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

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

Neuron 1724 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: 0.0688. Min Range: -0.0688

Max Act: 0.0688. Min Act: -0.0002

Data Index: 1512219 (Python Code)

Max Activating Token Index: 400

Text #1

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0640. Min Act: -0.0002

Data Index: 114786 (C4 (Web Text))

Max Activating Token Index: 922

Text #2

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0669. Min Act: -0.0002

Data Index: 1275038 (C4 (Web Text))

Max Activating Token Index: 666

Text #3

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0679. Min Act: -0.0002

Data Index: 629629 (C4 (Web Text))

Max Activating Token Index: 958

Text #4

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0681. Min Act: -0.0002

Data Index: 1489607 (Python Code)

Max Activating Token Index: 611

Text #5

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0670. Min Act: -0.0002

Data Index: 323159 (C4 (Web Text))

Max Activating Token Index: 777

Text #6

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0643. Min Act: -0.0002

Data Index: 710288 (C4 (Web Text))

Max Activating Token Index: 963

Text #7

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0666. Min Act: -0.0002

Data Index: 324328 (C4 (Web Text))

Max Activating Token Index: 106

Text #8

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0655. Min Act: -0.0002

Data Index: 438971 (C4 (Web Text))

Max Activating Token Index: 227

Text #9

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0628. Min Act: -0.0002

Data Index: 1186022 (C4 (Web Text))

Max Activating Token Index: 808

Text #10

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0641. Min Act: -0.0002

Data Index: 180302 (C4 (Web Text))

Max Activating Token Index: 970

Text #11

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0621. Min Act: -0.0002

Data Index: 284134 (C4 (Web Text))

Max Activating Token Index: 580

Text #12

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0644. Min Act: -0.0002

Data Index: 1600849 (Python Code)

Max Activating Token Index: 909

Text #13

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0661. Min Act: -0.0002

Data Index: 1339206 (C4 (Web Text))

Max Activating Token Index: 262

Text #14

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0668. Min Act: -0.0002

Data Index: 1184179 (C4 (Web Text))

Max Activating Token Index: 361

Text #15

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0629. Min Act: -0.0002

Data Index: 1645711 (Python Code)

Max Activating Token Index: 373

Text #16

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0639. Min Act: -0.0002

Data Index: 260038 (C4 (Web Text))

Max Activating Token Index: 140

Text #17

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0614. Min Act: -0.0002

Data Index: 486046 (C4 (Web Text))

Max Activating Token Index: 185

Text #18

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0650. Min Act: -0.0002

Data Index: 904160 (C4 (Web Text))

Max Activating Token Index: 648

Text #19

Max Range: 0.0688. Min Range: -0.0688

Max Act: 0.0656. Min Act: -0.0001

Data Index: 1075451 (C4 (Web Text))

Max Activating Token Index: 1004