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

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

Neuron 1522 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: 5.3239. Min Range: -5.3239

Max Act: 5.3239. Min Act: -0.1700

Data Index: 970074 (C4 (Web Text))

Max Activating Token Index: 947

Text #1

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.3083. Min Act: -0.1700

Data Index: 1635 (C4 (Web Text))

Max Activating Token Index: 849

Text #2

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2614. Min Act: -0.1700

Data Index: 357226 (C4 (Web Text))

Max Activating Token Index: 203

Text #3

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2926. Min Act: -0.1700

Data Index: 323485 (C4 (Web Text))

Max Activating Token Index: 149

Text #4

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2614. Min Act: -0.1700

Data Index: 68125 (C4 (Web Text))

Max Activating Token Index: 449

Text #5

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2614. Min Act: -0.1700

Data Index: 1503025 (Python Code)

Max Activating Token Index: 27

Text #6

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2145. Min Act: -0.1700

Data Index: 642444 (C4 (Web Text))

Max Activating Token Index: 56

Text #7

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1989. Min Act: -0.1700

Data Index: 1266067 (C4 (Web Text))

Max Activating Token Index: 425

Text #8

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.3083. Min Act: -0.1700

Data Index: 443445 (C4 (Web Text))

Max Activating Token Index: 190

Text #9

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2301. Min Act: -0.1700

Data Index: 458234 (C4 (Web Text))

Max Activating Token Index: 493

Text #10

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1208. Min Act: -0.1700

Data Index: 331469 (C4 (Web Text))

Max Activating Token Index: 840

Text #11

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1989. Min Act: -0.1700

Data Index: 230943 (C4 (Web Text))

Max Activating Token Index: 838

Text #12

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1989. Min Act: -0.1700

Data Index: 1651861 (Python Code)

Max Activating Token Index: 962

Text #13

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.2926. Min Act: -0.1700

Data Index: 1294672 (C4 (Web Text))

Max Activating Token Index: 671

Text #14

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1520. Min Act: -0.1700

Data Index: 436590 (C4 (Web Text))

Max Activating Token Index: 835

Text #15

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1051. Min Act: -0.1700

Data Index: 865185 (C4 (Web Text))

Max Activating Token Index: 990

Text #16

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.0739. Min Act: -0.1700

Data Index: 200861 (C4 (Web Text))

Max Activating Token Index: 489

Text #17

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.0426. Min Act: -0.1700

Data Index: 432551 (C4 (Web Text))

Max Activating Token Index: 371

Text #18

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.0739. Min Act: -0.1700

Data Index: 64997 (C4 (Web Text))

Max Activating Token Index: 306

Text #19

Max Range: 5.3239. Min Range: -5.3239

Max Act: 5.1676. Min Act: -0.1700

Data Index: 855413 (C4 (Web Text))

Max Activating Token Index: 27