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

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

Neuron 72 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: 0.0621. Min Range: -0.0621

Max Act: 0.0621. Min Act: -0.0002

Data Index: 1241618 (C4 (Web Text))

Max Activating Token Index: 204

Text #1

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0606. Min Act: -0.0002

Data Index: 1427809 (Python Code)

Max Activating Token Index: 197

Text #2

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0578. Min Act: -0.0001

Data Index: 1212323 (C4 (Web Text))

Max Activating Token Index: 293

Text #3

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0548. Min Act: -0.0002

Data Index: 1100568 (C4 (Web Text))

Max Activating Token Index: 544

Text #4

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0579. Min Act: -0.0002

Data Index: 985214 (C4 (Web Text))

Max Activating Token Index: 522

Text #5

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0546. Min Act: -0.0002

Data Index: 361026 (C4 (Web Text))

Max Activating Token Index: 794

Text #6

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0560. Min Act: -0.0002

Data Index: 679611 (C4 (Web Text))

Max Activating Token Index: 576

Text #7

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0530. Min Act: -0.0002

Data Index: 745343 (C4 (Web Text))

Max Activating Token Index: 386

Text #8

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0536. Min Act: -0.0002

Data Index: 345416 (C4 (Web Text))

Max Activating Token Index: 540

Text #9

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0516. Min Act: -0.0002

Data Index: 847987 (C4 (Web Text))

Max Activating Token Index: 298

Text #10

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0553. Min Act: -0.0002

Data Index: 1647176 (Python Code)

Max Activating Token Index: 128

Text #11

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0496. Min Act: -0.0002

Data Index: 1475653 (Python Code)

Max Activating Token Index: 305

Text #12

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0518. Min Act: -0.0002

Data Index: 1114974 (C4 (Web Text))

Max Activating Token Index: 453

Text #13

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0557. Min Act: -0.0002

Data Index: 859863 (C4 (Web Text))

Max Activating Token Index: 312

Text #14

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0512. Min Act: -0.0002

Data Index: 1044486 (C4 (Web Text))

Max Activating Token Index: 104

Text #15

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0513. Min Act: -0.0002

Data Index: 1464391 (Python Code)

Max Activating Token Index: 219

Text #16

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0502. Min Act: -0.0002

Data Index: 581236 (C4 (Web Text))

Max Activating Token Index: 545

Text #17

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0550. Min Act: -0.0002

Data Index: 297313 (C4 (Web Text))

Max Activating Token Index: 420

Text #18

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0531. Min Act: -0.0002

Data Index: 1328796 (C4 (Web Text))

Max Activating Token Index: 583

Text #19

Max Range: 0.0621. Min Range: -0.0621

Max Act: 0.0509. Min Act: -0.0002

Data Index: 5889 (C4 (Web Text))

Max Activating Token Index: 83