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

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

Neuron 1444 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-4l')

Text #0

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0050. Min Act: -0.0006

Data Index: 1779 (C4 (Web Text))

Max Activating Token Index: 640

Text #1

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0046. Min Act: -0.0005

Data Index: 833999 (C4 (Web Text))

Max Activating Token Index: 813

Text #2

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0047. Min Act: -0.0006

Data Index: 520880 (C4 (Web Text))

Max Activating Token Index: 302

Text #3

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0045. Min Act: -0.0006

Data Index: 1381853 (Python Code)

Max Activating Token Index: 973

Text #4

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0045. Min Act: -0.0005

Data Index: 1253710 (C4 (Web Text))

Max Activating Token Index: 743

Text #5

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0045. Min Act: -0.0006

Data Index: 839508 (C4 (Web Text))

Max Activating Token Index: 496

Text #6

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0044. Min Act: -0.0006

Data Index: 706730 (C4 (Web Text))

Max Activating Token Index: 798

Text #7

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0044. Min Act: -0.0005

Data Index: 794379 (C4 (Web Text))

Max Activating Token Index: 787

Text #8

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0043. Min Act: -0.0006

Data Index: 136123 (C4 (Web Text))

Max Activating Token Index: 795

Text #9

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0044. Min Act: -0.0005

Data Index: 29044 (C4 (Web Text))

Max Activating Token Index: 611

Text #10

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0042. Min Act: -0.0005

Data Index: 202415 (C4 (Web Text))

Max Activating Token Index: 552

Text #11

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0044. Min Act: -0.0006

Data Index: 449091 (C4 (Web Text))

Max Activating Token Index: 791

Text #12

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0042. Min Act: -0.0006

Data Index: 1034662 (C4 (Web Text))

Max Activating Token Index: 900

Text #13

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0042. Min Act: -0.0005

Data Index: 692862 (C4 (Web Text))

Max Activating Token Index: 509

Text #14

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0041. Min Act: -0.0006

Data Index: 635655 (C4 (Web Text))

Max Activating Token Index: 652

Text #15

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0042. Min Act: -0.0006

Data Index: 990307 (C4 (Web Text))

Max Activating Token Index: 287

Text #16

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0042. Min Act: -0.0005

Data Index: 924678 (C4 (Web Text))

Max Activating Token Index: 812

Text #17

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0042. Min Act: -0.0005

Data Index: 14579 (C4 (Web Text))

Max Activating Token Index: 972

Text #18

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0040. Min Act: -0.0006

Data Index: 1579160 (Python Code)

Max Activating Token Index: 805

Text #19

Max Range: 0.0050. Min Range: -0.0050

Max Act: 0.0040. Min Act: -0.0006

Data Index: 218127 (C4 (Web Text))

Max Activating Token Index: 565