Model: SoLU Model: 8 Layers, 4096 Neurons per Layer

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

Neuron 3650 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-8l')

Text #0

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6311. Min Act: -0.0002

Data Index: 103159 (C4 (Web Text))

Max Activating Token Index: 45

Text #1

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6075. Min Act: -0.0002

Data Index: 351754 (C4 (Web Text))

Max Activating Token Index: 600

Text #2

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6729. Min Act: -0.0002

Data Index: 666571 (C4 (Web Text))

Max Activating Token Index: 808

Text #3

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6114. Min Act: -0.0002

Data Index: 633596 (C4 (Web Text))

Max Activating Token Index: 119

Text #4

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5691. Min Act: -0.0002

Data Index: 110040 (C4 (Web Text))

Max Activating Token Index: 918

Text #5

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6165. Min Act: -0.0002

Data Index: 544234 (C4 (Web Text))

Max Activating Token Index: 753

Text #6

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6176. Min Act: -0.0002

Data Index: 790183 (C4 (Web Text))

Max Activating Token Index: 519

Text #7

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.6273. Min Act: -0.0001

Data Index: 414190 (C4 (Web Text))

Max Activating Token Index: 68

Text #8

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5429. Min Act: -0.0002

Data Index: 530574 (C4 (Web Text))

Max Activating Token Index: 727

Text #9

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5687. Min Act: -0.0002

Data Index: 30127 (C4 (Web Text))

Max Activating Token Index: 577

Text #10

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5581. Min Act: -0.0002

Data Index: 471934 (C4 (Web Text))

Max Activating Token Index: 947

Text #11

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5493. Min Act: -0.0002

Data Index: 914997 (C4 (Web Text))

Max Activating Token Index: 875

Text #12

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5507. Min Act: -0.0002

Data Index: 1045229 (C4 (Web Text))

Max Activating Token Index: 985

Text #13

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5439. Min Act: -0.0001

Data Index: 1469420 (Python Code)

Max Activating Token Index: 8

Text #14

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5587. Min Act: -0.0001

Data Index: 1092963 (C4 (Web Text))

Max Activating Token Index: 288

Text #15

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5335. Min Act: -0.0001

Data Index: 807284 (C4 (Web Text))

Max Activating Token Index: 938

Text #16

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5359. Min Act: -0.0002

Data Index: 1017158 (C4 (Web Text))

Max Activating Token Index: 721

Text #17

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5662. Min Act: -0.0002

Data Index: 1292414 (C4 (Web Text))

Max Activating Token Index: 183

Text #18

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5293. Min Act: -0.0001

Data Index: 294987 (C4 (Web Text))

Max Activating Token Index: 15

Text #19

Max Range: 0.6729. Min Range: -0.6729

Max Act: 0.5554. Min Act: -0.0002

Data Index: 1153063 (C4 (Web Text))

Max Activating Token Index: 1022