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

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

Neuron 3593 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.8440. Min Range: -3.8440

Max Act: 3.8440. Min Act: -0.0001

Data Index: 1313335 (C4 (Web Text))

Max Activating Token Index: 910

Text #1

Max Range: 3.8440. Min Range: -3.8440

Max Act: 2.3687. Min Act: -0.0001

Data Index: 852297 (C4 (Web Text))

Max Activating Token Index: 574

Text #2

Max Range: 3.8440. Min Range: -3.8440

Max Act: 2.1990. Min Act: -0.0000

Data Index: 812909 (C4 (Web Text))

Max Activating Token Index: 902

Text #3

Max Range: 3.8440. Min Range: -3.8440

Max Act: 2.1817. Min Act: -0.0001

Data Index: 38877 (C4 (Web Text))

Max Activating Token Index: 443

Text #4

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.9303. Min Act: -0.0001

Data Index: 1286548 (C4 (Web Text))

Max Activating Token Index: 719

Text #5

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.9189. Min Act: -0.0000

Data Index: 293123 (C4 (Web Text))

Max Activating Token Index: 818

Text #6

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.9646. Min Act: -0.0001

Data Index: 683042 (C4 (Web Text))

Max Activating Token Index: 37

Text #7

Max Range: 3.8440. Min Range: -3.8440

Max Act: 2.0712. Min Act: -0.0000

Data Index: 550223 (C4 (Web Text))

Max Activating Token Index: 641

Text #8

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.8129. Min Act: -0.0001

Data Index: 574396 (C4 (Web Text))

Max Activating Token Index: 320

Text #9

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.7646. Min Act: -0.0001

Data Index: 1203753 (C4 (Web Text))

Max Activating Token Index: 913

Text #10

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.7304. Min Act: -0.0001

Data Index: 653246 (C4 (Web Text))

Max Activating Token Index: 940

Text #11

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.8277. Min Act: -0.0000

Data Index: 1518538 (Python Code)

Max Activating Token Index: 691

Text #12

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.7230. Min Act: -0.0000

Data Index: 362147 (C4 (Web Text))

Max Activating Token Index: 212

Text #13

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.8251. Min Act: -0.0001

Data Index: 1438336 (Python Code)

Max Activating Token Index: 317

Text #14

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.8119. Min Act: -0.0000

Data Index: 1613138 (Python Code)

Max Activating Token Index: 404

Text #15

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.7218. Min Act: -0.0001

Data Index: 1455589 (Python Code)

Max Activating Token Index: 442

Text #16

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.8022. Min Act: -0.0000

Data Index: 1617984 (Python Code)

Max Activating Token Index: 440

Text #17

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.7959. Min Act: -0.0001

Data Index: 17015 (C4 (Web Text))

Max Activating Token Index: 727

Text #18

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.7863. Min Act: -0.0001

Data Index: 1219305 (C4 (Web Text))

Max Activating Token Index: 608

Text #19

Max Range: 3.8440. Min Range: -3.8440

Max Act: 1.8905. Min Act: -0.0001

Data Index: 1413998 (Python Code)

Max Activating Token Index: 226