Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

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

Neuron 3508 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9649. Min Act: -0.0001

Data Index: 1224805 (C4 (Web Text))

Max Activating Token Index: 807

Text #1

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.0614. Min Act: -0.0001

Data Index: 1207378 (C4 (Web Text))

Max Activating Token Index: 914

Text #2

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.1697. Min Act: -0.0001

Data Index: 1001392 (C4 (Web Text))

Max Activating Token Index: 443

Text #3

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.0895. Min Act: -0.0000

Data Index: 1455230 (Python Code)

Max Activating Token Index: 754

Text #4

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.1192. Min Act: -0.0000

Data Index: 1578802 (Python Code)

Max Activating Token Index: 867

Text #5

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.1134. Min Act: -0.0001

Data Index: 1575302 (Python Code)

Max Activating Token Index: 412

Text #6

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.0681. Min Act: -0.0001

Data Index: 9041 (C4 (Web Text))

Max Activating Token Index: 677

Text #7

Max Range: 1.1697. Min Range: -1.1697

Max Act: 1.0367. Min Act: -0.0001

Data Index: 1424281 (Python Code)

Max Activating Token Index: 698

Text #8

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9771. Min Act: -0.0001

Data Index: 370028 (C4 (Web Text))

Max Activating Token Index: 428

Text #9

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9329. Min Act: -0.0000

Data Index: 516046 (C4 (Web Text))

Max Activating Token Index: 174

Text #10

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9104. Min Act: -0.0001

Data Index: 210279 (C4 (Web Text))

Max Activating Token Index: 286

Text #11

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9320. Min Act: -0.0001

Data Index: 1019401 (C4 (Web Text))

Max Activating Token Index: 259

Text #12

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9236. Min Act: -0.0001

Data Index: 1464298 (Python Code)

Max Activating Token Index: 486

Text #13

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9528. Min Act: -0.0001

Data Index: 722069 (C4 (Web Text))

Max Activating Token Index: 736

Text #14

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9960. Min Act: -0.0000

Data Index: 553001 (C4 (Web Text))

Max Activating Token Index: 710

Text #15

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.9424. Min Act: -0.0001

Data Index: 1568407 (Python Code)

Max Activating Token Index: 457

Text #16

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.8837. Min Act: -0.0001

Data Index: 1093110 (C4 (Web Text))

Max Activating Token Index: 62

Text #17

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.8868. Min Act: -0.0000

Data Index: 1539789 (Python Code)

Max Activating Token Index: 609

Text #18

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.8490. Min Act: -0.0001

Data Index: 541264 (C4 (Web Text))

Max Activating Token Index: 99

Text #19

Max Range: 1.1697. Min Range: -1.1697

Max Act: 0.7696. Min Act: -0.0001

Data Index: 29387 (C4 (Web Text))

Max Activating Token Index: 203