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

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

Neuron 496 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: 0.0600. Min Range: -0.0600

Max Act: 0.0563. Min Act: -0.0002

Data Index: 1464494 (Python Code)

Max Activating Token Index: 641

Text #1

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0600. Min Act: -0.0002

Data Index: 727855 (C4 (Web Text))

Max Activating Token Index: 214

Text #2

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0551. Min Act: -0.0002

Data Index: 289053 (C4 (Web Text))

Max Activating Token Index: 837

Text #3

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0591. Min Act: -0.0002

Data Index: 617698 (C4 (Web Text))

Max Activating Token Index: 1010

Text #4

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0558. Min Act: -0.0002

Data Index: 420217 (C4 (Web Text))

Max Activating Token Index: 18

Text #5

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0574. Min Act: -0.0002

Data Index: 1482634 (Python Code)

Max Activating Token Index: 720

Text #6

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0540. Min Act: -0.0002

Data Index: 1163507 (C4 (Web Text))

Max Activating Token Index: 762

Text #7

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0562. Min Act: -0.0002

Data Index: 909599 (C4 (Web Text))

Max Activating Token Index: 375

Text #8

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0524. Min Act: -0.0002

Data Index: 1069575 (C4 (Web Text))

Max Activating Token Index: 595

Text #9

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0543. Min Act: -0.0002

Data Index: 565167 (C4 (Web Text))

Max Activating Token Index: 398

Text #10

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0532. Min Act: -0.0001

Data Index: 17927 (C4 (Web Text))

Max Activating Token Index: 971

Text #11

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0522. Min Act: -0.0002

Data Index: 318596 (C4 (Web Text))

Max Activating Token Index: 278

Text #12

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0538. Min Act: -0.0003

Data Index: 1425662 (Python Code)

Max Activating Token Index: 928

Text #13

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0545. Min Act: -0.0002

Data Index: 601487 (C4 (Web Text))

Max Activating Token Index: 344

Text #14

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0528. Min Act: -0.0002

Data Index: 1465934 (Python Code)

Max Activating Token Index: 335

Text #15

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0529. Min Act: -0.0000

Data Index: 934012 (C4 (Web Text))

Max Activating Token Index: 814

Text #16

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0515. Min Act: -0.0002

Data Index: 632192 (C4 (Web Text))

Max Activating Token Index: 339

Text #17

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0517. Min Act: -0.0002

Data Index: 1448840 (Python Code)

Max Activating Token Index: 629

Text #18

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0460. Min Act: -0.0002

Data Index: 152893 (C4 (Web Text))

Max Activating Token Index: 493

Text #19

Max Range: 0.0600. Min Range: -0.0600

Max Act: 0.0504. Min Act: -0.0002

Data Index: 1375169 (Python Code)

Max Activating Token Index: 865