Model: SoLU Model: 12 Layers, 6144 Neurons per Layer

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

Neuron 3582 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-12l')

Text #0

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1628. Min Act: -0.0001

Data Index: 95238 (C4 (Web Text))

Max Activating Token Index: 665

Text #1

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1440. Min Act: -0.0001

Data Index: 1616053 (Python Code)

Max Activating Token Index: 656

Text #2

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1514. Min Act: -0.0001

Data Index: 948893 (C4 (Web Text))

Max Activating Token Index: 232

Text #3

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1601. Min Act: -0.0001

Data Index: 484843 (C4 (Web Text))

Max Activating Token Index: 250

Text #4

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1363. Min Act: -0.0001

Data Index: 854005 (C4 (Web Text))

Max Activating Token Index: 409

Text #5

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1366. Min Act: -0.0001

Data Index: 1513149 (Python Code)

Max Activating Token Index: 605

Text #6

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1422. Min Act: -0.0001

Data Index: 1425058 (Python Code)

Max Activating Token Index: 132

Text #7

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1620. Min Act: -0.0001

Data Index: 1417292 (Python Code)

Max Activating Token Index: 816

Text #8

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1263. Min Act: -0.0001

Data Index: 584147 (C4 (Web Text))

Max Activating Token Index: 180

Text #9

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1340. Min Act: -0.0001

Data Index: 413516 (C4 (Web Text))

Max Activating Token Index: 355

Text #10

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1393. Min Act: -0.0001

Data Index: 1174658 (C4 (Web Text))

Max Activating Token Index: 326

Text #11

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1320. Min Act: -0.0001

Data Index: 1594326 (Python Code)

Max Activating Token Index: 366

Text #12

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1366. Min Act: -0.0001

Data Index: 265307 (C4 (Web Text))

Max Activating Token Index: 418

Text #13

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1373. Min Act: -0.0001

Data Index: 1403577 (Python Code)

Max Activating Token Index: 520

Text #14

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1329. Min Act: -0.0001

Data Index: 450491 (C4 (Web Text))

Max Activating Token Index: 218

Text #15

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1354. Min Act: -0.0001

Data Index: 1384657 (Python Code)

Max Activating Token Index: 500

Text #16

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1299. Min Act: -0.0001

Data Index: 1374229 (Python Code)

Max Activating Token Index: 250

Text #17

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1334. Min Act: -0.0001

Data Index: 370060 (C4 (Web Text))

Max Activating Token Index: 950

Text #18

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1332. Min Act: -0.0001

Data Index: 248946 (C4 (Web Text))

Max Activating Token Index: 299

Text #19

Max Range: 0.1628. Min Range: -0.1628

Max Act: 0.1235. Min Act: -0.0001

Data Index: 98188 (C4 (Web Text))

Max Activating Token Index: 1007