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

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

Neuron 4743 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: 2.3518. Min Range: -2.3518

Max Act: 2.3221. Min Act: -0.0000

Data Index: 1437706 (Python Code)

Max Activating Token Index: 953

Text #1

Max Range: 2.3518. Min Range: -2.3518

Max Act: 2.3518. Min Act: -0.0000

Data Index: 1412340 (Python Code)

Max Activating Token Index: 307

Text #2

Max Range: 2.3518. Min Range: -2.3518

Max Act: 2.1405. Min Act: -0.0000

Data Index: 1578262 (Python Code)

Max Activating Token Index: 279

Text #3

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.9983. Min Act: 0.0000

Data Index: 1418238 (Python Code)

Max Activating Token Index: 502

Text #4

Max Range: 2.3518. Min Range: -2.3518

Max Act: 2.0888. Min Act: -0.0000

Data Index: 287825 (C4 (Web Text))

Max Activating Token Index: 21

Text #5

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.9667. Min Act: -0.0000

Data Index: 1352751 (C4 (Web Text))

Max Activating Token Index: 23

Text #6

Max Range: 2.3518. Min Range: -2.3518

Max Act: 2.0570. Min Act: -0.0000

Data Index: 242417 (C4 (Web Text))

Max Activating Token Index: 386

Text #7

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.7674. Min Act: 0.0001

Data Index: 1389198 (Python Code)

Max Activating Token Index: 823

Text #8

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.9219. Min Act: -0.0000

Data Index: 572858 (C4 (Web Text))

Max Activating Token Index: 332

Text #9

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.9238. Min Act: -0.0000

Data Index: 1480407 (Python Code)

Max Activating Token Index: 730

Text #10

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.7943. Min Act: -0.0000

Data Index: 255195 (C4 (Web Text))

Max Activating Token Index: 119

Text #11

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.8740. Min Act: -0.0000

Data Index: 1413714 (Python Code)

Max Activating Token Index: 36

Text #12

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.9058. Min Act: -0.0000

Data Index: 140857 (C4 (Web Text))

Max Activating Token Index: 472

Text #13

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.7744. Min Act: 0.0000

Data Index: 1592725 (Python Code)

Max Activating Token Index: 64

Text #14

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.7610. Min Act: -0.0000

Data Index: 1025672 (C4 (Web Text))

Max Activating Token Index: 183

Text #15

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.7156. Min Act: -0.0000

Data Index: 1507801 (Python Code)

Max Activating Token Index: 169

Text #16

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.7487. Min Act: -0.0000

Data Index: 238187 (C4 (Web Text))

Max Activating Token Index: 924

Text #17

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.8133. Min Act: -0.0000

Data Index: 1365769 (Python Code)

Max Activating Token Index: 245

Text #18

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.6135. Min Act: 0.0000

Data Index: 519144 (C4 (Web Text))

Max Activating Token Index: 364

Text #19

Max Range: 2.3518. Min Range: -2.3518

Max Act: 1.6088. Min Act: 0.0001

Data Index: 407852 (C4 (Web Text))

Max Activating Token Index: 262