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

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

Neuron 3082 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: 0.3408. Min Range: -0.3408

Max Act: 0.3408. Min Act: -0.0001

Data Index: 154189 (C4 (Web Text))

Max Activating Token Index: 742

Text #1

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2672. Min Act: -0.0001

Data Index: 131863 (C4 (Web Text))

Max Activating Token Index: 510

Text #2

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2661. Min Act: -0.0001

Data Index: 503448 (C4 (Web Text))

Max Activating Token Index: 442

Text #3

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2644. Min Act: -0.0001

Data Index: 213516 (C4 (Web Text))

Max Activating Token Index: 457

Text #4

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2623. Min Act: -0.0001

Data Index: 470528 (C4 (Web Text))

Max Activating Token Index: 10

Text #5

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2583. Min Act: -0.0001

Data Index: 1393879 (Python Code)

Max Activating Token Index: 894

Text #6

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2428. Min Act: -0.0001

Data Index: 112392 (C4 (Web Text))

Max Activating Token Index: 378

Text #7

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2496. Min Act: -0.0001

Data Index: 926790 (C4 (Web Text))

Max Activating Token Index: 214

Text #8

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2480. Min Act: -0.0001

Data Index: 690028 (C4 (Web Text))

Max Activating Token Index: 700

Text #9

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2279. Min Act: -0.0001

Data Index: 12000 (C4 (Web Text))

Max Activating Token Index: 324

Text #10

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2152. Min Act: -0.0001

Data Index: 228549 (C4 (Web Text))

Max Activating Token Index: 206

Text #11

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2220. Min Act: -0.0001

Data Index: 604164 (C4 (Web Text))

Max Activating Token Index: 583

Text #12

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2201. Min Act: -0.0001

Data Index: 1002319 (C4 (Web Text))

Max Activating Token Index: 846

Text #13

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2206. Min Act: -0.0001

Data Index: 1644646 (Python Code)

Max Activating Token Index: 211

Text #14

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2177. Min Act: -0.0001

Data Index: 1046980 (C4 (Web Text))

Max Activating Token Index: 720

Text #15

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2142. Min Act: -0.0001

Data Index: 322873 (C4 (Web Text))

Max Activating Token Index: 699

Text #16

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2111. Min Act: -0.0001

Data Index: 734319 (C4 (Web Text))

Max Activating Token Index: 741

Text #17

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2099. Min Act: -0.0001

Data Index: 530959 (C4 (Web Text))

Max Activating Token Index: 249

Text #18

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.2035. Min Act: -0.0001

Data Index: 288238 (C4 (Web Text))

Max Activating Token Index: 852

Text #19

Max Range: 0.3408. Min Range: -0.3408

Max Act: 0.1935. Min Act: -0.0001

Data Index: 8900 (C4 (Web Text))

Max Activating Token Index: 566