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

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

Neuron 1640 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0308. Min Act: -0.0003

Data Index: 974629 (C4 (Web Text))

Max Activating Token Index: 395

Text #1

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0251. Min Act: -0.0002

Data Index: 960040 (C4 (Web Text))

Max Activating Token Index: 96

Text #2

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0221. Min Act: -0.0003

Data Index: 109693 (C4 (Web Text))

Max Activating Token Index: 530

Text #3

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0219. Min Act: -0.0002

Data Index: 552671 (C4 (Web Text))

Max Activating Token Index: 601

Text #4

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0220. Min Act: -0.0003

Data Index: 683217 (C4 (Web Text))

Max Activating Token Index: 239

Text #5

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0215. Min Act: -0.0003

Data Index: 575587 (C4 (Web Text))

Max Activating Token Index: 935

Text #6

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0228. Min Act: -0.0003

Data Index: 875273 (C4 (Web Text))

Max Activating Token Index: 535

Text #7

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0212. Min Act: -0.0002

Data Index: 931604 (C4 (Web Text))

Max Activating Token Index: 141

Text #8

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0203. Min Act: -0.0002

Data Index: 540622 (C4 (Web Text))

Max Activating Token Index: 756

Text #9

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0210. Min Act: -0.0002

Data Index: 1078202 (C4 (Web Text))

Max Activating Token Index: 294

Text #10

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0200. Min Act: -0.0003

Data Index: 493007 (C4 (Web Text))

Max Activating Token Index: 318

Text #11

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0222. Min Act: -0.0002

Data Index: 437136 (C4 (Web Text))

Max Activating Token Index: 127

Text #12

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0209. Min Act: -0.0002

Data Index: 338182 (C4 (Web Text))

Max Activating Token Index: 212

Text #13

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0218. Min Act: -0.0003

Data Index: 1034932 (C4 (Web Text))

Max Activating Token Index: 114

Text #14

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0210. Min Act: -0.0003

Data Index: 1564631 (Python Code)

Max Activating Token Index: 723

Text #15

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0208. Min Act: -0.0003

Data Index: 1157703 (C4 (Web Text))

Max Activating Token Index: 558

Text #16

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0217. Min Act: -0.0002

Data Index: 89135 (C4 (Web Text))

Max Activating Token Index: 360

Text #17

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0205. Min Act: -0.0003

Data Index: 220275 (C4 (Web Text))

Max Activating Token Index: 229

Text #18

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0206. Min Act: -0.0003

Data Index: 453726 (C4 (Web Text))

Max Activating Token Index: 60

Text #19

Max Range: 0.0308. Min Range: -0.0308

Max Act: 0.0191. Min Act: -0.0002

Data Index: 1072522 (C4 (Web Text))

Max Activating Token Index: 531