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

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

Neuron 536 in Layer 7

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6559. Min Act: -0.0001

Data Index: 596780 (C4 (Web Text))

Max Activating Token Index: 431

Text #1

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6447. Min Act: -0.0001

Data Index: 995894 (C4 (Web Text))

Max Activating Token Index: 885

Text #2

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6332. Min Act: -0.0000

Data Index: 519957 (C4 (Web Text))

Max Activating Token Index: 309

Text #3

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6231. Min Act: -0.0000

Data Index: 1237850 (C4 (Web Text))

Max Activating Token Index: 359

Text #4

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6423. Min Act: -0.0001

Data Index: 407625 (C4 (Web Text))

Max Activating Token Index: 805

Text #5

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6030. Min Act: -0.0001

Data Index: 196843 (C4 (Web Text))

Max Activating Token Index: 779

Text #6

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5556. Min Act: -0.0001

Data Index: 1529069 (Python Code)

Max Activating Token Index: 550

Text #7

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6119. Min Act: -0.0001

Data Index: 1284253 (C4 (Web Text))

Max Activating Token Index: 408

Text #8

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5947. Min Act: -0.0000

Data Index: 1067873 (C4 (Web Text))

Max Activating Token Index: 914

Text #9

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5432. Min Act: -0.0000

Data Index: 669787 (C4 (Web Text))

Max Activating Token Index: 690

Text #10

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5891. Min Act: -0.0001

Data Index: 714698 (C4 (Web Text))

Max Activating Token Index: 861

Text #11

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5758. Min Act: -0.0000

Data Index: 136672 (C4 (Web Text))

Max Activating Token Index: 257

Text #12

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5893. Min Act: -0.0001

Data Index: 1508800 (Python Code)

Max Activating Token Index: 691

Text #13

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.6019. Min Act: -0.0001

Data Index: 438281 (C4 (Web Text))

Max Activating Token Index: 40

Text #14

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5983. Min Act: -0.0000

Data Index: 65104 (C4 (Web Text))

Max Activating Token Index: 720

Text #15

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5762. Min Act: -0.0000

Data Index: 604878 (C4 (Web Text))

Max Activating Token Index: 419

Text #16

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5772. Min Act: -0.0001

Data Index: 161639 (C4 (Web Text))

Max Activating Token Index: 553

Text #17

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5562. Min Act: -0.0001

Data Index: 1291727 (C4 (Web Text))

Max Activating Token Index: 975

Text #18

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5519. Min Act: -0.0001

Data Index: 1247053 (C4 (Web Text))

Max Activating Token Index: 9

Text #19

Max Range: 0.6559. Min Range: -0.6559

Max Act: 0.5138. Min Act: -0.0001

Data Index: 1275945 (C4 (Web Text))

Max Activating Token Index: 508