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

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

Neuron 1745 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: 5.1563. Min Range: -5.1563

Max Act: 5.1563. Min Act: -0.1700

Data Index: 1302697 (C4 (Web Text))

Max Activating Token Index: 1023

Text #1

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.6250. Min Act: -0.1700

Data Index: 867696 (C4 (Web Text))

Max Activating Token Index: 596

Text #2

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.5156. Min Act: -0.1700

Data Index: 1289789 (C4 (Web Text))

Max Activating Token Index: 401

Text #3

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.5781. Min Act: -0.1700

Data Index: 876832 (C4 (Web Text))

Max Activating Token Index: 462

Text #4

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.5781. Min Act: -0.1700

Data Index: 702507 (C4 (Web Text))

Max Activating Token Index: 499

Text #5

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.5313. Min Act: -0.1700

Data Index: 1038335 (C4 (Web Text))

Max Activating Token Index: 795

Text #6

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.4844. Min Act: -0.1700

Data Index: 1267666 (C4 (Web Text))

Max Activating Token Index: 814

Text #7

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3750. Min Act: -0.1700

Data Index: 1400396 (Python Code)

Max Activating Token Index: 584

Text #8

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3437. Min Act: -0.1700

Data Index: 994280 (C4 (Web Text))

Max Activating Token Index: 762

Text #9

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3750. Min Act: -0.1700

Data Index: 567820 (C4 (Web Text))

Max Activating Token Index: 657

Text #10

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3750. Min Act: -0.1700

Data Index: 422172 (C4 (Web Text))

Max Activating Token Index: 794

Text #11

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3281. Min Act: -0.1700

Data Index: 1557474 (Python Code)

Max Activating Token Index: 358

Text #12

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.2812. Min Act: -0.1700

Data Index: 59029 (C4 (Web Text))

Max Activating Token Index: 634

Text #13

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3437. Min Act: -0.1700

Data Index: 515033 (C4 (Web Text))

Max Activating Token Index: 118

Text #14

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3437. Min Act: -0.1700

Data Index: 1502316 (Python Code)

Max Activating Token Index: 173

Text #15

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3594. Min Act: -0.1700

Data Index: 74356 (C4 (Web Text))

Max Activating Token Index: 40

Text #16

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.2969. Min Act: -0.1700

Data Index: 752877 (C4 (Web Text))

Max Activating Token Index: 902

Text #17

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.2969. Min Act: -0.1700

Data Index: 646780 (C4 (Web Text))

Max Activating Token Index: 14

Text #18

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.3125. Min Act: -0.1700

Data Index: 1463816 (Python Code)

Max Activating Token Index: 434

Text #19

Max Range: 5.1563. Min Range: -5.1563

Max Act: 4.1875. Min Act: -0.1700

Data Index: 624437 (C4 (Web Text))

Max Activating Token Index: 537