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

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

Neuron 150 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: 0.0548. Min Range: -0.0548

Max Act: 0.0548. Min Act: -0.0001

Data Index: 483507 (C4 (Web Text))

Max Activating Token Index: 484

Text #1

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0520. Min Act: -0.0001

Data Index: 886374 (C4 (Web Text))

Max Activating Token Index: 159

Text #2

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0489. Min Act: -0.0001

Data Index: 362044 (C4 (Web Text))

Max Activating Token Index: 982

Text #3

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0497. Min Act: -0.0001

Data Index: 103979 (C4 (Web Text))

Max Activating Token Index: 857

Text #4

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0456. Min Act: -0.0001

Data Index: 663557 (C4 (Web Text))

Max Activating Token Index: 555

Text #5

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0400. Min Act: -0.0001

Data Index: 1514711 (Python Code)

Max Activating Token Index: 689

Text #6

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0421. Min Act: 0.0000

Data Index: 972677 (C4 (Web Text))

Max Activating Token Index: 877

Text #7

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0360. Min Act: -0.0002

Data Index: 54681 (C4 (Web Text))

Max Activating Token Index: 953

Text #8

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0405. Min Act: -0.0001

Data Index: 1344572 (C4 (Web Text))

Max Activating Token Index: 901

Text #9

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0413. Min Act: -0.0002

Data Index: 1592319 (Python Code)

Max Activating Token Index: 344

Text #10

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0389. Min Act: -0.0002

Data Index: 1608509 (Python Code)

Max Activating Token Index: 656

Text #11

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0379. Min Act: -0.0000

Data Index: 379560 (C4 (Web Text))

Max Activating Token Index: 562

Text #12

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0383. Min Act: -0.0001

Data Index: 1000132 (C4 (Web Text))

Max Activating Token Index: 820

Text #13

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0373. Min Act: -0.0001

Data Index: 505105 (C4 (Web Text))

Max Activating Token Index: 593

Text #14

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0378. Min Act: -0.0002

Data Index: 1583085 (Python Code)

Max Activating Token Index: 434

Text #15

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0369. Min Act: -0.0001

Data Index: 237649 (C4 (Web Text))

Max Activating Token Index: 746

Text #16

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0367. Min Act: -0.0001

Data Index: 643284 (C4 (Web Text))

Max Activating Token Index: 411

Text #17

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0371. Min Act: -0.0002

Data Index: 735954 (C4 (Web Text))

Max Activating Token Index: 715

Text #18

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0381. Min Act: -0.0002

Data Index: 1564993 (Python Code)

Max Activating Token Index: 713

Text #19

Max Range: 0.0548. Min Range: -0.0548

Max Act: 0.0370. Min Act: -0.0002

Data Index: 1539207 (Python Code)

Max Activating Token Index: 462