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

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

Neuron 658 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: 5.4703. Min Range: -5.4703

Max Act: 5.4703. Min Act: -0.1699

Data Index: 1188760 (C4 (Web Text))

Max Activating Token Index: 393

Text #1

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.3141. Min Act: -0.1700

Data Index: 221763 (C4 (Web Text))

Max Activating Token Index: 279

Text #2

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.1578. Min Act: -0.1700

Data Index: 1579275 (Python Code)

Max Activating Token Index: 362

Text #3

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.1891. Min Act: -0.1700

Data Index: 85802 (C4 (Web Text))

Max Activating Token Index: 390

Text #4

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.1578. Min Act: -0.1700

Data Index: 894097 (C4 (Web Text))

Max Activating Token Index: 280

Text #5

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.1266. Min Act: -0.1700

Data Index: 289453 (C4 (Web Text))

Max Activating Token Index: 981

Text #6

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.0016. Min Act: -0.1700

Data Index: 61658 (C4 (Web Text))

Max Activating Token Index: 727

Text #7

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.0328. Min Act: -0.1700

Data Index: 886795 (C4 (Web Text))

Max Activating Token Index: 934

Text #8

Max Range: 5.4703. Min Range: -5.4703

Max Act: 5.0016. Min Act: -0.1700

Data Index: 1341302 (C4 (Web Text))

Max Activating Token Index: 513

Text #9

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.9703. Min Act: -0.1700

Data Index: 1538478 (Python Code)

Max Activating Token Index: 188

Text #10

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.9078. Min Act: -0.1700

Data Index: 1356998 (C4 (Web Text))

Max Activating Token Index: 437

Text #11

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.9391. Min Act: -0.1700

Data Index: 627639 (C4 (Web Text))

Max Activating Token Index: 783

Text #12

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.9391. Min Act: -0.1699

Data Index: 1000736 (C4 (Web Text))

Max Activating Token Index: 440

Text #13

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.9078. Min Act: -0.1700

Data Index: 1539719 (Python Code)

Max Activating Token Index: 912

Text #14

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.8453. Min Act: -0.1699

Data Index: 965231 (C4 (Web Text))

Max Activating Token Index: 444

Text #15

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.9078. Min Act: -0.1700

Data Index: 1460527 (Python Code)

Max Activating Token Index: 844

Text #16

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.8766. Min Act: -0.1700

Data Index: 584084 (C4 (Web Text))

Max Activating Token Index: 26

Text #17

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.8453. Min Act: -0.1700

Data Index: 834893 (C4 (Web Text))

Max Activating Token Index: 551

Text #18

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.7516. Min Act: -0.1700

Data Index: 1235319 (C4 (Web Text))

Max Activating Token Index: 1002

Text #19

Max Range: 5.4703. Min Range: -5.4703

Max Act: 4.8141. Min Act: -0.1700

Data Index: 308803 (C4 (Web Text))

Max Activating Token Index: 355