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

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

Neuron 30 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.1360. Min Act: -0.0005

Data Index: 316881 (C4 (Web Text))

Max Activating Token Index: 246

Text #1

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0368. Min Act: -0.0005

Data Index: 507807 (C4 (Web Text))

Max Activating Token Index: 442

Text #2

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0260. Min Act: -0.0005

Data Index: 682823 (C4 (Web Text))

Max Activating Token Index: 915

Text #3

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0249. Min Act: -0.0005

Data Index: 845299 (C4 (Web Text))

Max Activating Token Index: 998

Text #4

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0274. Min Act: -0.0005

Data Index: 286237 (C4 (Web Text))

Max Activating Token Index: 966

Text #5

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0240. Min Act: -0.0005

Data Index: 1425548 (Python Code)

Max Activating Token Index: 772

Text #6

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0158. Min Act: -0.0005

Data Index: 504562 (C4 (Web Text))

Max Activating Token Index: 847

Text #7

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0198. Min Act: -0.0005

Data Index: 1145184 (C4 (Web Text))

Max Activating Token Index: 583

Text #8

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0200. Min Act: -0.0005

Data Index: 879781 (C4 (Web Text))

Max Activating Token Index: 828

Text #9

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0189. Min Act: -0.0005

Data Index: 210038 (C4 (Web Text))

Max Activating Token Index: 1010

Text #10

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0196. Min Act: -0.0005

Data Index: 742437 (C4 (Web Text))

Max Activating Token Index: 300

Text #11

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0191. Min Act: -0.0005

Data Index: 1412980 (Python Code)

Max Activating Token Index: 478

Text #12

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0188. Min Act: -0.0004

Data Index: 1267479 (C4 (Web Text))

Max Activating Token Index: 813

Text #13

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0196. Min Act: -0.0005

Data Index: 585541 (C4 (Web Text))

Max Activating Token Index: 710

Text #14

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0193. Min Act: -0.0005

Data Index: 1538202 (Python Code)

Max Activating Token Index: 978

Text #15

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0155. Min Act: -0.0005

Data Index: 497799 (C4 (Web Text))

Max Activating Token Index: 985

Text #16

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0181. Min Act: -0.0005

Data Index: 734670 (C4 (Web Text))

Max Activating Token Index: 318

Text #17

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0170. Min Act: -0.0005

Data Index: 537383 (C4 (Web Text))

Max Activating Token Index: 15

Text #18

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0181. Min Act: -0.0005

Data Index: 1612475 (Python Code)

Max Activating Token Index: 981

Text #19

Max Range: 0.1360. Min Range: -0.1360

Max Act: 0.0157. Min Act: -0.0005

Data Index: 770782 (C4 (Web Text))

Max Activating Token Index: 366