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

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

Neuron 3041 in Layer 6

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.7442. Min Act: -0.0000

Data Index: 369517 (C4 (Web Text))

Max Activating Token Index: 790

Text #1

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.6620. Min Act: -0.0000

Data Index: 486108 (C4 (Web Text))

Max Activating Token Index: 326

Text #2

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.6392. Min Act: -0.0000

Data Index: 1441956 (Python Code)

Max Activating Token Index: 35

Text #3

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.7742. Min Act: -0.0000

Data Index: 330965 (C4 (Web Text))

Max Activating Token Index: 849

Text #4

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.7210. Min Act: -0.0000

Data Index: 487368 (C4 (Web Text))

Max Activating Token Index: 638

Text #5

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.4621. Min Act: -0.0000

Data Index: 1457006 (Python Code)

Max Activating Token Index: 558

Text #6

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.6832. Min Act: -0.0000

Data Index: 765535 (C4 (Web Text))

Max Activating Token Index: 972

Text #7

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.3326. Min Act: -0.0000

Data Index: 1146463 (C4 (Web Text))

Max Activating Token Index: 847

Text #8

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.1947. Min Act: -0.0000

Data Index: 1385891 (Python Code)

Max Activating Token Index: 425

Text #9

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.5290. Min Act: -0.0000

Data Index: 1613949 (Python Code)

Max Activating Token Index: 863

Text #10

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.1111. Min Act: -0.0000

Data Index: 1206648 (C4 (Web Text))

Max Activating Token Index: 294

Text #11

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.3649. Min Act: -0.0000

Data Index: 1619090 (Python Code)

Max Activating Token Index: 68

Text #12

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.3644. Min Act: -0.0000

Data Index: 919579 (C4 (Web Text))

Max Activating Token Index: 125

Text #13

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.3506. Min Act: -0.0000

Data Index: 1065360 (C4 (Web Text))

Max Activating Token Index: 241

Text #14

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.2141. Min Act: -0.0000

Data Index: 268002 (C4 (Web Text))

Max Activating Token Index: 535

Text #15

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.3342. Min Act: -0.0000

Data Index: 1654064 (Python Code)

Max Activating Token Index: 670

Text #16

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.1785. Min Act: -0.0000

Data Index: 802819 (C4 (Web Text))

Max Activating Token Index: 305

Text #17

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.1430. Min Act: -0.0000

Data Index: 1533860 (Python Code)

Max Activating Token Index: 552

Text #18

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.2369. Min Act: -0.0000

Data Index: 1449966 (Python Code)

Max Activating Token Index: 651

Text #19

Max Range: 2.7742. Min Range: -2.7742

Max Act: 2.2424. Min Act: -0.0000

Data Index: 1440433 (Python Code)

Max Activating Token Index: 43