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

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

Neuron 1506 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: 5.1225. Min Range: -5.1225

Max Act: 5.1225. Min Act: -0.1700

Data Index: 723970 (C4 (Web Text))

Max Activating Token Index: 790

Text #1

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.9350. Min Act: -0.1700

Data Index: 477015 (C4 (Web Text))

Max Activating Token Index: 354

Text #2

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.9037. Min Act: -0.1700

Data Index: 536294 (C4 (Web Text))

Max Activating Token Index: 870

Text #3

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.8725. Min Act: -0.1700

Data Index: 165909 (C4 (Web Text))

Max Activating Token Index: 303

Text #4

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.7787. Min Act: -0.1700

Data Index: 1005705 (C4 (Web Text))

Max Activating Token Index: 162

Text #5

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.7162. Min Act: -0.1700

Data Index: 251598 (C4 (Web Text))

Max Activating Token Index: 866

Text #6

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.7474. Min Act: -0.1700

Data Index: 1130100 (C4 (Web Text))

Max Activating Token Index: 597

Text #7

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6849. Min Act: -0.1700

Data Index: 1443112 (Python Code)

Max Activating Token Index: 238

Text #8

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6537. Min Act: -0.1700

Data Index: 629365 (C4 (Web Text))

Max Activating Token Index: 250

Text #9

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6537. Min Act: -0.1700

Data Index: 288101 (C4 (Web Text))

Max Activating Token Index: 947

Text #10

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.5912. Min Act: -0.1700

Data Index: 882169 (C4 (Web Text))

Max Activating Token Index: 611

Text #11

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6537. Min Act: -0.1700

Data Index: 254016 (C4 (Web Text))

Max Activating Token Index: 968

Text #12

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6224. Min Act: -0.1700

Data Index: 945448 (C4 (Web Text))

Max Activating Token Index: 116

Text #13

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.5912. Min Act: -0.1700

Data Index: 21543 (C4 (Web Text))

Max Activating Token Index: 58

Text #14

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.5599. Min Act: -0.1700

Data Index: 106604 (C4 (Web Text))

Max Activating Token Index: 298

Text #15

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6537. Min Act: -0.1700

Data Index: 676426 (C4 (Web Text))

Max Activating Token Index: 223

Text #16

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.6224. Min Act: -0.1700

Data Index: 524142 (C4 (Web Text))

Max Activating Token Index: 785

Text #17

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.5912. Min Act: -0.1700

Data Index: 1388101 (Python Code)

Max Activating Token Index: 476

Text #18

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.5599. Min Act: -0.1700

Data Index: 66429 (C4 (Web Text))

Max Activating Token Index: 733

Text #19

Max Range: 5.1225. Min Range: -5.1225

Max Act: 4.5912. Min Act: -0.1700

Data Index: 778056 (C4 (Web Text))

Max Activating Token Index: 262