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

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

Neuron 75 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: 0.0691. Min Range: -0.0691

Max Act: 0.0691. Min Act: -0.0002

Data Index: 344940 (C4 (Web Text))

Max Activating Token Index: 192

Text #1

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0645. Min Act: -0.0002

Data Index: 790814 (C4 (Web Text))

Max Activating Token Index: 77

Text #2

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0594. Min Act: -0.0002

Data Index: 643417 (C4 (Web Text))

Max Activating Token Index: 263

Text #3

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0579. Min Act: -0.0002

Data Index: 1224330 (C4 (Web Text))

Max Activating Token Index: 565

Text #4

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0579. Min Act: -0.0002

Data Index: 116520 (C4 (Web Text))

Max Activating Token Index: 886

Text #5

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0583. Min Act: -0.0002

Data Index: 822718 (C4 (Web Text))

Max Activating Token Index: 527

Text #6

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0567. Min Act: -0.0002

Data Index: 378109 (C4 (Web Text))

Max Activating Token Index: 935

Text #7

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0580. Min Act: -0.0002

Data Index: 656243 (C4 (Web Text))

Max Activating Token Index: 91

Text #8

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0585. Min Act: -0.0002

Data Index: 1096660 (C4 (Web Text))

Max Activating Token Index: 806

Text #9

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0575. Min Act: -0.0002

Data Index: 430001 (C4 (Web Text))

Max Activating Token Index: 842

Text #10

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0572. Min Act: -0.0002

Data Index: 1262733 (C4 (Web Text))

Max Activating Token Index: 844

Text #11

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0521. Min Act: -0.0002

Data Index: 1558598 (Python Code)

Max Activating Token Index: 882

Text #12

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0564. Min Act: -0.0002

Data Index: 700410 (C4 (Web Text))

Max Activating Token Index: 365

Text #13

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0522. Min Act: -0.0002

Data Index: 551624 (C4 (Web Text))

Max Activating Token Index: 438

Text #14

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0556. Min Act: -0.0002

Data Index: 1305809 (C4 (Web Text))

Max Activating Token Index: 111

Text #15

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0576. Min Act: -0.0002

Data Index: 103094 (C4 (Web Text))

Max Activating Token Index: 85

Text #16

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0566. Min Act: -0.0002

Data Index: 453017 (C4 (Web Text))

Max Activating Token Index: 410

Text #17

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0571. Min Act: -0.0002

Data Index: 207041 (C4 (Web Text))

Max Activating Token Index: 120

Text #18

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0565. Min Act: -0.0002

Data Index: 622246 (C4 (Web Text))

Max Activating Token Index: 398

Text #19

Max Range: 0.0691. Min Range: -0.0691

Max Act: 0.0540. Min Act: -0.0002

Data Index: 1057649 (C4 (Web Text))

Max Activating Token Index: 195