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

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

Neuron 20 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 1.1038. Min Range: -1.1038

Max Act: 1.0663. Min Act: -0.0001

Data Index: 620827 (C4 (Web Text))

Max Activating Token Index: 357

Text #1

Max Range: 1.1038. Min Range: -1.1038

Max Act: 1.1038. Min Act: -0.0001

Data Index: 456400 (C4 (Web Text))

Max Activating Token Index: 432

Text #2

Max Range: 1.1038. Min Range: -1.1038

Max Act: 1.0547. Min Act: -0.0001

Data Index: 1410848 (Python Code)

Max Activating Token Index: 371

Text #3

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.9877. Min Act: -0.0001

Data Index: 1045944 (C4 (Web Text))

Max Activating Token Index: 949

Text #4

Max Range: 1.1038. Min Range: -1.1038

Max Act: 1.0247. Min Act: -0.0001

Data Index: 742829 (C4 (Web Text))

Max Activating Token Index: 559

Text #5

Max Range: 1.1038. Min Range: -1.1038

Max Act: 1.0174. Min Act: -0.0001

Data Index: 1040320 (C4 (Web Text))

Max Activating Token Index: 835

Text #6

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.9703. Min Act: -0.0001

Data Index: 611907 (C4 (Web Text))

Max Activating Token Index: 368

Text #7

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.9340. Min Act: -0.0001

Data Index: 862997 (C4 (Web Text))

Max Activating Token Index: 326

Text #8

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.9838. Min Act: -0.0001

Data Index: 1394378 (Python Code)

Max Activating Token Index: 997

Text #9

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.9141. Min Act: -0.0001

Data Index: 1361477 (Python Code)

Max Activating Token Index: 258

Text #10

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8853. Min Act: -0.0001

Data Index: 281401 (C4 (Web Text))

Max Activating Token Index: 902

Text #11

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8520. Min Act: -0.0001

Data Index: 1441286 (Python Code)

Max Activating Token Index: 673

Text #12

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8727. Min Act: -0.0001

Data Index: 9839 (C4 (Web Text))

Max Activating Token Index: 873

Text #13

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8733. Min Act: -0.0001

Data Index: 739805 (C4 (Web Text))

Max Activating Token Index: 959

Text #14

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8836. Min Act: -0.0001

Data Index: 100069 (C4 (Web Text))

Max Activating Token Index: 390

Text #15

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8723. Min Act: -0.0001

Data Index: 915948 (C4 (Web Text))

Max Activating Token Index: 415

Text #16

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8156. Min Act: -0.0001

Data Index: 42659 (C4 (Web Text))

Max Activating Token Index: 365

Text #17

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8375. Min Act: -0.0001

Data Index: 830558 (C4 (Web Text))

Max Activating Token Index: 921

Text #18

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8948. Min Act: -0.0001

Data Index: 151465 (C4 (Web Text))

Max Activating Token Index: 812

Text #19

Max Range: 1.1038. Min Range: -1.1038

Max Act: 0.8917. Min Act: -0.0001

Data Index: 43328 (C4 (Web Text))

Max Activating Token Index: 1007