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

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

Neuron 1403 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-3l')

Text #0

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0659. Min Act: -0.0003

Data Index: 1591532 (Python Code)

Max Activating Token Index: 735

Text #1

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0595. Min Act: -0.0003

Data Index: 1115450 (C4 (Web Text))

Max Activating Token Index: 952

Text #2

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0564. Min Act: -0.0003

Data Index: 500987 (C4 (Web Text))

Max Activating Token Index: 639

Text #3

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0569. Min Act: -0.0004

Data Index: 199852 (C4 (Web Text))

Max Activating Token Index: 783

Text #4

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0564. Min Act: -0.0003

Data Index: 603999 (C4 (Web Text))

Max Activating Token Index: 447

Text #5

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0553. Min Act: -0.0003

Data Index: 1594018 (Python Code)

Max Activating Token Index: 690

Text #6

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0542. Min Act: -0.0003

Data Index: 954123 (C4 (Web Text))

Max Activating Token Index: 489

Text #7

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0541. Min Act: -0.0003

Data Index: 1367107 (Python Code)

Max Activating Token Index: 975

Text #8

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0550. Min Act: -0.0003

Data Index: 1598507 (Python Code)

Max Activating Token Index: 602

Text #9

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0541. Min Act: -0.0003

Data Index: 600406 (C4 (Web Text))

Max Activating Token Index: 306

Text #10

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0543. Min Act: -0.0003

Data Index: 826366 (C4 (Web Text))

Max Activating Token Index: 275

Text #11

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0533. Min Act: -0.0003

Data Index: 1491412 (Python Code)

Max Activating Token Index: 975

Text #12

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0504. Min Act: -0.0003

Data Index: 1034984 (C4 (Web Text))

Max Activating Token Index: 944

Text #13

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0523. Min Act: -0.0003

Data Index: 477000 (C4 (Web Text))

Max Activating Token Index: 747

Text #14

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0543. Min Act: -0.0003

Data Index: 1060662 (C4 (Web Text))

Max Activating Token Index: 604

Text #15

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0500. Min Act: -0.0003

Data Index: 393616 (C4 (Web Text))

Max Activating Token Index: 722

Text #16

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0507. Min Act: -0.0002

Data Index: 565059 (C4 (Web Text))

Max Activating Token Index: 362

Text #17

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0504. Min Act: -0.0003

Data Index: 69291 (C4 (Web Text))

Max Activating Token Index: 958

Text #18

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0516. Min Act: -0.0003

Data Index: 1418740 (Python Code)

Max Activating Token Index: 279

Text #19

Max Range: 0.0659. Min Range: -0.0659

Max Act: 0.0501. Min Act: -0.0003

Data Index: 512149 (C4 (Web Text))

Max Activating Token Index: 775