Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

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

Neuron 70 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4963. Min Act: -0.0000

Data Index: 596218 (C4 (Web Text))

Max Activating Token Index: 652

Text #1

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4935. Min Act: -0.0001

Data Index: 1382169 (Python Code)

Max Activating Token Index: 682

Text #2

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4530. Min Act: -0.0001

Data Index: 1591497 (Python Code)

Max Activating Token Index: 909

Text #3

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4447. Min Act: -0.0001

Data Index: 1213449 (C4 (Web Text))

Max Activating Token Index: 599

Text #4

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4331. Min Act: -0.0001

Data Index: 196041 (C4 (Web Text))

Max Activating Token Index: 363

Text #5

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4611. Min Act: -0.0001

Data Index: 1587202 (Python Code)

Max Activating Token Index: 490

Text #6

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4452. Min Act: -0.0001

Data Index: 1499329 (Python Code)

Max Activating Token Index: 368

Text #7

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4176. Min Act: -0.0001

Data Index: 1145806 (C4 (Web Text))

Max Activating Token Index: 123

Text #8

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4438. Min Act: -0.0001

Data Index: 1393288 (Python Code)

Max Activating Token Index: 282

Text #9

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4536. Min Act: -0.0001

Data Index: 780004 (C4 (Web Text))

Max Activating Token Index: 180

Text #10

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4400. Min Act: -0.0001

Data Index: 388523 (C4 (Web Text))

Max Activating Token Index: 935

Text #11

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4399. Min Act: -0.0001

Data Index: 1213654 (C4 (Web Text))

Max Activating Token Index: 144

Text #12

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4516. Min Act: -0.0001

Data Index: 1579910 (Python Code)

Max Activating Token Index: 386

Text #13

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4378. Min Act: -0.0001

Data Index: 1429875 (Python Code)

Max Activating Token Index: 208

Text #14

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4341. Min Act: -0.0001

Data Index: 1224566 (C4 (Web Text))

Max Activating Token Index: 573

Text #15

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4277. Min Act: -0.0001

Data Index: 1093711 (C4 (Web Text))

Max Activating Token Index: 184

Text #16

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4397. Min Act: -0.0001

Data Index: 584831 (C4 (Web Text))

Max Activating Token Index: 197

Text #17

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4243. Min Act: -0.0001

Data Index: 334533 (C4 (Web Text))

Max Activating Token Index: 494

Text #18

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4510. Min Act: -0.0000

Data Index: 1125163 (C4 (Web Text))

Max Activating Token Index: 267

Text #19

Max Range: 0.4963. Min Range: -0.4963

Max Act: 0.4099. Min Act: -0.0001

Data Index: 109980 (C4 (Web Text))

Max Activating Token Index: 448