Model: SoLU Model: 12 Layers, 6144 Neurons per Layer

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

Neuron 3904 in Layer 7

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-12l')

Text #0

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.5872. Min Act: -0.0000

Data Index: 837915 (C4 (Web Text))

Max Activating Token Index: 109

Text #1

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.5192. Min Act: -0.0000

Data Index: 589315 (C4 (Web Text))

Max Activating Token Index: 791

Text #2

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.5024. Min Act: -0.0000

Data Index: 664379 (C4 (Web Text))

Max Activating Token Index: 995

Text #3

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4896. Min Act: -0.0000

Data Index: 1379711 (Python Code)

Max Activating Token Index: 371

Text #4

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4574. Min Act: -0.0000

Data Index: 47976 (C4 (Web Text))

Max Activating Token Index: 915

Text #5

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4687. Min Act: -0.0000

Data Index: 161908 (C4 (Web Text))

Max Activating Token Index: 964

Text #6

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4821. Min Act: -0.0000

Data Index: 336809 (C4 (Web Text))

Max Activating Token Index: 706

Text #7

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4996. Min Act: -0.0000

Data Index: 889563 (C4 (Web Text))

Max Activating Token Index: 93

Text #8

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4790. Min Act: -0.0000

Data Index: 1563097 (Python Code)

Max Activating Token Index: 140

Text #9

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4877. Min Act: -0.0000

Data Index: 145382 (C4 (Web Text))

Max Activating Token Index: 913

Text #10

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4625. Min Act: -0.0000

Data Index: 1336362 (C4 (Web Text))

Max Activating Token Index: 333

Text #11

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4910. Min Act: -0.0000

Data Index: 69372 (C4 (Web Text))

Max Activating Token Index: 262

Text #12

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4570. Min Act: -0.0000

Data Index: 392132 (C4 (Web Text))

Max Activating Token Index: 986

Text #13

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4646. Min Act: -0.0000

Data Index: 244344 (C4 (Web Text))

Max Activating Token Index: 999

Text #14

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4349. Min Act: -0.0000

Data Index: 824727 (C4 (Web Text))

Max Activating Token Index: 664

Text #15

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4416. Min Act: -0.0000

Data Index: 1332166 (C4 (Web Text))

Max Activating Token Index: 600

Text #16

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4600. Min Act: -0.0000

Data Index: 1560979 (Python Code)

Max Activating Token Index: 807

Text #17

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.4063. Min Act: -0.0000

Data Index: 51560 (C4 (Web Text))

Max Activating Token Index: 929

Text #18

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.3903. Min Act: -0.0000

Data Index: 225576 (C4 (Web Text))

Max Activating Token Index: 534

Text #19

Max Range: 0.5872. Min Range: -0.5872

Max Act: 0.3907. Min Act: -0.0000

Data Index: 754281 (C4 (Web Text))

Max Activating Token Index: 280