Model: SoLU Model: 6 Layers, 3072 Neurons per Layer

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

Neuron 933 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-6l')

Text #0

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0545. Min Act: -0.0001

Data Index: 1404122 (Python Code)

Max Activating Token Index: 764

Text #1

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0560. Min Act: -0.0001

Data Index: 19307 (C4 (Web Text))

Max Activating Token Index: 29

Text #2

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0486. Min Act: -0.0001

Data Index: 1226555 (C4 (Web Text))

Max Activating Token Index: 223

Text #3

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0499. Min Act: -0.0001

Data Index: 143130 (C4 (Web Text))

Max Activating Token Index: 509

Text #4

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0493. Min Act: -0.0001

Data Index: 472735 (C4 (Web Text))

Max Activating Token Index: 166

Text #5

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0478. Min Act: -0.0001

Data Index: 632434 (C4 (Web Text))

Max Activating Token Index: 505

Text #6

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0519. Min Act: -0.0001

Data Index: 445502 (C4 (Web Text))

Max Activating Token Index: 647

Text #7

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0529. Min Act: -0.0001

Data Index: 189768 (C4 (Web Text))

Max Activating Token Index: 1022

Text #8

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0476. Min Act: -0.0001

Data Index: 771939 (C4 (Web Text))

Max Activating Token Index: 651

Text #9

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0461. Min Act: -0.0001

Data Index: 1416871 (Python Code)

Max Activating Token Index: 492

Text #10

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0463. Min Act: -0.0001

Data Index: 1336357 (C4 (Web Text))

Max Activating Token Index: 814

Text #11

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0418. Min Act: -0.0001

Data Index: 1092724 (C4 (Web Text))

Max Activating Token Index: 126

Text #12

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0418. Min Act: -0.0001

Data Index: 574566 (C4 (Web Text))

Max Activating Token Index: 64

Text #13

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0425. Min Act: -0.0000

Data Index: 356113 (C4 (Web Text))

Max Activating Token Index: 553

Text #14

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0433. Min Act: -0.0001

Data Index: 981874 (C4 (Web Text))

Max Activating Token Index: 418

Text #15

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0474. Min Act: -0.0001

Data Index: 663150 (C4 (Web Text))

Max Activating Token Index: 639

Text #16

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0435. Min Act: -0.0001

Data Index: 1570788 (Python Code)

Max Activating Token Index: 167

Text #17

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0430. Min Act: -0.0001

Data Index: 1169686 (C4 (Web Text))

Max Activating Token Index: 753

Text #18

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0449. Min Act: -0.0001

Data Index: 1118108 (C4 (Web Text))

Max Activating Token Index: 879

Text #19

Max Range: 0.0560. Min Range: -0.0560

Max Act: 0.0382. Min Act: -0.0001

Data Index: 279032 (C4 (Web Text))

Max Activating Token Index: 577