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

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

Neuron 4653 in Layer 9

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.5772. Min Act: -0.0000

Data Index: 1022609 (C4 (Web Text))

Max Activating Token Index: 849

Text #1

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.5724. Min Act: -0.0000

Data Index: 935962 (C4 (Web Text))

Max Activating Token Index: 755

Text #2

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.5404. Min Act: -0.0000

Data Index: 800343 (C4 (Web Text))

Max Activating Token Index: 127

Text #3

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4197. Min Act: -0.0000

Data Index: 32372 (C4 (Web Text))

Max Activating Token Index: 335

Text #4

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4488. Min Act: -0.0000

Data Index: 141155 (C4 (Web Text))

Max Activating Token Index: 349

Text #5

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4607. Min Act: -0.0000

Data Index: 965771 (C4 (Web Text))

Max Activating Token Index: 634

Text #6

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4291. Min Act: -0.0000

Data Index: 943695 (C4 (Web Text))

Max Activating Token Index: 102

Text #7

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4494. Min Act: -0.0000

Data Index: 468384 (C4 (Web Text))

Max Activating Token Index: 727

Text #8

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4520. Min Act: -0.0000

Data Index: 1367563 (Python Code)

Max Activating Token Index: 712

Text #9

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4153. Min Act: -0.0000

Data Index: 1435221 (Python Code)

Max Activating Token Index: 861

Text #10

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4357. Min Act: -0.0000

Data Index: 27927 (C4 (Web Text))

Max Activating Token Index: 652

Text #11

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4282. Min Act: -0.0000

Data Index: 8010 (C4 (Web Text))

Max Activating Token Index: 89

Text #12

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4166. Min Act: -0.0000

Data Index: 1132878 (C4 (Web Text))

Max Activating Token Index: 99

Text #13

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.3935. Min Act: -0.0000

Data Index: 1090844 (C4 (Web Text))

Max Activating Token Index: 965

Text #14

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.3998. Min Act: -0.0000

Data Index: 1536624 (Python Code)

Max Activating Token Index: 638

Text #15

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4057. Min Act: -0.0000

Data Index: 996696 (C4 (Web Text))

Max Activating Token Index: 196

Text #16

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4010. Min Act: -0.0000

Data Index: 849700 (C4 (Web Text))

Max Activating Token Index: 558

Text #17

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.4166. Min Act: -0.0000

Data Index: 673756 (C4 (Web Text))

Max Activating Token Index: 361

Text #18

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.3750. Min Act: -0.0000

Data Index: 341267 (C4 (Web Text))

Max Activating Token Index: 961

Text #19

Max Range: 0.5772. Min Range: -0.5772

Max Act: 0.3881. Min Act: -0.0000

Data Index: 632930 (C4 (Web Text))

Max Activating Token Index: 267