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

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

Neuron 3885 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: 0.0847. Min Range: -0.0847

Max Act: 0.0806. Min Act: -0.0000

Data Index: 138283 (C4 (Web Text))

Max Activating Token Index: 46

Text #1

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0847. Min Act: -0.0001

Data Index: 1155613 (C4 (Web Text))

Max Activating Token Index: 898

Text #2

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0806. Min Act: -0.0001

Data Index: 73805 (C4 (Web Text))

Max Activating Token Index: 146

Text #3

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0843. Min Act: -0.0001

Data Index: 698325 (C4 (Web Text))

Max Activating Token Index: 623

Text #4

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0837. Min Act: -0.0000

Data Index: 491699 (C4 (Web Text))

Max Activating Token Index: 993

Text #5

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0842. Min Act: -0.0001

Data Index: 1566428 (Python Code)

Max Activating Token Index: 273

Text #6

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0792. Min Act: -0.0001

Data Index: 856345 (C4 (Web Text))

Max Activating Token Index: 6

Text #7

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0720. Min Act: -0.0001

Data Index: 1576718 (Python Code)

Max Activating Token Index: 96

Text #8

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0689. Min Act: -0.0001

Data Index: 900901 (C4 (Web Text))

Max Activating Token Index: 892

Text #9

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0610. Min Act: -0.0001

Data Index: 340874 (C4 (Web Text))

Max Activating Token Index: 311

Text #10

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0634. Min Act: -0.0001

Data Index: 527200 (C4 (Web Text))

Max Activating Token Index: 170

Text #11

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0669. Min Act: -0.0001

Data Index: 1481176 (Python Code)

Max Activating Token Index: 990

Text #12

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0671. Min Act: -0.0000

Data Index: 491642 (C4 (Web Text))

Max Activating Token Index: 74

Text #13

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0674. Min Act: -0.0000

Data Index: 853490 (C4 (Web Text))

Max Activating Token Index: 609

Text #14

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0672. Min Act: -0.0001

Data Index: 1478294 (Python Code)

Max Activating Token Index: 348

Text #15

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0619. Min Act: -0.0001

Data Index: 814832 (C4 (Web Text))

Max Activating Token Index: 665

Text #16

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0669. Min Act: -0.0000

Data Index: 1388333 (Python Code)

Max Activating Token Index: 85

Text #17

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0471. Min Act: -0.0001

Data Index: 1243666 (C4 (Web Text))

Max Activating Token Index: 614

Text #18

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0700. Min Act: -0.0000

Data Index: 36822 (C4 (Web Text))

Max Activating Token Index: 757

Text #19

Max Range: 0.0847. Min Range: -0.0847

Max Act: 0.0574. Min Act: -0.0001

Data Index: 869524 (C4 (Web Text))

Max Activating Token Index: 560