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

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

Neuron 102 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: 0.0852. Min Range: -0.0852

Max Act: 0.0852. Min Act: -0.0001

Data Index: 389718 (C4 (Web Text))

Max Activating Token Index: 265

Text #1

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0727. Min Act: -0.0001

Data Index: 1292863 (C4 (Web Text))

Max Activating Token Index: 894

Text #2

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0723. Min Act: -0.0001

Data Index: 598997 (C4 (Web Text))

Max Activating Token Index: 993

Text #3

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0698. Min Act: -0.0001

Data Index: 279431 (C4 (Web Text))

Max Activating Token Index: 448

Text #4

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0691. Min Act: -0.0001

Data Index: 769955 (C4 (Web Text))

Max Activating Token Index: 548

Text #5

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0687. Min Act: -0.0001

Data Index: 887853 (C4 (Web Text))

Max Activating Token Index: 205

Text #6

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0647. Min Act: -0.0001

Data Index: 884285 (C4 (Web Text))

Max Activating Token Index: 376

Text #7

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0648. Min Act: -0.0000

Data Index: 158924 (C4 (Web Text))

Max Activating Token Index: 438

Text #8

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0627. Min Act: -0.0001

Data Index: 247717 (C4 (Web Text))

Max Activating Token Index: 234

Text #9

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0599. Min Act: -0.0001

Data Index: 1592999 (Python Code)

Max Activating Token Index: 876

Text #10

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0617. Min Act: -0.0001

Data Index: 729591 (C4 (Web Text))

Max Activating Token Index: 987

Text #11

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0598. Min Act: -0.0000

Data Index: 546088 (C4 (Web Text))

Max Activating Token Index: 348

Text #12

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0592. Min Act: -0.0001

Data Index: 1494130 (Python Code)

Max Activating Token Index: 306

Text #13

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0615. Min Act: -0.0001

Data Index: 148750 (C4 (Web Text))

Max Activating Token Index: 642

Text #14

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0609. Min Act: -0.0000

Data Index: 894763 (C4 (Web Text))

Max Activating Token Index: 274

Text #15

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0585. Min Act: -0.0000

Data Index: 305027 (C4 (Web Text))

Max Activating Token Index: 632

Text #16

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0622. Min Act: 0.0000

Data Index: 220074 (C4 (Web Text))

Max Activating Token Index: 513

Text #17

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0602. Min Act: -0.0001

Data Index: 1614235 (Python Code)

Max Activating Token Index: 711

Text #18

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0582. Min Act: -0.0001

Data Index: 1109216 (C4 (Web Text))

Max Activating Token Index: 339

Text #19

Max Range: 0.0852. Min Range: -0.0852

Max Act: 0.0610. Min Act: -0.0002

Data Index: 575863 (C4 (Web Text))

Max Activating Token Index: 225