Model: SoLU Model: 6 Layers, 3072 Neurons per Layer

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

Neuron 2025 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-6l')

Text #0

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.1182. Min Act: -0.0001

Data Index: 261390 (C4 (Web Text))

Max Activating Token Index: 825

Text #1

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0970. Min Act: -0.0001

Data Index: 1488054 (Python Code)

Max Activating Token Index: 410

Text #2

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0907. Min Act: -0.0001

Data Index: 684718 (C4 (Web Text))

Max Activating Token Index: 873

Text #3

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0861. Min Act: -0.0001

Data Index: 88423 (C4 (Web Text))

Max Activating Token Index: 940

Text #4

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0896. Min Act: -0.0001

Data Index: 415650 (C4 (Web Text))

Max Activating Token Index: 39

Text #5

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0859. Min Act: -0.0001

Data Index: 89446 (C4 (Web Text))

Max Activating Token Index: 237

Text #6

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0853. Min Act: -0.0001

Data Index: 1091570 (C4 (Web Text))

Max Activating Token Index: 981

Text #7

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0891. Min Act: -0.0001

Data Index: 832527 (C4 (Web Text))

Max Activating Token Index: 648

Text #8

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0893. Min Act: -0.0001

Data Index: 570382 (C4 (Web Text))

Max Activating Token Index: 527

Text #9

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0821. Min Act: -0.0001

Data Index: 193527 (C4 (Web Text))

Max Activating Token Index: 507

Text #10

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0819. Min Act: -0.0001

Data Index: 333122 (C4 (Web Text))

Max Activating Token Index: 761

Text #11

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0829. Min Act: -0.0001

Data Index: 486379 (C4 (Web Text))

Max Activating Token Index: 903

Text #12

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0827. Min Act: -0.0001

Data Index: 979411 (C4 (Web Text))

Max Activating Token Index: 1002

Text #13

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0738. Min Act: -0.0001

Data Index: 1304981 (C4 (Web Text))

Max Activating Token Index: 799

Text #14

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0790. Min Act: -0.0001

Data Index: 1500223 (Python Code)

Max Activating Token Index: 109

Text #15

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0749. Min Act: -0.0001

Data Index: 1484014 (Python Code)

Max Activating Token Index: 25

Text #16

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0805. Min Act: -0.0001

Data Index: 442528 (C4 (Web Text))

Max Activating Token Index: 34

Text #17

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0796. Min Act: -0.0001

Data Index: 1034993 (C4 (Web Text))

Max Activating Token Index: 1009

Text #18

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0794. Min Act: -0.0001

Data Index: 625621 (C4 (Web Text))

Max Activating Token Index: 715

Text #19

Max Range: 0.1182. Min Range: -0.1182

Max Act: 0.0760. Min Act: -0.0000

Data Index: 1042513 (C4 (Web Text))

Max Activating Token Index: 915