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

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

Neuron 3850 in Layer 6

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1739. Min Act: -0.0000

Data Index: 914773 (C4 (Web Text))

Max Activating Token Index: 43

Text #1

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1552. Min Act: -0.0001

Data Index: 534163 (C4 (Web Text))

Max Activating Token Index: 753

Text #2

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1413. Min Act: -0.0000

Data Index: 316606 (C4 (Web Text))

Max Activating Token Index: 58

Text #3

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1422. Min Act: -0.0000

Data Index: 398081 (C4 (Web Text))

Max Activating Token Index: 50

Text #4

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1310. Min Act: -0.0000

Data Index: 590811 (C4 (Web Text))

Max Activating Token Index: 799

Text #5

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1358. Min Act: -0.0000

Data Index: 1023894 (C4 (Web Text))

Max Activating Token Index: 455

Text #6

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1300. Min Act: -0.0001

Data Index: 287237 (C4 (Web Text))

Max Activating Token Index: 278

Text #7

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1307. Min Act: -0.0000

Data Index: 82421 (C4 (Web Text))

Max Activating Token Index: 945

Text #8

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1276. Min Act: -0.0000

Data Index: 16858 (C4 (Web Text))

Max Activating Token Index: 331

Text #9

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1212. Min Act: -0.0000

Data Index: 1340745 (C4 (Web Text))

Max Activating Token Index: 90

Text #10

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1200. Min Act: -0.0000

Data Index: 707030 (C4 (Web Text))

Max Activating Token Index: 986

Text #11

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1273. Min Act: -0.0001

Data Index: 918451 (C4 (Web Text))

Max Activating Token Index: 481

Text #12

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1165. Min Act: -0.0000

Data Index: 337177 (C4 (Web Text))

Max Activating Token Index: 828

Text #13

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1231. Min Act: -0.0000

Data Index: 88574 (C4 (Web Text))

Max Activating Token Index: 675

Text #14

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1147. Min Act: -0.0001

Data Index: 1496245 (Python Code)

Max Activating Token Index: 160

Text #15

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1099. Min Act: -0.0000

Data Index: 768117 (C4 (Web Text))

Max Activating Token Index: 716

Text #16

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1029. Min Act: -0.0001

Data Index: 956315 (C4 (Web Text))

Max Activating Token Index: 675

Text #17

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1181. Min Act: -0.0000

Data Index: 979574 (C4 (Web Text))

Max Activating Token Index: 801

Text #18

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1101. Min Act: -0.0000

Data Index: 621733 (C4 (Web Text))

Max Activating Token Index: 51

Text #19

Max Range: 0.1739. Min Range: -0.1739

Max Act: 0.1188. Min Act: -0.0000

Data Index: 859723 (C4 (Web Text))

Max Activating Token Index: 878