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

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

Neuron 1312 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0315. Min Act: -0.0006

Data Index: 1113310 (C4 (Web Text))

Max Activating Token Index: 603

Text #1

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0295. Min Act: -0.0006

Data Index: 489995 (C4 (Web Text))

Max Activating Token Index: 519

Text #2

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0290. Min Act: -0.0006

Data Index: 1377033 (Python Code)

Max Activating Token Index: 418

Text #3

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0275. Min Act: -0.0006

Data Index: 145093 (C4 (Web Text))

Max Activating Token Index: 946

Text #4

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0287. Min Act: -0.0006

Data Index: 320403 (C4 (Web Text))

Max Activating Token Index: 621

Text #5

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0280. Min Act: -0.0006

Data Index: 920358 (C4 (Web Text))

Max Activating Token Index: 869

Text #6

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0250. Min Act: -0.0006

Data Index: 1219515 (C4 (Web Text))

Max Activating Token Index: 719

Text #7

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0272. Min Act: -0.0006

Data Index: 1445840 (Python Code)

Max Activating Token Index: 15

Text #8

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0265. Min Act: -0.0006

Data Index: 524441 (C4 (Web Text))

Max Activating Token Index: 833

Text #9

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0257. Min Act: -0.0006

Data Index: 114019 (C4 (Web Text))

Max Activating Token Index: 522

Text #10

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0258. Min Act: -0.0006

Data Index: 175418 (C4 (Web Text))

Max Activating Token Index: 831

Text #11

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0253. Min Act: -0.0006

Data Index: 1523766 (Python Code)

Max Activating Token Index: 832

Text #12

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0257. Min Act: -0.0006

Data Index: 731583 (C4 (Web Text))

Max Activating Token Index: 409

Text #13

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0255. Min Act: -0.0006

Data Index: 1343890 (C4 (Web Text))

Max Activating Token Index: 759

Text #14

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0249. Min Act: -0.0006

Data Index: 781993 (C4 (Web Text))

Max Activating Token Index: 161

Text #15

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0265. Min Act: -0.0006

Data Index: 1365796 (Python Code)

Max Activating Token Index: 417

Text #16

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0278. Min Act: -0.0006

Data Index: 340840 (C4 (Web Text))

Max Activating Token Index: 781

Text #17

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0271. Min Act: -0.0006

Data Index: 422358 (C4 (Web Text))

Max Activating Token Index: 976

Text #18

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0253. Min Act: -0.0006

Data Index: 7333 (C4 (Web Text))

Max Activating Token Index: 767

Text #19

Max Range: 0.0315. Min Range: -0.0315

Max Act: 0.0252. Min Act: -0.0006

Data Index: 1557861 (Python Code)

Max Activating Token Index: 8