Model: SoLU Model: 4 Layers, 2048 Neurons per Layer

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

Neuron 453 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-4l')

Text #0

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0489. Min Act: -0.0002

Data Index: 1324396 (C4 (Web Text))

Max Activating Token Index: 1012

Text #1

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0465. Min Act: -0.0001

Data Index: 1298056 (C4 (Web Text))

Max Activating Token Index: 684

Text #2

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0504. Min Act: -0.0001

Data Index: 1167997 (C4 (Web Text))

Max Activating Token Index: 852

Text #3

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0462. Min Act: -0.0002

Data Index: 410103 (C4 (Web Text))

Max Activating Token Index: 596

Text #4

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0423. Min Act: -0.0001

Data Index: 53464 (C4 (Web Text))

Max Activating Token Index: 1016

Text #5

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0447. Min Act: -0.0001

Data Index: 1009001 (C4 (Web Text))

Max Activating Token Index: 550

Text #6

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0411. Min Act: -0.0001

Data Index: 790459 (C4 (Web Text))

Max Activating Token Index: 747

Text #7

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0404. Min Act: -0.0001

Data Index: 717053 (C4 (Web Text))

Max Activating Token Index: 143

Text #8

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0407. Min Act: -0.0001

Data Index: 449884 (C4 (Web Text))

Max Activating Token Index: 287

Text #9

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0348. Min Act: 0.0000

Data Index: 1394846 (Python Code)

Max Activating Token Index: 749

Text #10

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0309. Min Act: -0.0002

Data Index: 826238 (C4 (Web Text))

Max Activating Token Index: 676

Text #11

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0332. Min Act: -0.0001

Data Index: 1281131 (C4 (Web Text))

Max Activating Token Index: 649

Text #12

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0313. Min Act: -0.0001

Data Index: 610170 (C4 (Web Text))

Max Activating Token Index: 213

Text #13

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0325. Min Act: 0.0000

Data Index: 1584057 (Python Code)

Max Activating Token Index: 25

Text #14

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0341. Min Act: -0.0001

Data Index: 969177 (C4 (Web Text))

Max Activating Token Index: 899

Text #15

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0307. Min Act: -0.0001

Data Index: 55940 (C4 (Web Text))

Max Activating Token Index: 552

Text #16

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0310. Min Act: -0.0002

Data Index: 1233397 (C4 (Web Text))

Max Activating Token Index: 147

Text #17

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0293. Min Act: -0.0001

Data Index: 1279080 (C4 (Web Text))

Max Activating Token Index: 397

Text #18

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0281. Min Act: -0.0001

Data Index: 155787 (C4 (Web Text))

Max Activating Token Index: 722

Text #19

Max Range: 0.0504. Min Range: -0.0504

Max Act: 0.0289. Min Act: 0.0001

Data Index: 1439303 (Python Code)

Max Activating Token Index: 31