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

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

Neuron 3834 in Layer 5

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.5489. Min Act: -0.0001

Data Index: 216247 (C4 (Web Text))

Max Activating Token Index: 7

Text #1

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.2543. Min Act: -0.0001

Data Index: 263041 (C4 (Web Text))

Max Activating Token Index: 315

Text #2

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.2423. Min Act: -0.0001

Data Index: 785143 (C4 (Web Text))

Max Activating Token Index: 829

Text #3

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.2177. Min Act: -0.0001

Data Index: 734930 (C4 (Web Text))

Max Activating Token Index: 732

Text #4

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.2473. Min Act: -0.0001

Data Index: 1647932 (Python Code)

Max Activating Token Index: 990

Text #5

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.1923. Min Act: -0.0001

Data Index: 533639 (C4 (Web Text))

Max Activating Token Index: 668

Text #6

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.1819. Min Act: -0.0001

Data Index: 273687 (C4 (Web Text))

Max Activating Token Index: 784

Text #7

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.1830. Min Act: -0.0000

Data Index: 1012187 (C4 (Web Text))

Max Activating Token Index: 430

Text #8

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.2876. Min Act: -0.0000

Data Index: 451961 (C4 (Web Text))

Max Activating Token Index: 832

Text #9

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.1392. Min Act: -0.0001

Data Index: 55 (C4 (Web Text))

Max Activating Token Index: 146

Text #10

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0607. Min Act: -0.0001

Data Index: 1272199 (C4 (Web Text))

Max Activating Token Index: 170

Text #11

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0561. Min Act: -0.0001

Data Index: 886624 (C4 (Web Text))

Max Activating Token Index: 665

Text #12

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0502. Min Act: -0.0001

Data Index: 1111418 (C4 (Web Text))

Max Activating Token Index: 542

Text #13

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0201. Min Act: -0.0001

Data Index: 1623453 (Python Code)

Max Activating Token Index: 800

Text #14

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0022. Min Act: -0.0001

Data Index: 919832 (C4 (Web Text))

Max Activating Token Index: 51

Text #15

Max Range: 1.5489. Min Range: -1.5489

Max Act: 0.9682. Min Act: -0.0001

Data Index: 1478001 (Python Code)

Max Activating Token Index: 840

Text #16

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0013. Min Act: -0.0001

Data Index: 279379 (C4 (Web Text))

Max Activating Token Index: 81

Text #17

Max Range: 1.5489. Min Range: -1.5489

Max Act: 1.0135. Min Act: -0.0001

Data Index: 939992 (C4 (Web Text))

Max Activating Token Index: 513

Text #18

Max Range: 1.5489. Min Range: -1.5489

Max Act: 0.8981. Min Act: -0.0001

Data Index: 855030 (C4 (Web Text))

Max Activating Token Index: 84

Text #19

Max Range: 1.5489. Min Range: -1.5489

Max Act: 0.9787. Min Act: -0.0000

Data Index: 1186573 (C4 (Web Text))

Max Activating Token Index: 621