Model: SoLU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 817 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-3l')

Text #0

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0260. Min Act: -0.0006

Data Index: 348475 (C4 (Web Text))

Max Activating Token Index: 869

Text #1

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0242. Min Act: -0.0006

Data Index: 1418062 (Python Code)

Max Activating Token Index: 403

Text #2

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0248. Min Act: -0.0006

Data Index: 1167124 (C4 (Web Text))

Max Activating Token Index: 21

Text #3

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0250. Min Act: -0.0006

Data Index: 1644596 (Python Code)

Max Activating Token Index: 560

Text #4

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0222. Min Act: -0.0006

Data Index: 97699 (C4 (Web Text))

Max Activating Token Index: 716

Text #5

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0222. Min Act: -0.0006

Data Index: 1035626 (C4 (Web Text))

Max Activating Token Index: 932

Text #6

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0228. Min Act: -0.0006

Data Index: 1576273 (Python Code)

Max Activating Token Index: 875

Text #7

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0217. Min Act: -0.0005

Data Index: 1539484 (Python Code)

Max Activating Token Index: 1023

Text #8

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0217. Min Act: -0.0006

Data Index: 317564 (C4 (Web Text))

Max Activating Token Index: 457

Text #9

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0226. Min Act: -0.0006

Data Index: 831889 (C4 (Web Text))

Max Activating Token Index: 1023

Text #10

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0211. Min Act: -0.0006

Data Index: 921383 (C4 (Web Text))

Max Activating Token Index: 692

Text #11

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0205. Min Act: -0.0006

Data Index: 808680 (C4 (Web Text))

Max Activating Token Index: 949

Text #12

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0209. Min Act: -0.0006

Data Index: 1549920 (Python Code)

Max Activating Token Index: 112

Text #13

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0199. Min Act: -0.0006

Data Index: 290567 (C4 (Web Text))

Max Activating Token Index: 572

Text #14

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0186. Min Act: -0.0005

Data Index: 1515428 (Python Code)

Max Activating Token Index: 936

Text #15

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0203. Min Act: -0.0006

Data Index: 858189 (C4 (Web Text))

Max Activating Token Index: 283

Text #16

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0202. Min Act: -0.0006

Data Index: 1166305 (C4 (Web Text))

Max Activating Token Index: 290

Text #17

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0202. Min Act: -0.0006

Data Index: 85216 (C4 (Web Text))

Max Activating Token Index: 738

Text #18

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0182. Min Act: -0.0006

Data Index: 392493 (C4 (Web Text))

Max Activating Token Index: 974

Text #19

Max Range: 0.0260. Min Range: -0.0260

Max Act: 0.0197. Min Act: -0.0006

Data Index: 1282814 (C4 (Web Text))

Max Activating Token Index: 921