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

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

Neuron 14 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: 9.2063. Min Range: -9.2063

Max Act: 9.2063. Min Act: 0.0389

Data Index: 33331 (C4 (Web Text))

Max Activating Token Index: 61

Text #1

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.8313. Min Act: -0.1695

Data Index: 391115 (C4 (Web Text))

Max Activating Token Index: 575

Text #2

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.6438. Min Act: -0.1700

Data Index: 1181148 (C4 (Web Text))

Max Activating Token Index: 662

Text #3

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.7063. Min Act: -0.1700

Data Index: 881261 (C4 (Web Text))

Max Activating Token Index: 110

Text #4

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.6438. Min Act: -0.1700

Data Index: 898891 (C4 (Web Text))

Max Activating Token Index: 131

Text #5

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.4563. Min Act: -0.1700

Data Index: 543147 (C4 (Web Text))

Max Activating Token Index: 640

Text #6

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.5188. Min Act: -0.1698

Data Index: 200542 (C4 (Web Text))

Max Activating Token Index: 267

Text #7

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.5188. Min Act: -0.1700

Data Index: 1191763 (C4 (Web Text))

Max Activating Token Index: 463

Text #8

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.4563. Min Act: -0.1700

Data Index: 130786 (C4 (Web Text))

Max Activating Token Index: 417

Text #9

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.5188. Min Act: -0.1700

Data Index: 495404 (C4 (Web Text))

Max Activating Token Index: 609

Text #10

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.3938. Min Act: -0.1700

Data Index: 1336452 (C4 (Web Text))

Max Activating Token Index: 805

Text #11

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.3938. Min Act: -0.1699

Data Index: 887325 (C4 (Web Text))

Max Activating Token Index: 824

Text #12

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.3313. Min Act: -0.1700

Data Index: 1077031 (C4 (Web Text))

Max Activating Token Index: 424

Text #13

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.3313. Min Act: -0.1004

Data Index: 1012308 (C4 (Web Text))

Max Activating Token Index: 34

Text #14

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.3938. Min Act: -0.1696

Data Index: 485699 (C4 (Web Text))

Max Activating Token Index: 226

Text #15

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.2688. Min Act: -0.1337

Data Index: 19948 (C4 (Web Text))

Max Activating Token Index: 420

Text #16

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.2063. Min Act: -0.1700

Data Index: 1422180 (Python Code)

Max Activating Token Index: 670

Text #17

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.2063. Min Act: -0.1699

Data Index: 375452 (C4 (Web Text))

Max Activating Token Index: 396

Text #18

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.2063. Min Act: -0.1700

Data Index: 649853 (C4 (Web Text))

Max Activating Token Index: 779

Text #19

Max Range: 9.2063. Min Range: -9.2063

Max Act: 8.2063. Min Act: -0.1699

Data Index: 7025 (C4 (Web Text))

Max Activating Token Index: 540