Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

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

Neuron 4817 in Layer 8

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.2928. Min Act: -0.0000

Data Index: 922451 (C4 (Web Text))

Max Activating Token Index: 892

Text #1

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1948. Min Act: -0.0000

Data Index: 773758 (C4 (Web Text))

Max Activating Token Index: 470

Text #2

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1678. Min Act: -0.0000

Data Index: 69621 (C4 (Web Text))

Max Activating Token Index: 677

Text #3

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1658. Min Act: -0.0000

Data Index: 1228142 (C4 (Web Text))

Max Activating Token Index: 503

Text #4

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1751. Min Act: -0.0000

Data Index: 200327 (C4 (Web Text))

Max Activating Token Index: 65

Text #5

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1599. Min Act: 0.0000

Data Index: 582050 (C4 (Web Text))

Max Activating Token Index: 365

Text #6

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1392. Min Act: -0.0000

Data Index: 1395262 (Python Code)

Max Activating Token Index: 307

Text #7

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1400. Min Act: -0.0000

Data Index: 1286092 (C4 (Web Text))

Max Activating Token Index: 450

Text #8

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1250. Min Act: -0.0000

Data Index: 262043 (C4 (Web Text))

Max Activating Token Index: 981

Text #9

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1248. Min Act: -0.0000

Data Index: 1584770 (Python Code)

Max Activating Token Index: 629

Text #10

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1425. Min Act: -0.0000

Data Index: 339435 (C4 (Web Text))

Max Activating Token Index: 155

Text #11

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1360. Min Act: -0.0000

Data Index: 1476517 (Python Code)

Max Activating Token Index: 390

Text #12

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1286. Min Act: -0.0000

Data Index: 61074 (C4 (Web Text))

Max Activating Token Index: 122

Text #13

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1293. Min Act: -0.0000

Data Index: 1491013 (Python Code)

Max Activating Token Index: 696

Text #14

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1278. Min Act: -0.0000

Data Index: 1637773 (Python Code)

Max Activating Token Index: 973

Text #15

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1314. Min Act: -0.0000

Data Index: 344454 (C4 (Web Text))

Max Activating Token Index: 154

Text #16

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1259. Min Act: -0.0000

Data Index: 1571442 (Python Code)

Max Activating Token Index: 377

Text #17

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1251. Min Act: 0.0000

Data Index: 25365 (C4 (Web Text))

Max Activating Token Index: 321

Text #18

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1247. Min Act: -0.0000

Data Index: 971482 (C4 (Web Text))

Max Activating Token Index: 249

Text #19

Max Range: 0.2928. Min Range: -0.2928

Max Act: 0.1197. Min Act: -0.0000

Data Index: 601466 (C4 (Web Text))

Max Activating Token Index: 312