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

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

Neuron 127 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.7674. Min Act: -0.0001

Data Index: 1361179 (Python Code)

Max Activating Token Index: 3

Text #1

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.7037. Min Act: -0.0001

Data Index: 1294321 (C4 (Web Text))

Max Activating Token Index: 290

Text #2

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.6573. Min Act: -0.0001

Data Index: 1254075 (C4 (Web Text))

Max Activating Token Index: 21

Text #3

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.6584. Min Act: -0.0001

Data Index: 855014 (C4 (Web Text))

Max Activating Token Index: 25

Text #4

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.6689. Min Act: -0.0001

Data Index: 1155261 (C4 (Web Text))

Max Activating Token Index: 887

Text #5

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.6349. Min Act: -0.0001

Data Index: 261879 (C4 (Web Text))

Max Activating Token Index: 399

Text #6

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.6540. Min Act: -0.0001

Data Index: 1456564 (Python Code)

Max Activating Token Index: 12

Text #7

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.6163. Min Act: -0.0001

Data Index: 1552485 (Python Code)

Max Activating Token Index: 61

Text #8

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5537. Min Act: -0.0001

Data Index: 1607269 (Python Code)

Max Activating Token Index: 105

Text #9

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5936. Min Act: -0.0001

Data Index: 1102855 (C4 (Web Text))

Max Activating Token Index: 42

Text #10

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5810. Min Act: -0.0001

Data Index: 907643 (C4 (Web Text))

Max Activating Token Index: 422

Text #11

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5702. Min Act: -0.0001

Data Index: 95467 (C4 (Web Text))

Max Activating Token Index: 274

Text #12

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5487. Min Act: -0.0001

Data Index: 1429376 (Python Code)

Max Activating Token Index: 44

Text #13

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5597. Min Act: -0.0001

Data Index: 458605 (C4 (Web Text))

Max Activating Token Index: 487

Text #14

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5359. Min Act: -0.0001

Data Index: 1464118 (Python Code)

Max Activating Token Index: 514

Text #15

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.4881. Min Act: -0.0001

Data Index: 425743 (C4 (Web Text))

Max Activating Token Index: 853

Text #16

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5322. Min Act: -0.0001

Data Index: 743612 (C4 (Web Text))

Max Activating Token Index: 417

Text #17

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5285. Min Act: -0.0001

Data Index: 1159818 (C4 (Web Text))

Max Activating Token Index: 650

Text #18

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5225. Min Act: -0.0001

Data Index: 738193 (C4 (Web Text))

Max Activating Token Index: 131

Text #19

Max Range: 0.7674. Min Range: -0.7674

Max Act: 0.5493. Min Act: -0.0001

Data Index: 798674 (C4 (Web Text))

Max Activating Token Index: 920