Model: SoLU Model: 8 Layers, 4096 Neurons per Layer

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

Neuron 3472 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-8l')

Text #0

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.3481. Min Act: -0.0001

Data Index: 221395 (C4 (Web Text))

Max Activating Token Index: 311

Text #1

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.4965. Min Act: -0.0001

Data Index: 598304 (C4 (Web Text))

Max Activating Token Index: 170

Text #2

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.3626. Min Act: -0.0001

Data Index: 1065151 (C4 (Web Text))

Max Activating Token Index: 338

Text #3

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.3110. Min Act: -0.0001

Data Index: 1569224 (Python Code)

Max Activating Token Index: 263

Text #4

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.3113. Min Act: -0.0001

Data Index: 836514 (C4 (Web Text))

Max Activating Token Index: 608

Text #5

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2864. Min Act: -0.0002

Data Index: 316869 (C4 (Web Text))

Max Activating Token Index: 783

Text #6

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.3201. Min Act: -0.0002

Data Index: 355068 (C4 (Web Text))

Max Activating Token Index: 1017

Text #7

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2843. Min Act: -0.0001

Data Index: 1435243 (Python Code)

Max Activating Token Index: 298

Text #8

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2876. Min Act: -0.0001

Data Index: 408403 (C4 (Web Text))

Max Activating Token Index: 658

Text #9

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.3021. Min Act: -0.0001

Data Index: 455550 (C4 (Web Text))

Max Activating Token Index: 328

Text #10

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2447. Min Act: -0.0001

Data Index: 1594556 (Python Code)

Max Activating Token Index: 749

Text #11

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2816. Min Act: -0.0001

Data Index: 898649 (C4 (Web Text))

Max Activating Token Index: 399

Text #12

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2691. Min Act: -0.0001

Data Index: 383831 (C4 (Web Text))

Max Activating Token Index: 597

Text #13

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2773. Min Act: -0.0001

Data Index: 936513 (C4 (Web Text))

Max Activating Token Index: 455

Text #14

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2850. Min Act: -0.0001

Data Index: 396148 (C4 (Web Text))

Max Activating Token Index: 951

Text #15

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2708. Min Act: -0.0001

Data Index: 1552143 (Python Code)

Max Activating Token Index: 956

Text #16

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2601. Min Act: -0.0001

Data Index: 407149 (C4 (Web Text))

Max Activating Token Index: 975

Text #17

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2796. Min Act: -0.0001

Data Index: 656516 (C4 (Web Text))

Max Activating Token Index: 166

Text #18

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2574. Min Act: -0.0001

Data Index: 151128 (C4 (Web Text))

Max Activating Token Index: 286

Text #19

Max Range: 0.4965. Min Range: -0.4965

Max Act: 0.2063. Min Act: -0.0001

Data Index: 579693 (C4 (Web Text))

Max Activating Token Index: 545