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

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

Neuron 437 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: 5.4432. Min Range: -5.4432

Max Act: 5.4432. Min Act: -0.1690

Data Index: 164119 (C4 (Web Text))

Max Activating Token Index: 586

Text #1

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.1619. Min Act: -0.1700

Data Index: 661466 (C4 (Web Text))

Max Activating Token Index: 354

Text #2

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.0994. Min Act: -0.1700

Data Index: 718161 (C4 (Web Text))

Max Activating Token Index: 499

Text #3

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.0994. Min Act: -0.1700

Data Index: 788600 (C4 (Web Text))

Max Activating Token Index: 500

Text #4

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.0369. Min Act: -0.1700

Data Index: 117659 (C4 (Web Text))

Max Activating Token Index: 85

Text #5

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.0369. Min Act: -0.1700

Data Index: 707675 (C4 (Web Text))

Max Activating Token Index: 592

Text #6

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.0369. Min Act: -0.1700

Data Index: 595103 (C4 (Web Text))

Max Activating Token Index: 386

Text #7

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.9744. Min Act: -0.1697

Data Index: 1641061 (Python Code)

Max Activating Token Index: 336

Text #8

Max Range: 5.4432. Min Range: -5.4432

Max Act: 5.0369. Min Act: -0.1700

Data Index: 1038775 (C4 (Web Text))

Max Activating Token Index: 1020

Text #9

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.9744. Min Act: -0.1700

Data Index: 90835 (C4 (Web Text))

Max Activating Token Index: 793

Text #10

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.9744. Min Act: -0.1700

Data Index: 227189 (C4 (Web Text))

Max Activating Token Index: 428

Text #11

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.9432. Min Act: -0.1700

Data Index: 1375550 (Python Code)

Max Activating Token Index: 9

Text #12

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.9119. Min Act: -0.1700

Data Index: 1644743 (Python Code)

Max Activating Token Index: 625

Text #13

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8807. Min Act: -0.1700

Data Index: 1628790 (Python Code)

Max Activating Token Index: 49

Text #14

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8494. Min Act: -0.1700

Data Index: 1568821 (Python Code)

Max Activating Token Index: 68

Text #15

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8494. Min Act: -0.1700

Data Index: 167119 (C4 (Web Text))

Max Activating Token Index: 991

Text #16

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8494. Min Act: -0.1700

Data Index: 860809 (C4 (Web Text))

Max Activating Token Index: 143

Text #17

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8807. Min Act: -0.1674

Data Index: 1239363 (C4 (Web Text))

Max Activating Token Index: 828

Text #18

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8807. Min Act: -0.1700

Data Index: 56384 (C4 (Web Text))

Max Activating Token Index: 699

Text #19

Max Range: 5.4432. Min Range: -5.4432

Max Act: 4.8494. Min Act: -0.1700

Data Index: 1284387 (C4 (Web Text))

Max Activating Token Index: 966