Model: SoLU Model: 6 Layers, 3072 Neurons per Layer

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

Neuron 36 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-6l')

Text #0

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2846. Min Act: -0.0001

Data Index: 467247 (C4 (Web Text))

Max Activating Token Index: 72

Text #1

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2980. Min Act: -0.0001

Data Index: 241830 (C4 (Web Text))

Max Activating Token Index: 925

Text #2

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2737. Min Act: -0.0001

Data Index: 1244937 (C4 (Web Text))

Max Activating Token Index: 334

Text #3

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2248. Min Act: -0.0001

Data Index: 1026641 (C4 (Web Text))

Max Activating Token Index: 289

Text #4

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2736. Min Act: -0.0001

Data Index: 1290619 (C4 (Web Text))

Max Activating Token Index: 399

Text #5

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2486. Min Act: -0.0001

Data Index: 946135 (C4 (Web Text))

Max Activating Token Index: 137

Text #6

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2665. Min Act: -0.0001

Data Index: 700722 (C4 (Web Text))

Max Activating Token Index: 199

Text #7

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2573. Min Act: -0.0001

Data Index: 585810 (C4 (Web Text))

Max Activating Token Index: 5

Text #8

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2358. Min Act: -0.0001

Data Index: 1442894 (Python Code)

Max Activating Token Index: 708

Text #9

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2138. Min Act: -0.0001

Data Index: 418619 (C4 (Web Text))

Max Activating Token Index: 1007

Text #10

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2223. Min Act: -0.0001

Data Index: 1498383 (Python Code)

Max Activating Token Index: 929

Text #11

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2399. Min Act: -0.0001

Data Index: 639959 (C4 (Web Text))

Max Activating Token Index: 147

Text #12

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2289. Min Act: -0.0001

Data Index: 885266 (C4 (Web Text))

Max Activating Token Index: 220

Text #13

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2240. Min Act: -0.0001

Data Index: 7973 (C4 (Web Text))

Max Activating Token Index: 367

Text #14

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2363. Min Act: -0.0001

Data Index: 677048 (C4 (Web Text))

Max Activating Token Index: 182

Text #15

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2478. Min Act: -0.0001

Data Index: 917083 (C4 (Web Text))

Max Activating Token Index: 211

Text #16

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2206. Min Act: -0.0001

Data Index: 1525968 (Python Code)

Max Activating Token Index: 92

Text #17

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2215. Min Act: -0.0001

Data Index: 405176 (C4 (Web Text))

Max Activating Token Index: 998

Text #18

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2132. Min Act: -0.0001

Data Index: 30550 (C4 (Web Text))

Max Activating Token Index: 624

Text #19

Max Range: 0.2980. Min Range: -0.2980

Max Act: 0.2448. Min Act: -0.0001

Data Index: 754354 (C4 (Web Text))

Max Activating Token Index: 204