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

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

Neuron 50 in Layer 0

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.4824. Min Act: -0.1700

Data Index: 720638 (C4 (Web Text))

Max Activating Token Index: 863

Text #1

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.4040. Min Act: -0.1700

Data Index: 1258433 (C4 (Web Text))

Max Activating Token Index: 134

Text #2

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.3726. Min Act: -0.1700

Data Index: 1336034 (C4 (Web Text))

Max Activating Token Index: 506

Text #3

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2784. Min Act: -0.1700

Data Index: 1315487 (C4 (Web Text))

Max Activating Token Index: 770

Text #4

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2627. Min Act: -0.1700

Data Index: 293409 (C4 (Web Text))

Max Activating Token Index: 491

Text #5

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2470. Min Act: -0.1700

Data Index: 355017 (C4 (Web Text))

Max Activating Token Index: 948

Text #6

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2470. Min Act: -0.1699

Data Index: 1571493 (Python Code)

Max Activating Token Index: 3

Text #7

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2313. Min Act: -0.1700

Data Index: 857978 (C4 (Web Text))

Max Activating Token Index: 247

Text #8

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2155. Min Act: -0.1700

Data Index: 259638 (C4 (Web Text))

Max Activating Token Index: 44

Text #9

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.2155. Min Act: -0.1700

Data Index: 988037 (C4 (Web Text))

Max Activating Token Index: 10

Text #10

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1841. Min Act: -0.1700

Data Index: 129923 (C4 (Web Text))

Max Activating Token Index: 132

Text #11

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1368. Min Act: -0.1700

Data Index: 762295 (C4 (Web Text))

Max Activating Token Index: 859

Text #12

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1526. Min Act: -0.1700

Data Index: 1022456 (C4 (Web Text))

Max Activating Token Index: 243

Text #13

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1368. Min Act: -0.1700

Data Index: 1377382 (Python Code)

Max Activating Token Index: 761

Text #14

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1211. Min Act: -0.1700

Data Index: 1317594 (C4 (Web Text))

Max Activating Token Index: 786

Text #15

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1526. Min Act: -0.1700

Data Index: 890661 (C4 (Web Text))

Max Activating Token Index: 657

Text #16

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1053. Min Act: -0.1700

Data Index: 1104070 (C4 (Web Text))

Max Activating Token Index: 829

Text #17

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1053. Min Act: -0.1700

Data Index: 734396 (C4 (Web Text))

Max Activating Token Index: 337

Text #18

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1053. Min Act: -0.1700

Data Index: 1173968 (C4 (Web Text))

Max Activating Token Index: 551

Text #19

Max Range: 3.4824. Min Range: -3.4824

Max Act: 3.1211. Min Act: -0.1700

Data Index: 1300872 (C4 (Web Text))

Max Activating Token Index: 639