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

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

Neuron 567 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: 0.0566. Min Range: -0.0566

Max Act: 0.0566. Min Act: -0.0006

Data Index: 149389 (C4 (Web Text))

Max Activating Token Index: 378

Text #1

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0550. Min Act: -0.0006

Data Index: 998368 (C4 (Web Text))

Max Activating Token Index: 965

Text #2

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0475. Min Act: -0.0006

Data Index: 663918 (C4 (Web Text))

Max Activating Token Index: 184

Text #3

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0476. Min Act: -0.0006

Data Index: 934795 (C4 (Web Text))

Max Activating Token Index: 473

Text #4

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0468. Min Act: -0.0006

Data Index: 1191995 (C4 (Web Text))

Max Activating Token Index: 998

Text #5

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0447. Min Act: -0.0006

Data Index: 1245397 (C4 (Web Text))

Max Activating Token Index: 777

Text #6

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0438. Min Act: -0.0006

Data Index: 52504 (C4 (Web Text))

Max Activating Token Index: 369

Text #7

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0417. Min Act: -0.0006

Data Index: 1463599 (Python Code)

Max Activating Token Index: 387

Text #8

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0416. Min Act: -0.0006

Data Index: 176548 (C4 (Web Text))

Max Activating Token Index: 500

Text #9

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0413. Min Act: -0.0006

Data Index: 377576 (C4 (Web Text))

Max Activating Token Index: 465

Text #10

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0416. Min Act: -0.0006

Data Index: 1484595 (Python Code)

Max Activating Token Index: 984

Text #11

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0414. Min Act: -0.0006

Data Index: 137960 (C4 (Web Text))

Max Activating Token Index: 272

Text #12

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0404. Min Act: -0.0006

Data Index: 37311 (C4 (Web Text))

Max Activating Token Index: 273

Text #13

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0397. Min Act: -0.0006

Data Index: 734024 (C4 (Web Text))

Max Activating Token Index: 432

Text #14

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0413. Min Act: -0.0006

Data Index: 774162 (C4 (Web Text))

Max Activating Token Index: 711

Text #15

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0411. Min Act: -0.0006

Data Index: 1021110 (C4 (Web Text))

Max Activating Token Index: 975

Text #16

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0404. Min Act: -0.0006

Data Index: 991696 (C4 (Web Text))

Max Activating Token Index: 201

Text #17

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0401. Min Act: -0.0006

Data Index: 804335 (C4 (Web Text))

Max Activating Token Index: 623

Text #18

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0408. Min Act: -0.0006

Data Index: 1535479 (Python Code)

Max Activating Token Index: 457

Text #19

Max Range: 0.0566. Min Range: -0.0566

Max Act: 0.0391. Min Act: -0.0006

Data Index: 332559 (C4 (Web Text))

Max Activating Token Index: 187