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

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

Neuron 653 in Layer 7

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-8l')

Text #0

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5787. Min Act: -0.0001

Data Index: 289962 (C4 (Web Text))

Max Activating Token Index: 361

Text #1

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5783. Min Act: -0.0000

Data Index: 470140 (C4 (Web Text))

Max Activating Token Index: 338

Text #2

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5603. Min Act: -0.0000

Data Index: 983615 (C4 (Web Text))

Max Activating Token Index: 973

Text #3

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5523. Min Act: -0.0001

Data Index: 582482 (C4 (Web Text))

Max Activating Token Index: 388

Text #4

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5022. Min Act: -0.0000

Data Index: 419649 (C4 (Web Text))

Max Activating Token Index: 699

Text #5

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5093. Min Act: -0.0000

Data Index: 128101 (C4 (Web Text))

Max Activating Token Index: 182

Text #6

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4849. Min Act: -0.0000

Data Index: 683786 (C4 (Web Text))

Max Activating Token Index: 151

Text #7

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4783. Min Act: -0.0001

Data Index: 1233001 (C4 (Web Text))

Max Activating Token Index: 336

Text #8

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5294. Min Act: -0.0000

Data Index: 452581 (C4 (Web Text))

Max Activating Token Index: 993

Text #9

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5320. Min Act: -0.0000

Data Index: 65846 (C4 (Web Text))

Max Activating Token Index: 362

Text #10

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5004. Min Act: -0.0001

Data Index: 529217 (C4 (Web Text))

Max Activating Token Index: 383

Text #11

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4776. Min Act: -0.0000

Data Index: 839655 (C4 (Web Text))

Max Activating Token Index: 9

Text #12

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5257. Min Act: -0.0001

Data Index: 849158 (C4 (Web Text))

Max Activating Token Index: 781

Text #13

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4535. Min Act: -0.0001

Data Index: 698527 (C4 (Web Text))

Max Activating Token Index: 162

Text #14

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4672. Min Act: -0.0001

Data Index: 1269420 (C4 (Web Text))

Max Activating Token Index: 133

Text #15

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.5202. Min Act: -0.0001

Data Index: 1449844 (Python Code)

Max Activating Token Index: 89

Text #16

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4387. Min Act: -0.0001

Data Index: 372245 (C4 (Web Text))

Max Activating Token Index: 253

Text #17

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4026. Min Act: -0.0001

Data Index: 1349856 (C4 (Web Text))

Max Activating Token Index: 769

Text #18

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4630. Min Act: -0.0000

Data Index: 810499 (C4 (Web Text))

Max Activating Token Index: 692

Text #19

Max Range: 0.5787. Min Range: -0.5787

Max Act: 0.4565. Min Act: -0.0001

Data Index: 521052 (C4 (Web Text))

Max Activating Token Index: 144