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

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

Neuron 1397 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0640. Min Act: -0.0005

Data Index: 872525 (C4 (Web Text))

Max Activating Token Index: 841

Text #1

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0322. Min Act: -0.0005

Data Index: 1187822 (C4 (Web Text))

Max Activating Token Index: 203

Text #2

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0212. Min Act: -0.0005

Data Index: 61073 (C4 (Web Text))

Max Activating Token Index: 736

Text #3

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0242. Min Act: -0.0005

Data Index: 1493388 (Python Code)

Max Activating Token Index: 7

Text #4

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0221. Min Act: -0.0005

Data Index: 685042 (C4 (Web Text))

Max Activating Token Index: 771

Text #5

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0214. Min Act: -0.0005

Data Index: 323549 (C4 (Web Text))

Max Activating Token Index: 753

Text #6

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0261. Min Act: -0.0005

Data Index: 345611 (C4 (Web Text))

Max Activating Token Index: 917

Text #7

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0190. Min Act: -0.0005

Data Index: 1489775 (Python Code)

Max Activating Token Index: 3

Text #8

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0224. Min Act: -0.0005

Data Index: 1334662 (C4 (Web Text))

Max Activating Token Index: 815

Text #9

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0185. Min Act: -0.0005

Data Index: 1110188 (C4 (Web Text))

Max Activating Token Index: 300

Text #10

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0176. Min Act: -0.0005

Data Index: 1296234 (C4 (Web Text))

Max Activating Token Index: 334

Text #11

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0168. Min Act: -0.0005

Data Index: 841751 (C4 (Web Text))

Max Activating Token Index: 901

Text #12

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0159. Min Act: -0.0005

Data Index: 609725 (C4 (Web Text))

Max Activating Token Index: 829

Text #13

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0159. Min Act: -0.0005

Data Index: 1533536 (Python Code)

Max Activating Token Index: 44

Text #14

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0189. Min Act: -0.0005

Data Index: 44956 (C4 (Web Text))

Max Activating Token Index: 552

Text #15

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0148. Min Act: -0.0005

Data Index: 1576111 (Python Code)

Max Activating Token Index: 11

Text #16

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0146. Min Act: -0.0005

Data Index: 275813 (C4 (Web Text))

Max Activating Token Index: 578

Text #17

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0136. Min Act: -0.0005

Data Index: 470500 (C4 (Web Text))

Max Activating Token Index: 517

Text #18

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0127. Min Act: -0.0005

Data Index: 1644464 (Python Code)

Max Activating Token Index: 10

Text #19

Max Range: 0.0640. Min Range: -0.0640

Max Act: 0.0138. Min Act: -0.0005

Data Index: 611675 (C4 (Web Text))

Max Activating Token Index: 9