Model: SoLU Model: 6 Layers, 3072 Neurons per Layer

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

Neuron 50 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-6l')

Text #0

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0830. Min Act: -0.0001

Data Index: 314273 (C4 (Web Text))

Max Activating Token Index: 612

Text #1

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0722. Min Act: -0.0001

Data Index: 1208657 (C4 (Web Text))

Max Activating Token Index: 141

Text #2

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0625. Min Act: -0.0001

Data Index: 793187 (C4 (Web Text))

Max Activating Token Index: 246

Text #3

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0647. Min Act: -0.0001

Data Index: 263865 (C4 (Web Text))

Max Activating Token Index: 345

Text #4

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0679. Min Act: -0.0001

Data Index: 410921 (C4 (Web Text))

Max Activating Token Index: 184

Text #5

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0627. Min Act: -0.0001

Data Index: 1067235 (C4 (Web Text))

Max Activating Token Index: 153

Text #6

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0668. Min Act: -0.0001

Data Index: 23071 (C4 (Web Text))

Max Activating Token Index: 730

Text #7

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0639. Min Act: -0.0001

Data Index: 373513 (C4 (Web Text))

Max Activating Token Index: 391

Text #8

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0611. Min Act: -0.0001

Data Index: 1607576 (Python Code)

Max Activating Token Index: 629

Text #9

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0633. Min Act: -0.0001

Data Index: 484733 (C4 (Web Text))

Max Activating Token Index: 830

Text #10

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0643. Min Act: -0.0001

Data Index: 433952 (C4 (Web Text))

Max Activating Token Index: 473

Text #11

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0667. Min Act: -0.0001

Data Index: 306620 (C4 (Web Text))

Max Activating Token Index: 700

Text #12

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0591. Min Act: -0.0001

Data Index: 1022097 (C4 (Web Text))

Max Activating Token Index: 381

Text #13

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0615. Min Act: -0.0001

Data Index: 487896 (C4 (Web Text))

Max Activating Token Index: 414

Text #14

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0526. Min Act: -0.0001

Data Index: 493401 (C4 (Web Text))

Max Activating Token Index: 846

Text #15

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0565. Min Act: -0.0001

Data Index: 1134714 (C4 (Web Text))

Max Activating Token Index: 543

Text #16

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0587. Min Act: -0.0001

Data Index: 412432 (C4 (Web Text))

Max Activating Token Index: 381

Text #17

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0584. Min Act: -0.0001

Data Index: 107655 (C4 (Web Text))

Max Activating Token Index: 999

Text #18

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0610. Min Act: -0.0001

Data Index: 1032252 (C4 (Web Text))

Max Activating Token Index: 129

Text #19

Max Range: 0.0830. Min Range: -0.0830

Max Act: 0.0577. Min Act: -0.0001

Data Index: 273826 (C4 (Web Text))

Max Activating Token Index: 318