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

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

Neuron 1863 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: 4.0169. Min Range: -4.0169

Max Act: 4.0169. Min Act: -0.1700

Data Index: 564055 (C4 (Web Text))

Max Activating Token Index: 529

Text #1

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.9856. Min Act: -0.1700

Data Index: 1532209 (Python Code)

Max Activating Token Index: 37

Text #2

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8762. Min Act: -0.1700

Data Index: 1213864 (C4 (Web Text))

Max Activating Token Index: 569

Text #3

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8762. Min Act: -0.1700

Data Index: 1208391 (C4 (Web Text))

Max Activating Token Index: 954

Text #4

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8605. Min Act: -0.1700

Data Index: 401562 (C4 (Web Text))

Max Activating Token Index: 141

Text #5

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8918. Min Act: -0.1700

Data Index: 897605 (C4 (Web Text))

Max Activating Token Index: 658

Text #6

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8918. Min Act: -0.1700

Data Index: 1177381 (C4 (Web Text))

Max Activating Token Index: 351

Text #7

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8605. Min Act: -0.1700

Data Index: 794460 (C4 (Web Text))

Max Activating Token Index: 802

Text #8

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8293. Min Act: -0.1700

Data Index: 616262 (C4 (Web Text))

Max Activating Token Index: 196

Text #9

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8136. Min Act: -0.1699

Data Index: 1551192 (Python Code)

Max Activating Token Index: 769

Text #10

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.7980. Min Act: -0.1700

Data Index: 1039705 (C4 (Web Text))

Max Activating Token Index: 714

Text #11

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8136. Min Act: -0.1700

Data Index: 1366887 (Python Code)

Max Activating Token Index: 6

Text #12

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8136. Min Act: -0.1700

Data Index: 1190381 (C4 (Web Text))

Max Activating Token Index: 12

Text #13

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8449. Min Act: -0.1700

Data Index: 1479600 (Python Code)

Max Activating Token Index: 219

Text #14

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.7667. Min Act: -0.1700

Data Index: 719973 (C4 (Web Text))

Max Activating Token Index: 899

Text #15

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.8449. Min Act: -0.1700

Data Index: 1316614 (C4 (Web Text))

Max Activating Token Index: 702

Text #16

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.7667. Min Act: -0.1700

Data Index: 1401436 (Python Code)

Max Activating Token Index: 905

Text #17

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.7667. Min Act: -0.1700

Data Index: 1426339 (Python Code)

Max Activating Token Index: 18

Text #18

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.7354. Min Act: -0.1700

Data Index: 912065 (C4 (Web Text))

Max Activating Token Index: 390

Text #19

Max Range: 4.0169. Min Range: -4.0169

Max Act: 3.7041. Min Act: -0.1700

Data Index: 1272594 (C4 (Web Text))

Max Activating Token Index: 784