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

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

Neuron 2016 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')

Text #0

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.6798. Min Act: -0.1700

Data Index: 1302524 (C4 (Web Text))

Max Activating Token Index: 978

Text #1

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.4610. Min Act: -0.1700

Data Index: 1187585 (C4 (Web Text))

Max Activating Token Index: 1016

Text #2

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.2735. Min Act: -0.1700

Data Index: 126173 (C4 (Web Text))

Max Activating Token Index: 923

Text #3

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.3360. Min Act: -0.1700

Data Index: 1442218 (Python Code)

Max Activating Token Index: 704

Text #4

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.3673. Min Act: -0.1700

Data Index: 516605 (C4 (Web Text))

Max Activating Token Index: 897

Text #5

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.2735. Min Act: -0.1700

Data Index: 900572 (C4 (Web Text))

Max Activating Token Index: 231

Text #6

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.2423. Min Act: -0.1700

Data Index: 1216232 (C4 (Web Text))

Max Activating Token Index: 864

Text #7

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.2735. Min Act: -0.1700

Data Index: 1613698 (Python Code)

Max Activating Token Index: 878

Text #8

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.2110. Min Act: -0.1700

Data Index: 1023228 (C4 (Web Text))

Max Activating Token Index: 1001

Text #9

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.1485. Min Act: -0.1700

Data Index: 1413920 (Python Code)

Max Activating Token Index: 798

Text #10

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.2110. Min Act: -0.1700

Data Index: 1544528 (Python Code)

Max Activating Token Index: 891

Text #11

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.1798. Min Act: -0.1700

Data Index: 1024752 (C4 (Web Text))

Max Activating Token Index: 728

Text #12

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.1173. Min Act: -0.1700

Data Index: 1369394 (Python Code)

Max Activating Token Index: 211

Text #13

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.1173. Min Act: -0.1700

Data Index: 1522348 (Python Code)

Max Activating Token Index: 532

Text #14

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.0235. Min Act: -0.1700

Data Index: 801746 (C4 (Web Text))

Max Activating Token Index: 499

Text #15

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.0235. Min Act: -0.1700

Data Index: 1623659 (Python Code)

Max Activating Token Index: 611

Text #16

Max Range: 5.6798. Min Range: -5.6798

Max Act: 4.9923. Min Act: -0.1700

Data Index: 1485844 (Python Code)

Max Activating Token Index: 937

Text #17

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.0548. Min Act: -0.1700

Data Index: 1522589 (Python Code)

Max Activating Token Index: 159

Text #18

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.0860. Min Act: -0.1700

Data Index: 1448671 (Python Code)

Max Activating Token Index: 190

Text #19

Max Range: 5.6798. Min Range: -5.6798

Max Act: 5.1173. Min Act: -0.1700

Data Index: 60054 (C4 (Web Text))

Max Activating Token Index: 950