Model: SoLU Model: 4 Layers, 2048 Neurons per Layer

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

Neuron 768 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-4l')

Text #0

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0235. Min Act: -0.0006

Data Index: 1448028 (Python Code)

Max Activating Token Index: 964

Text #1

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0219. Min Act: -0.0006

Data Index: 1537435 (Python Code)

Max Activating Token Index: 930

Text #2

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0220. Min Act: -0.0006

Data Index: 1292339 (C4 (Web Text))

Max Activating Token Index: 194

Text #3

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0209. Min Act: -0.0006

Data Index: 844165 (C4 (Web Text))

Max Activating Token Index: 544

Text #4

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0190. Min Act: -0.0006

Data Index: 1630477 (Python Code)

Max Activating Token Index: 325

Text #5

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0198. Min Act: -0.0006

Data Index: 647359 (C4 (Web Text))

Max Activating Token Index: 1010

Text #6

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0196. Min Act: -0.0006

Data Index: 1410856 (Python Code)

Max Activating Token Index: 978

Text #7

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0200. Min Act: -0.0006

Data Index: 805959 (C4 (Web Text))

Max Activating Token Index: 202

Text #8

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0189. Min Act: -0.0006

Data Index: 1069631 (C4 (Web Text))

Max Activating Token Index: 407

Text #9

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0206. Min Act: -0.0006

Data Index: 593184 (C4 (Web Text))

Max Activating Token Index: 647

Text #10

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0197. Min Act: -0.0006

Data Index: 576748 (C4 (Web Text))

Max Activating Token Index: 852

Text #11

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0191. Min Act: -0.0006

Data Index: 1423362 (Python Code)

Max Activating Token Index: 528

Text #12

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0205. Min Act: -0.0006

Data Index: 1543573 (Python Code)

Max Activating Token Index: 417

Text #13

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0191. Min Act: -0.0006

Data Index: 1564594 (Python Code)

Max Activating Token Index: 337

Text #14

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0182. Min Act: -0.0006

Data Index: 1544561 (Python Code)

Max Activating Token Index: 87

Text #15

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0191. Min Act: -0.0006

Data Index: 1636411 (Python Code)

Max Activating Token Index: 738

Text #16

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0200. Min Act: -0.0006

Data Index: 1528514 (Python Code)

Max Activating Token Index: 63

Text #17

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0187. Min Act: -0.0006

Data Index: 1583079 (Python Code)

Max Activating Token Index: 580

Text #18

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0184. Min Act: -0.0006

Data Index: 374939 (C4 (Web Text))

Max Activating Token Index: 668

Text #19

Max Range: 0.0235. Min Range: -0.0235

Max Act: 0.0198. Min Act: -0.0006

Data Index: 1567603 (Python Code)

Max Activating Token Index: 574