Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

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

Neuron 4787 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.3926. Min Act: -0.0002

Data Index: 1179165 (C4 (Web Text))

Max Activating Token Index: 39

Text #1

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.3951. Min Act: -0.0002

Data Index: 1505141 (Python Code)

Max Activating Token Index: 738

Text #2

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2232. Min Act: -0.0002

Data Index: 511225 (C4 (Web Text))

Max Activating Token Index: 148

Text #3

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2703. Min Act: -0.0002

Data Index: 53374 (C4 (Web Text))

Max Activating Token Index: 509

Text #4

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.0369. Min Act: -0.0002

Data Index: 688477 (C4 (Web Text))

Max Activating Token Index: 323

Text #5

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.3244. Min Act: -0.0002

Data Index: 1586590 (Python Code)

Max Activating Token Index: 159

Text #6

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.3134. Min Act: -0.0002

Data Index: 1449056 (Python Code)

Max Activating Token Index: 301

Text #7

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2565. Min Act: -0.0002

Data Index: 877236 (C4 (Web Text))

Max Activating Token Index: 420

Text #8

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.1555. Min Act: -0.0002

Data Index: 700988 (C4 (Web Text))

Max Activating Token Index: 287

Text #9

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2484. Min Act: -0.0002

Data Index: 251971 (C4 (Web Text))

Max Activating Token Index: 32

Text #10

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.1892. Min Act: -0.0002

Data Index: 171663 (C4 (Web Text))

Max Activating Token Index: 719

Text #11

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.0358. Min Act: -0.0002

Data Index: 1479713 (Python Code)

Max Activating Token Index: 122

Text #12

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2473. Min Act: -0.0002

Data Index: 170054 (C4 (Web Text))

Max Activating Token Index: 911

Text #13

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.3220. Min Act: -0.0002

Data Index: 1610704 (Python Code)

Max Activating Token Index: 42

Text #14

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.1836. Min Act: -0.0002

Data Index: 706984 (C4 (Web Text))

Max Activating Token Index: 498

Text #15

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2130. Min Act: -0.0002

Data Index: 321596 (C4 (Web Text))

Max Activating Token Index: 9

Text #16

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.1809. Min Act: -0.0002

Data Index: 506389 (C4 (Web Text))

Max Activating Token Index: 492

Text #17

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.2714. Min Act: -0.0002

Data Index: 708 (C4 (Web Text))

Max Activating Token Index: 826

Text #18

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.0775. Min Act: -0.0002

Data Index: 1300308 (C4 (Web Text))

Max Activating Token Index: 841

Text #19

Max Range: 1.3951. Min Range: -1.3951

Max Act: 1.1672. Min Act: -0.0002

Data Index: 396894 (C4 (Web Text))

Max Activating Token Index: 142