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

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

Neuron 379 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.5708. Min Act: -0.1700

Data Index: 463 (C4 (Web Text))

Max Activating Token Index: 807

Text #1

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.4768. Min Act: -0.1700

Data Index: 1267538 (C4 (Web Text))

Max Activating Token Index: 32

Text #2

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.3670. Min Act: -0.1700

Data Index: 512864 (C4 (Web Text))

Max Activating Token Index: 750

Text #3

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.3200. Min Act: -0.1700

Data Index: 1346729 (C4 (Web Text))

Max Activating Token Index: 208

Text #4

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.3043. Min Act: -0.1700

Data Index: 755560 (C4 (Web Text))

Max Activating Token Index: 595

Text #5

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.3357. Min Act: -0.1700

Data Index: 548950 (C4 (Web Text))

Max Activating Token Index: 739

Text #6

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2885. Min Act: -0.1700

Data Index: 1220848 (C4 (Web Text))

Max Activating Token Index: 830

Text #7

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2885. Min Act: -0.1700

Data Index: 354018 (C4 (Web Text))

Max Activating Token Index: 918

Text #8

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2728. Min Act: -0.1700

Data Index: 960332 (C4 (Web Text))

Max Activating Token Index: 705

Text #9

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2414. Min Act: -0.1700

Data Index: 1166399 (C4 (Web Text))

Max Activating Token Index: 387

Text #10

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2728. Min Act: -0.1700

Data Index: 1093033 (C4 (Web Text))

Max Activating Token Index: 248

Text #11

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2728. Min Act: -0.1700

Data Index: 784369 (C4 (Web Text))

Max Activating Token Index: 320

Text #12

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2414. Min Act: -0.1700

Data Index: 915003 (C4 (Web Text))

Max Activating Token Index: 40

Text #13

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2257. Min Act: -0.1700

Data Index: 160149 (C4 (Web Text))

Max Activating Token Index: 975

Text #14

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2257. Min Act: -0.1700

Data Index: 1018620 (C4 (Web Text))

Max Activating Token Index: 312

Text #15

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2414. Min Act: -0.1700

Data Index: 346601 (C4 (Web Text))

Max Activating Token Index: 647

Text #16

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2099. Min Act: -0.1700

Data Index: 541020 (C4 (Web Text))

Max Activating Token Index: 492

Text #17

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.1942. Min Act: -0.1700

Data Index: 1625998 (Python Code)

Max Activating Token Index: 355

Text #18

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.2571. Min Act: -0.1700

Data Index: 1198414 (C4 (Web Text))

Max Activating Token Index: 255

Text #19

Max Range: 3.5708. Min Range: -3.5708

Max Act: 3.1942. Min Act: -0.1700

Data Index: 485055 (C4 (Web Text))

Max Activating Token Index: 870