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

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

Neuron 650 in Layer 7

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.4351. Min Act: -0.0000

Data Index: 1200833 (C4 (Web Text))

Max Activating Token Index: 836

Text #1

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.4249. Min Act: -0.0000

Data Index: 475004 (C4 (Web Text))

Max Activating Token Index: 874

Text #2

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3480. Min Act: -0.0000

Data Index: 759215 (C4 (Web Text))

Max Activating Token Index: 534

Text #3

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3687. Min Act: -0.0000

Data Index: 803567 (C4 (Web Text))

Max Activating Token Index: 903

Text #4

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3684. Min Act: -0.0000

Data Index: 834762 (C4 (Web Text))

Max Activating Token Index: 184

Text #5

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3272. Min Act: -0.0000

Data Index: 188653 (C4 (Web Text))

Max Activating Token Index: 865

Text #6

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3448. Min Act: -0.0000

Data Index: 440268 (C4 (Web Text))

Max Activating Token Index: 517

Text #7

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3218. Min Act: -0.0000

Data Index: 1571365 (Python Code)

Max Activating Token Index: 890

Text #8

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3246. Min Act: -0.0000

Data Index: 1445276 (Python Code)

Max Activating Token Index: 980

Text #9

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3090. Min Act: -0.0000

Data Index: 1418601 (Python Code)

Max Activating Token Index: 536

Text #10

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2905. Min Act: -0.0000

Data Index: 1454893 (Python Code)

Max Activating Token Index: 280

Text #11

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2976. Min Act: -0.0000

Data Index: 1434693 (Python Code)

Max Activating Token Index: 341

Text #12

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2881. Min Act: -0.0000

Data Index: 562742 (C4 (Web Text))

Max Activating Token Index: 509

Text #13

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2902. Min Act: -0.0000

Data Index: 1373214 (Python Code)

Max Activating Token Index: 200

Text #14

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3169. Min Act: -0.0000

Data Index: 1378954 (Python Code)

Max Activating Token Index: 970

Text #15

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2855. Min Act: -0.0000

Data Index: 1466164 (Python Code)

Max Activating Token Index: 874

Text #16

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.3125. Min Act: -0.0000

Data Index: 1384980 (Python Code)

Max Activating Token Index: 220

Text #17

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2721. Min Act: -0.0000

Data Index: 1599360 (Python Code)

Max Activating Token Index: 546

Text #18

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2888. Min Act: -0.0000

Data Index: 1391375 (Python Code)

Max Activating Token Index: 483

Text #19

Max Range: 0.4351. Min Range: -0.4351

Max Act: 0.2872. Min Act: -0.0000

Data Index: 1524419 (Python Code)

Max Activating Token Index: 872