Model: SoLU Model: 8 Layers, 4096 Neurons per Layer

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

Neuron 336 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-8l')

Text #0

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1725. Min Act: -0.0001

Data Index: 513693 (C4 (Web Text))

Max Activating Token Index: 765

Text #1

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1544. Min Act: -0.0001

Data Index: 802594 (C4 (Web Text))

Max Activating Token Index: 37

Text #2

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1466. Min Act: -0.0001

Data Index: 295569 (C4 (Web Text))

Max Activating Token Index: 814

Text #3

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1409. Min Act: -0.0001

Data Index: 297851 (C4 (Web Text))

Max Activating Token Index: 231

Text #4

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1233. Min Act: -0.0001

Data Index: 643652 (C4 (Web Text))

Max Activating Token Index: 557

Text #5

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1341. Min Act: -0.0001

Data Index: 776645 (C4 (Web Text))

Max Activating Token Index: 760

Text #6

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1245. Min Act: -0.0001

Data Index: 11257 (C4 (Web Text))

Max Activating Token Index: 403

Text #7

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1285. Min Act: -0.0001

Data Index: 358523 (C4 (Web Text))

Max Activating Token Index: 617

Text #8

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1164. Min Act: -0.0001

Data Index: 876415 (C4 (Web Text))

Max Activating Token Index: 797

Text #9

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1232. Min Act: -0.0001

Data Index: 839092 (C4 (Web Text))

Max Activating Token Index: 338

Text #10

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1226. Min Act: -0.0001

Data Index: 1051777 (C4 (Web Text))

Max Activating Token Index: 965

Text #11

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1137. Min Act: -0.0001

Data Index: 1610431 (Python Code)

Max Activating Token Index: 631

Text #12

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1248. Min Act: -0.0001

Data Index: 940044 (C4 (Web Text))

Max Activating Token Index: 640

Text #13

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1226. Min Act: -0.0001

Data Index: 929834 (C4 (Web Text))

Max Activating Token Index: 298

Text #14

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1241. Min Act: -0.0001

Data Index: 1479850 (Python Code)

Max Activating Token Index: 252

Text #15

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1188. Min Act: -0.0001

Data Index: 1196456 (C4 (Web Text))

Max Activating Token Index: 155

Text #16

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1189. Min Act: -0.0001

Data Index: 986091 (C4 (Web Text))

Max Activating Token Index: 586

Text #17

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1234. Min Act: -0.0001

Data Index: 1067711 (C4 (Web Text))

Max Activating Token Index: 389

Text #18

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1174. Min Act: -0.0001

Data Index: 1487473 (Python Code)

Max Activating Token Index: 340

Text #19

Max Range: 0.1725. Min Range: -0.1725

Max Act: 0.1112. Min Act: -0.0001

Data Index: 1346116 (C4 (Web Text))

Max Activating Token Index: 658