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

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

Neuron 1980 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: 5.1598. Min Range: -5.1598

Max Act: 5.1598. Min Act: -0.1700

Data Index: 1543845 (Python Code)

Max Activating Token Index: 929

Text #1

Max Range: 5.1598. Min Range: -5.1598

Max Act: 5.0348. Min Act: -0.1700

Data Index: 381910 (C4 (Web Text))

Max Activating Token Index: 582

Text #2

Max Range: 5.1598. Min Range: -5.1598

Max Act: 5.0035. Min Act: -0.1700

Data Index: 440768 (C4 (Web Text))

Max Activating Token Index: 733

Text #3

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.9410. Min Act: -0.1700

Data Index: 633731 (C4 (Web Text))

Max Activating Token Index: 281

Text #4

Max Range: 5.1598. Min Range: -5.1598

Max Act: 5.0035. Min Act: -0.1700

Data Index: 456317 (C4 (Web Text))

Max Activating Token Index: 72

Text #5

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.9410. Min Act: -0.1700

Data Index: 1569478 (Python Code)

Max Activating Token Index: 645

Text #6

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8785. Min Act: -0.1700

Data Index: 1543242 (Python Code)

Max Activating Token Index: 198

Text #7

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8785. Min Act: -0.1700

Data Index: 54394 (C4 (Web Text))

Max Activating Token Index: 712

Text #8

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8785. Min Act: -0.1700

Data Index: 1555862 (Python Code)

Max Activating Token Index: 20

Text #9

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8473. Min Act: -0.1700

Data Index: 1627468 (Python Code)

Max Activating Token Index: 853

Text #10

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.9098. Min Act: -0.1700

Data Index: 111881 (C4 (Web Text))

Max Activating Token Index: 1016

Text #11

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8473. Min Act: -0.1700

Data Index: 1383817 (Python Code)

Max Activating Token Index: 24

Text #12

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8473. Min Act: -0.1700

Data Index: 670192 (C4 (Web Text))

Max Activating Token Index: 728

Text #13

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8160. Min Act: -0.1700

Data Index: 48748 (C4 (Web Text))

Max Activating Token Index: 220

Text #14

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8473. Min Act: -0.1700

Data Index: 598587 (C4 (Web Text))

Max Activating Token Index: 808

Text #15

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.7848. Min Act: -0.1700

Data Index: 553613 (C4 (Web Text))

Max Activating Token Index: 846

Text #16

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.7848. Min Act: -0.1700

Data Index: 1410498 (Python Code)

Max Activating Token Index: 732

Text #17

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.7223. Min Act: -0.1700

Data Index: 688251 (C4 (Web Text))

Max Activating Token Index: 818

Text #18

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.7848. Min Act: -0.1700

Data Index: 1135659 (C4 (Web Text))

Max Activating Token Index: 401

Text #19

Max Range: 5.1598. Min Range: -5.1598

Max Act: 4.8160. Min Act: -0.1700

Data Index: 558962 (C4 (Web Text))

Max Activating Token Index: 1021