Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

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

Neuron 4710 in Layer 7

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 1.3602. Min Range: -1.3602

Max Act: 1.3602. Min Act: -0.0000

Data Index: 1653106 (Python Code)

Max Activating Token Index: 365

Text #1

Max Range: 1.3602. Min Range: -1.3602

Max Act: 1.1445. Min Act: -0.0000

Data Index: 1199448 (C4 (Web Text))

Max Activating Token Index: 394

Text #2

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.9182. Min Act: -0.0000

Data Index: 1583960 (Python Code)

Max Activating Token Index: 687

Text #3

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.9098. Min Act: -0.0000

Data Index: 1363630 (Python Code)

Max Activating Token Index: 320

Text #4

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.8773. Min Act: -0.0000

Data Index: 1618626 (Python Code)

Max Activating Token Index: 180

Text #5

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.9075. Min Act: -0.0000

Data Index: 1504992 (Python Code)

Max Activating Token Index: 73

Text #6

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.8527. Min Act: -0.0000

Data Index: 136480 (C4 (Web Text))

Max Activating Token Index: 61

Text #7

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.8626. Min Act: -0.0000

Data Index: 87632 (C4 (Web Text))

Max Activating Token Index: 551

Text #8

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7870. Min Act: -0.0000

Data Index: 1461478 (Python Code)

Max Activating Token Index: 108

Text #9

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.8123. Min Act: -0.0000

Data Index: 476374 (C4 (Web Text))

Max Activating Token Index: 583

Text #10

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7834. Min Act: -0.0000

Data Index: 570692 (C4 (Web Text))

Max Activating Token Index: 102

Text #11

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7508. Min Act: -0.0000

Data Index: 1397424 (Python Code)

Max Activating Token Index: 563

Text #12

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7206. Min Act: -0.0000

Data Index: 834202 (C4 (Web Text))

Max Activating Token Index: 627

Text #13

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7401. Min Act: -0.0000

Data Index: 1150451 (C4 (Web Text))

Max Activating Token Index: 506

Text #14

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7494. Min Act: -0.0000

Data Index: 1505882 (Python Code)

Max Activating Token Index: 205

Text #15

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7153. Min Act: -0.0000

Data Index: 1480107 (Python Code)

Max Activating Token Index: 79

Text #16

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7596. Min Act: -0.0000

Data Index: 589320 (C4 (Web Text))

Max Activating Token Index: 369

Text #17

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7414. Min Act: -0.0000

Data Index: 658451 (C4 (Web Text))

Max Activating Token Index: 282

Text #18

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.6587. Min Act: -0.0000

Data Index: 1533037 (Python Code)

Max Activating Token Index: 25

Text #19

Max Range: 1.3602. Min Range: -1.3602

Max Act: 0.7046. Min Act: -0.0000

Data Index: 229071 (C4 (Web Text))

Max Activating Token Index: 223