #7

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.8068. Min Act: -0.1700

Data Index: 567468 (C4 (Web Text))

Max Activating Token Index: 686

Text #8

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7755. Min Act: -0.1700

Data Index: 1502213 (Python Code)

Max Activating Token Index: 997

Text #9

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.8380. Min Act: -0.1700

Data Index: 638868 (C4 (Web Text))

Max Activating Token Index: 176

Text #10

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7755. Min Act: -0.1700

Data Index: 980009 (C4 (Web Text))

Max Activating Token Index: 940

Text #11

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.8068. Min Act: -0.1700

Data Index: 240099 (C4 (Web Text))

Max Activating Token Index: 920

Text #12

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7755. Min Act: -0.1700

Data Index: 562358 (C4 (Web Text))

Max Activating Token Index: 959

Text #13

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7755. Min Act: -0.1700

Data Index: 708585 (C4 (Web Text))

Max Activating Token Index: 52

Text #14

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7130. Min Act: -0.1700

Data Index: 1650112 (Python Code)

Max Activating Token Index: 1005

Text #15

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7755. Min Act: -0.1700

Data Index: 805843 (C4 (Web Text))

Max Activating Token Index: 460

Text #16

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7443. Min Act: -0.1699

Data Index: 1011489 (C4 (Web Text))

Max Activating Token Index: 881

Text #17

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7755. Min Act: -0.1700

Data Index: 1181883 (C4 (Web Text))

Max Activating Token Index: 967

Text #18

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.7443. Min Act: -0.1700

Data Index: 763389 (C4 (Web Text))

Max Activating Token Index: 536

Text #19

Max Range: 5.1818. Min Range: -5.1818

Max Act: 4.8068. Min Act: -0.1700

Data Index: 1201300 (C4 (Web Text))

Max Activating Token Index: 723