Model: GELU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 43 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-3l')

Text #0

Max Range: 7.1185. Min Range: -7.1185

Max Act: 7.1185. Min Act: -0.1700

Data Index: 1600956 (Python Code)

Max Activating Token Index: 794

Text #1

Max Range: 7.1185. Min Range: -7.1185

Max Act: 7.0872. Min Act: -0.1700

Data Index: 199799 (C4 (Web Text))

Max Activating Token Index: 1009

Text #2

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.9310. Min Act: -0.1700

Data Index: 22070 (C4 (Web Text))

Max Activating Token Index: 1013

Text #3

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8997. Min Act: -0.1700

Data Index: 1650669 (Python Code)

Max Activating Token Index: 85

Text #4

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8997. Min Act: -0.1700

Data Index: 157393 (C4 (Web Text))

Max Activating Token Index: 857

Text #5

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8685. Min Act: -0.1700

Data Index: 1134447 (C4 (Web Text))

Max Activating Token Index: 904

Text #6

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8372. Min Act: -0.1700

Data Index: 397573 (C4 (Web Text))

Max Activating Token Index: 182

Text #7

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8060. Min Act: -0.1700

Data Index: 280821 (C4 (Web Text))

Max Activating Token Index: 657

Text #8

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8372. Min Act: -0.1700

Data Index: 538099 (C4 (Web Text))

Max Activating Token Index: 778

Text #9

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.7435. Min Act: -0.1700

Data Index: 1058502 (C4 (Web Text))

Max Activating Token Index: 893

Text #10

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8060. Min Act: -0.1700

Data Index: 23790 (C4 (Web Text))

Max Activating Token Index: 908

Text #11

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8060. Min Act: -0.1700

Data Index: 1612423 (Python Code)

Max Activating Token Index: 608

Text #12

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.5872. Min Act: -0.1700

Data Index: 1413630 (Python Code)

Max Activating Token Index: 139

Text #13

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.7747. Min Act: -0.1700

Data Index: 1434328 (Python Code)

Max Activating Token Index: 540

Text #14

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.7747. Min Act: -0.1700

Data Index: 1602152 (Python Code)

Max Activating Token Index: 942

Text #15

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.7747. Min Act: -0.1700

Data Index: 566054 (C4 (Web Text))

Max Activating Token Index: 331

Text #16

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8060. Min Act: -0.1700

Data Index: 739019 (C4 (Web Text))

Max Activating Token Index: 173

Text #17

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.8060. Min Act: -0.1700

Data Index: 1501122 (Python Code)

Max Activating Token Index: 776

Text #18

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.7435. Min Act: -0.1700

Data Index: 745573 (C4 (Web Text))

Max Activating Token Index: 551

Text #19

Max Range: 7.1185. Min Range: -7.1185

Max Act: 6.6810. Min Act: -0.1700

Data Index: 303704 (C4 (Web Text))

Max Activating Token Index: 449