Model: SoLU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 550 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-3l')

Text #0

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0297. Min Act: -0.0003

Data Index: 1596730 (Python Code)

Max Activating Token Index: 275

Text #1

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0255. Min Act: -0.0002

Data Index: 431953 (C4 (Web Text))

Max Activating Token Index: 593

Text #2

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0251. Min Act: -0.0002

Data Index: 1328714 (C4 (Web Text))

Max Activating Token Index: 388

Text #3

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0251. Min Act: -0.0002

Data Index: 1097660 (C4 (Web Text))

Max Activating Token Index: 800

Text #4

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0244. Min Act: -0.0003

Data Index: 1349892 (C4 (Web Text))

Max Activating Token Index: 834

Text #5

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0253. Min Act: -0.0002

Data Index: 364785 (C4 (Web Text))

Max Activating Token Index: 851

Text #6

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0225. Min Act: -0.0003

Data Index: 1192130 (C4 (Web Text))

Max Activating Token Index: 721

Text #7

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0221. Min Act: -0.0003

Data Index: 1215764 (C4 (Web Text))

Max Activating Token Index: 253

Text #8

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0227. Min Act: -0.0003

Data Index: 1460540 (Python Code)

Max Activating Token Index: 677

Text #9

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0198. Min Act: -0.0002

Data Index: 771479 (C4 (Web Text))

Max Activating Token Index: 198

Text #10

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0215. Min Act: -0.0003

Data Index: 867962 (C4 (Web Text))

Max Activating Token Index: 673

Text #11

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0209. Min Act: -0.0003

Data Index: 1055110 (C4 (Web Text))

Max Activating Token Index: 864

Text #12

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0206. Min Act: -0.0001

Data Index: 703968 (C4 (Web Text))

Max Activating Token Index: 511

Text #13

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0203. Min Act: -0.0002

Data Index: 1251634 (C4 (Web Text))

Max Activating Token Index: 541

Text #14

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0206. Min Act: -0.0002

Data Index: 302612 (C4 (Web Text))

Max Activating Token Index: 126

Text #15

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0206. Min Act: -0.0002

Data Index: 791919 (C4 (Web Text))

Max Activating Token Index: 859

Text #16

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0199. Min Act: -0.0002

Data Index: 349331 (C4 (Web Text))

Max Activating Token Index: 638

Text #17

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0198. Min Act: -0.0003

Data Index: 1216315 (C4 (Web Text))

Max Activating Token Index: 442

Text #18

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0197. Min Act: -0.0002

Data Index: 1378611 (Python Code)

Max Activating Token Index: 580

Text #19

Max Range: 0.0297. Min Range: -0.0297

Max Act: 0.0195. Min Act: -0.0003

Data Index: 1351026 (C4 (Web Text))

Max Activating Token Index: 282