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

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

Neuron 660 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-4l')

Text #0

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.3015. Min Act: -0.0002

Data Index: 522462 (C4 (Web Text))

Max Activating Token Index: 905

Text #1

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.2094. Min Act: -0.0002

Data Index: 1609144 (Python Code)

Max Activating Token Index: 951

Text #2

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.2061. Min Act: -0.0002

Data Index: 459925 (C4 (Web Text))

Max Activating Token Index: 825

Text #3

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.2034. Min Act: -0.0002

Data Index: 1214146 (C4 (Web Text))

Max Activating Token Index: 103

Text #4

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1671. Min Act: -0.0002

Data Index: 214466 (C4 (Web Text))

Max Activating Token Index: 447

Text #5

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1717. Min Act: -0.0002

Data Index: 1311078 (C4 (Web Text))

Max Activating Token Index: 910

Text #6

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1860. Min Act: -0.0002

Data Index: 36579 (C4 (Web Text))

Max Activating Token Index: 712

Text #7

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1793. Min Act: -0.0002

Data Index: 407899 (C4 (Web Text))

Max Activating Token Index: 925

Text #8

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1962. Min Act: -0.0002

Data Index: 506442 (C4 (Web Text))

Max Activating Token Index: 647

Text #9

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1843. Min Act: -0.0002

Data Index: 1624667 (Python Code)

Max Activating Token Index: 396

Text #10

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1791. Min Act: -0.0002

Data Index: 321775 (C4 (Web Text))

Max Activating Token Index: 385

Text #11

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1901. Min Act: -0.0002

Data Index: 492357 (C4 (Web Text))

Max Activating Token Index: 846

Text #12

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1817. Min Act: -0.0002

Data Index: 1486057 (Python Code)

Max Activating Token Index: 905

Text #13

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1733. Min Act: -0.0002

Data Index: 1046100 (C4 (Web Text))

Max Activating Token Index: 660

Text #14

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1772. Min Act: -0.0002

Data Index: 1628399 (Python Code)

Max Activating Token Index: 967

Text #15

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1717. Min Act: -0.0002

Data Index: 1389046 (Python Code)

Max Activating Token Index: 484

Text #16

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1726. Min Act: -0.0002

Data Index: 1524496 (Python Code)

Max Activating Token Index: 198

Text #17

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1700. Min Act: -0.0002

Data Index: 1307848 (C4 (Web Text))

Max Activating Token Index: 807

Text #18

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1800. Min Act: -0.0002

Data Index: 1277546 (C4 (Web Text))

Max Activating Token Index: 104

Text #19

Max Range: 0.3015. Min Range: -0.3015

Max Act: 0.1683. Min Act: -0.0002

Data Index: 686529 (C4 (Web Text))

Max Activating Token Index: 605