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

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

Neuron 3706 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1639. Min Act: -0.0001

Data Index: 939 (C4 (Web Text))

Max Activating Token Index: 334

Text #1

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1532. Min Act: -0.0001

Data Index: 336592 (C4 (Web Text))

Max Activating Token Index: 411

Text #2

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1563. Min Act: -0.0001

Data Index: 1141671 (C4 (Web Text))

Max Activating Token Index: 424

Text #3

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1485. Min Act: -0.0001

Data Index: 141358 (C4 (Web Text))

Max Activating Token Index: 902

Text #4

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1371. Min Act: -0.0001

Data Index: 492714 (C4 (Web Text))

Max Activating Token Index: 310

Text #5

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1384. Min Act: -0.0001

Data Index: 1653765 (Python Code)

Max Activating Token Index: 431

Text #6

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1313. Min Act: -0.0001

Data Index: 562106 (C4 (Web Text))

Max Activating Token Index: 13

Text #7

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1358. Min Act: -0.0001

Data Index: 1126146 (C4 (Web Text))

Max Activating Token Index: 193

Text #8

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1302. Min Act: -0.0001

Data Index: 439490 (C4 (Web Text))

Max Activating Token Index: 405

Text #9

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1308. Min Act: -0.0001

Data Index: 892627 (C4 (Web Text))

Max Activating Token Index: 778

Text #10

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1303. Min Act: -0.0001

Data Index: 121114 (C4 (Web Text))

Max Activating Token Index: 224

Text #11

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1286. Min Act: -0.0001

Data Index: 1419085 (Python Code)

Max Activating Token Index: 367

Text #12

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1335. Min Act: -0.0001

Data Index: 169865 (C4 (Web Text))

Max Activating Token Index: 676

Text #13

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1141. Min Act: -0.0001

Data Index: 1473110 (Python Code)

Max Activating Token Index: 623

Text #14

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1266. Min Act: -0.0001

Data Index: 473331 (C4 (Web Text))

Max Activating Token Index: 512

Text #15

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1101. Min Act: -0.0001

Data Index: 1551723 (Python Code)

Max Activating Token Index: 333

Text #16

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1171. Min Act: -0.0001

Data Index: 117123 (C4 (Web Text))

Max Activating Token Index: 231

Text #17

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1246. Min Act: -0.0001

Data Index: 1373999 (Python Code)

Max Activating Token Index: 222

Text #18

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1192. Min Act: -0.0001

Data Index: 475043 (C4 (Web Text))

Max Activating Token Index: 1011

Text #19

Max Range: 0.1639. Min Range: -0.1639

Max Act: 0.1245. Min Act: -0.0001

Data Index: 1439951 (Python Code)

Max Activating Token Index: 789