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

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

Neuron 1800 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: 0.0352. Min Range: -0.0352

Max Act: 0.0352. Min Act: -0.0006

Data Index: 973010 (C4 (Web Text))

Max Activating Token Index: 300

Text #1

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0335. Min Act: -0.0006

Data Index: 1553469 (Python Code)

Max Activating Token Index: 359

Text #2

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0326. Min Act: -0.0006

Data Index: 1380138 (Python Code)

Max Activating Token Index: 460

Text #3

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0317. Min Act: -0.0006

Data Index: 177365 (C4 (Web Text))

Max Activating Token Index: 628

Text #4

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0317. Min Act: -0.0006

Data Index: 814708 (C4 (Web Text))

Max Activating Token Index: 35

Text #5

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0311. Min Act: -0.0006

Data Index: 1151019 (C4 (Web Text))

Max Activating Token Index: 718

Text #6

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0307. Min Act: -0.0006

Data Index: 1443264 (Python Code)

Max Activating Token Index: 323

Text #7

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0307. Min Act: -0.0006

Data Index: 1015343 (C4 (Web Text))

Max Activating Token Index: 24

Text #8

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0293. Min Act: -0.0006

Data Index: 327455 (C4 (Web Text))

Max Activating Token Index: 683

Text #9

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0301. Min Act: -0.0006

Data Index: 1472108 (Python Code)

Max Activating Token Index: 196

Text #10

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0299. Min Act: -0.0006

Data Index: 253675 (C4 (Web Text))

Max Activating Token Index: 6

Text #11

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0291. Min Act: -0.0006

Data Index: 1412235 (Python Code)

Max Activating Token Index: 122

Text #12

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0289. Min Act: -0.0006

Data Index: 694402 (C4 (Web Text))

Max Activating Token Index: 87

Text #13

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0288. Min Act: -0.0006

Data Index: 1067563 (C4 (Web Text))

Max Activating Token Index: 8

Text #14

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0294. Min Act: -0.0006

Data Index: 1592299 (Python Code)

Max Activating Token Index: 69

Text #15

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0300. Min Act: -0.0006

Data Index: 68662 (C4 (Web Text))

Max Activating Token Index: 20

Text #16

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0299. Min Act: -0.0005

Data Index: 1280511 (C4 (Web Text))

Max Activating Token Index: 25

Text #17

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0287. Min Act: -0.0006

Data Index: 1413466 (Python Code)

Max Activating Token Index: 304

Text #18

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0306. Min Act: -0.0006

Data Index: 728964 (C4 (Web Text))

Max Activating Token Index: 720

Text #19

Max Range: 0.0352. Min Range: -0.0352

Max Act: 0.0299. Min Act: -0.0006

Data Index: 748754 (C4 (Web Text))

Max Activating Token Index: 327