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

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

Neuron 2012 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.1024. Min Act: -0.1700

Data Index: 473528 (C4 (Web Text))

Max Activating Token Index: 994

Text #1

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.0711. Min Act: -0.1700

Data Index: 1524083 (Python Code)

Max Activating Token Index: 335

Text #2

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.0399. Min Act: -0.1700

Data Index: 1157265 (C4 (Web Text))

Max Activating Token Index: 875

Text #3

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.0086. Min Act: -0.1700

Data Index: 490931 (C4 (Web Text))

Max Activating Token Index: 917

Text #4

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.0086. Min Act: -0.1700

Data Index: 1259101 (C4 (Web Text))

Max Activating Token Index: 742

Text #5

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.0399. Min Act: -0.1700

Data Index: 714737 (C4 (Web Text))

Max Activating Token Index: 128

Text #6

Max Range: 4.1024. Min Range: -4.1024

Max Act: 4.0086. Min Act: -0.1700

Data Index: 314034 (C4 (Web Text))

Max Activating Token Index: 184

Text #7

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9773. Min Act: -0.1700

Data Index: 1652527 (Python Code)

Max Activating Token Index: 775

Text #8

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9773. Min Act: -0.1700

Data Index: 367647 (C4 (Web Text))

Max Activating Token Index: 908

Text #9

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9461. Min Act: -0.1700

Data Index: 210630 (C4 (Web Text))

Max Activating Token Index: 424

Text #10

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9148. Min Act: -0.1700

Data Index: 375285 (C4 (Web Text))

Max Activating Token Index: 860

Text #11

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9148. Min Act: -0.1700

Data Index: 1408176 (Python Code)

Max Activating Token Index: 5

Text #12

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9148. Min Act: -0.1700

Data Index: 1570017 (Python Code)

Max Activating Token Index: 28

Text #13

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.9148. Min Act: -0.1700

Data Index: 547554 (C4 (Web Text))

Max Activating Token Index: 208

Text #14

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.8835. Min Act: -0.1700

Data Index: 1406700 (Python Code)

Max Activating Token Index: 5

Text #15

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.8523. Min Act: -0.1700

Data Index: 500219 (C4 (Web Text))

Max Activating Token Index: 461

Text #16

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.8523. Min Act: -0.1700

Data Index: 1625470 (Python Code)

Max Activating Token Index: 107

Text #17

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.8523. Min Act: -0.1700

Data Index: 1551535 (Python Code)

Max Activating Token Index: 254

Text #18

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.8523. Min Act: -0.1700

Data Index: 1568762 (Python Code)

Max Activating Token Index: 5

Text #19

Max Range: 4.1024. Min Range: -4.1024

Max Act: 3.8210. Min Act: -0.1700

Data Index: 1583432 (Python Code)

Max Activating Token Index: 4