Model: GELU Model: 4 Layers, 2048 Neurons per Layer

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

Neuron 975 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')

Text #0

Max Range: 5.0558. Min Range: -5.0558

Max Act: 5.0558. Min Act: -0.1700

Data Index: 1334676 (C4 (Web Text))

Max Activating Token Index: 965

Text #1

Max Range: 5.0558. Min Range: -5.0558

Max Act: 5.0558. Min Act: -0.1700

Data Index: 775886 (C4 (Web Text))

Max Activating Token Index: 991

Text #2

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8683. Min Act: -0.1700

Data Index: 394580 (C4 (Web Text))

Max Activating Token Index: 710

Text #3

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8683. Min Act: -0.1700

Data Index: 549046 (C4 (Web Text))

Max Activating Token Index: 437

Text #4

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8683. Min Act: -0.1700

Data Index: 1374700 (Python Code)

Max Activating Token Index: 375

Text #5

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8370. Min Act: -0.1700

Data Index: 511106 (C4 (Web Text))

Max Activating Token Index: 230

Text #6

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8058. Min Act: -0.1700

Data Index: 1374146 (Python Code)

Max Activating Token Index: 637

Text #7

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8058. Min Act: -0.1700

Data Index: 385668 (C4 (Web Text))

Max Activating Token Index: 885

Text #8

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8058. Min Act: -0.1700

Data Index: 486460 (C4 (Web Text))

Max Activating Token Index: 269

Text #9

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.7745. Min Act: -0.1699

Data Index: 1527402 (Python Code)

Max Activating Token Index: 10

Text #10

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.7433. Min Act: -0.1700

Data Index: 1403239 (Python Code)

Max Activating Token Index: 603

Text #11

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.7745. Min Act: -0.1700

Data Index: 1359943 (Python Code)

Max Activating Token Index: 475

Text #12

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.8058. Min Act: -0.1700

Data Index: 868415 (C4 (Web Text))

Max Activating Token Index: 297

Text #13

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.7433. Min Act: -0.1700

Data Index: 1637782 (Python Code)

Max Activating Token Index: 364

Text #14

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.6808. Min Act: -0.1700

Data Index: 1508952 (Python Code)

Max Activating Token Index: 728

Text #15

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.7433. Min Act: -0.1700

Data Index: 899597 (C4 (Web Text))

Max Activating Token Index: 747

Text #16

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.7433. Min Act: -0.1699

Data Index: 188587 (C4 (Web Text))

Max Activating Token Index: 702

Text #17

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.6808. Min Act: -0.1700

Data Index: 565106 (C4 (Web Text))

Max Activating Token Index: 966

Text #18

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.6808. Min Act: -0.1700

Data Index: 178047 (C4 (Web Text))

Max Activating Token Index: 90

Text #19

Max Range: 5.0558. Min Range: -5.0558

Max Act: 4.6495. Min Act: -0.1700

Data Index: 1380282 (Python Code)

Max Activating Token Index: 646