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

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

Neuron 48 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: 0.1082. Min Range: -0.1082

Max Act: 0.1075. Min Act: -0.0002

Data Index: 863539 (C4 (Web Text))

Max Activating Token Index: 338

Text #1

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.1080. Min Act: -0.0002

Data Index: 438523 (C4 (Web Text))

Max Activating Token Index: 756

Text #2

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.1028. Min Act: -0.0002

Data Index: 904537 (C4 (Web Text))

Max Activating Token Index: 497

Text #3

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.1082. Min Act: -0.0002

Data Index: 662849 (C4 (Web Text))

Max Activating Token Index: 408

Text #4

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0969. Min Act: -0.0002

Data Index: 1601927 (Python Code)

Max Activating Token Index: 968

Text #5

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.1017. Min Act: -0.0002

Data Index: 94108 (C4 (Web Text))

Max Activating Token Index: 978

Text #6

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0888. Min Act: -0.0002

Data Index: 1469552 (Python Code)

Max Activating Token Index: 725

Text #7

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.1038. Min Act: -0.0002

Data Index: 1414306 (Python Code)

Max Activating Token Index: 965

Text #8

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0927. Min Act: -0.0002

Data Index: 515087 (C4 (Web Text))

Max Activating Token Index: 577

Text #9

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0992. Min Act: -0.0002

Data Index: 301908 (C4 (Web Text))

Max Activating Token Index: 747

Text #10

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0934. Min Act: -0.0002

Data Index: 407167 (C4 (Web Text))

Max Activating Token Index: 430

Text #11

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0912. Min Act: -0.0002

Data Index: 1507005 (Python Code)

Max Activating Token Index: 642

Text #12

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0940. Min Act: -0.0002

Data Index: 1494241 (Python Code)

Max Activating Token Index: 616

Text #13

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0915. Min Act: -0.0002

Data Index: 60832 (C4 (Web Text))

Max Activating Token Index: 306

Text #14

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0945. Min Act: -0.0002

Data Index: 1646942 (Python Code)

Max Activating Token Index: 773

Text #15

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0900. Min Act: -0.0002

Data Index: 1000952 (C4 (Web Text))

Max Activating Token Index: 528

Text #16

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0822. Min Act: -0.0002

Data Index: 1620703 (Python Code)

Max Activating Token Index: 788

Text #17

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0878. Min Act: -0.0002

Data Index: 1422294 (Python Code)

Max Activating Token Index: 1019

Text #18

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0829. Min Act: -0.0002

Data Index: 153601 (C4 (Web Text))

Max Activating Token Index: 337

Text #19

Max Range: 0.1082. Min Range: -0.1082

Max Act: 0.0886. Min Act: -0.0002

Data Index: 210559 (C4 (Web Text))

Max Activating Token Index: 106