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

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

Neuron 768 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0497. Min Act: -0.0003

Data Index: 95222 (C4 (Web Text))

Max Activating Token Index: 607

Text #1

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0428. Min Act: -0.0003

Data Index: 184607 (C4 (Web Text))

Max Activating Token Index: 676

Text #2

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0433. Min Act: -0.0003

Data Index: 687593 (C4 (Web Text))

Max Activating Token Index: 420

Text #3

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0411. Min Act: -0.0003

Data Index: 304477 (C4 (Web Text))

Max Activating Token Index: 338

Text #4

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0420. Min Act: -0.0003

Data Index: 1249948 (C4 (Web Text))

Max Activating Token Index: 193

Text #5

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0434. Min Act: -0.0003

Data Index: 994298 (C4 (Web Text))

Max Activating Token Index: 507

Text #6

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0415. Min Act: -0.0003

Data Index: 1197225 (C4 (Web Text))

Max Activating Token Index: 884

Text #7

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0403. Min Act: -0.0003

Data Index: 613657 (C4 (Web Text))

Max Activating Token Index: 373

Text #8

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0409. Min Act: -0.0002

Data Index: 793536 (C4 (Web Text))

Max Activating Token Index: 160

Text #9

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0393. Min Act: -0.0003

Data Index: 1252345 (C4 (Web Text))

Max Activating Token Index: 401

Text #10

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0381. Min Act: -0.0004

Data Index: 548515 (C4 (Web Text))

Max Activating Token Index: 627

Text #11

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0389. Min Act: -0.0004

Data Index: 998378 (C4 (Web Text))

Max Activating Token Index: 584

Text #12

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0381. Min Act: -0.0003

Data Index: 139141 (C4 (Web Text))

Max Activating Token Index: 864

Text #13

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0410. Min Act: -0.0004

Data Index: 1337235 (C4 (Web Text))

Max Activating Token Index: 640

Text #14

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0391. Min Act: -0.0003

Data Index: 621135 (C4 (Web Text))

Max Activating Token Index: 885

Text #15

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0384. Min Act: -0.0003

Data Index: 1520803 (Python Code)

Max Activating Token Index: 718

Text #16

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0392. Min Act: -0.0002

Data Index: 1524608 (Python Code)

Max Activating Token Index: 964

Text #17

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0417. Min Act: -0.0002

Data Index: 1103536 (C4 (Web Text))

Max Activating Token Index: 167

Text #18

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0391. Min Act: -0.0003

Data Index: 987063 (C4 (Web Text))

Max Activating Token Index: 232

Text #19

Max Range: 0.0497. Min Range: -0.0497

Max Act: 0.0365. Min Act: -0.0002

Data Index: 259173 (C4 (Web Text))

Max Activating Token Index: 919