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

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

Neuron 1591 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: 5.2872. Min Range: -5.2872

Max Act: 5.2872. Min Act: -0.1700

Data Index: 883716 (C4 (Web Text))

Max Activating Token Index: 744

Text #1

Max Range: 5.2872. Min Range: -5.2872

Max Act: 5.1700. Min Act: -0.1700

Data Index: 1654364 (Python Code)

Max Activating Token Index: 484

Text #2

Max Range: 5.2872. Min Range: -5.2872

Max Act: 5.2013. Min Act: -0.1700

Data Index: 1203695 (C4 (Web Text))

Max Activating Token Index: 315

Text #3

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.9044. Min Act: -0.1700

Data Index: 357325 (C4 (Web Text))

Max Activating Token Index: 651

Text #4

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8888. Min Act: -0.1700

Data Index: 1606814 (Python Code)

Max Activating Token Index: 531

Text #5

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8106. Min Act: -0.1700

Data Index: 82362 (C4 (Web Text))

Max Activating Token Index: 819

Text #6

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8809. Min Act: -0.1700

Data Index: 238453 (C4 (Web Text))

Max Activating Token Index: 850

Text #7

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.9200. Min Act: -0.1700

Data Index: 210272 (C4 (Web Text))

Max Activating Token Index: 458

Text #8

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8731. Min Act: -0.1700

Data Index: 1042609 (C4 (Web Text))

Max Activating Token Index: 832

Text #9

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.7794. Min Act: -0.1700

Data Index: 1346870 (C4 (Web Text))

Max Activating Token Index: 280

Text #10

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.7638. Min Act: -0.1700

Data Index: 1633848 (Python Code)

Max Activating Token Index: 210

Text #11

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8184. Min Act: -0.1700

Data Index: 1424533 (Python Code)

Max Activating Token Index: 910

Text #12

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.7638. Min Act: -0.1700

Data Index: 1003762 (C4 (Web Text))

Max Activating Token Index: 945

Text #13

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8106. Min Act: -0.1700

Data Index: 660127 (C4 (Web Text))

Max Activating Token Index: 947

Text #14

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.8341. Min Act: -0.1700

Data Index: 228156 (C4 (Web Text))

Max Activating Token Index: 750

Text #15

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.7403. Min Act: -0.1700

Data Index: 1654426 (Python Code)

Max Activating Token Index: 740

Text #16

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.6934. Min Act: -0.1700

Data Index: 1568399 (Python Code)

Max Activating Token Index: 820

Text #17

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.6856. Min Act: -0.1700

Data Index: 1197970 (C4 (Web Text))

Max Activating Token Index: 374

Text #18

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.6778. Min Act: -0.1700

Data Index: 315854 (C4 (Web Text))

Max Activating Token Index: 846

Text #19

Max Range: 5.2872. Min Range: -5.2872

Max Act: 4.6075. Min Act: -0.1700

Data Index: 317729 (C4 (Web Text))

Max Activating Token Index: 290