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

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

Neuron 2012 in Layer 0

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.8149. Min Act: -0.1700

Data Index: 1100233 (C4 (Web Text))

Max Activating Token Index: 698

Text #1

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.7523. Min Act: -0.1700

Data Index: 1127125 (C4 (Web Text))

Max Activating Token Index: 833

Text #2

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.7836. Min Act: -0.1700

Data Index: 959429 (C4 (Web Text))

Max Activating Token Index: 221

Text #3

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.7523. Min Act: -0.1700

Data Index: 486924 (C4 (Web Text))

Max Activating Token Index: 264

Text #4

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.6897. Min Act: -0.1700

Data Index: 1081645 (C4 (Web Text))

Max Activating Token Index: 496

Text #5

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.6584. Min Act: -0.1700

Data Index: 1195691 (C4 (Web Text))

Max Activating Token Index: 493

Text #6

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.6271. Min Act: -0.1700

Data Index: 310663 (C4 (Web Text))

Max Activating Token Index: 432

Text #7

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.5958. Min Act: -0.1700

Data Index: 624046 (C4 (Web Text))

Max Activating Token Index: 200

Text #8

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.5018. Min Act: -0.1700

Data Index: 634700 (C4 (Web Text))

Max Activating Token Index: 188

Text #9

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.4705. Min Act: -0.1700

Data Index: 204303 (C4 (Web Text))

Max Activating Token Index: 40

Text #10

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.4391. Min Act: -0.1700

Data Index: 1525571 (Python Code)

Max Activating Token Index: 823

Text #11

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.4234. Min Act: -0.1700

Data Index: 1641106 (Python Code)

Max Activating Token Index: 316

Text #12

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.3764. Min Act: -0.1700

Data Index: 1393228 (Python Code)

Max Activating Token Index: 7

Text #13

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.3607. Min Act: -0.1700

Data Index: 1018725 (C4 (Web Text))

Max Activating Token Index: 228

Text #14

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.3764. Min Act: -0.1700

Data Index: 1131332 (C4 (Web Text))

Max Activating Token Index: 187

Text #15

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.3450. Min Act: -0.1700

Data Index: 1193822 (C4 (Web Text))

Max Activating Token Index: 730

Text #16

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.3607. Min Act: -0.1700

Data Index: 1513610 (Python Code)

Max Activating Token Index: 637

Text #17

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.3450. Min Act: -0.1700

Data Index: 116724 (C4 (Web Text))

Max Activating Token Index: 94

Text #18

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.2979. Min Act: -0.1700

Data Index: 1648354 (Python Code)

Max Activating Token Index: 966

Text #19

Max Range: 3.8149. Min Range: -3.8149

Max Act: 3.2979. Min Act: -0.1700

Data Index: 1567113 (Python Code)

Max Activating Token Index: 179