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

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

Neuron 1035 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: 3.4277. Min Range: -3.4277

Max Act: 3.4277. Min Act: -0.1700

Data Index: 1618546 (Python Code)

Max Activating Token Index: 1023

Text #1

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3649. Min Act: -0.1700

Data Index: 998605 (C4 (Web Text))

Max Activating Token Index: 795

Text #2

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3963. Min Act: -0.1700

Data Index: 787953 (C4 (Web Text))

Max Activating Token Index: 41

Text #3

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3806. Min Act: -0.1700

Data Index: 137358 (C4 (Web Text))

Max Activating Token Index: 370

Text #4

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3492. Min Act: -0.1700

Data Index: 1198204 (C4 (Web Text))

Max Activating Token Index: 97

Text #5

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3335. Min Act: -0.1700

Data Index: 1569755 (Python Code)

Max Activating Token Index: 255

Text #6

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3335. Min Act: -0.1700

Data Index: 853755 (C4 (Web Text))

Max Activating Token Index: 796

Text #7

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3492. Min Act: -0.1700

Data Index: 715404 (C4 (Web Text))

Max Activating Token Index: 285

Text #8

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3335. Min Act: -0.1700

Data Index: 1013330 (C4 (Web Text))

Max Activating Token Index: 841

Text #9

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3178. Min Act: -0.1700

Data Index: 67113 (C4 (Web Text))

Max Activating Token Index: 17

Text #10

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3335. Min Act: -0.1700

Data Index: 766211 (C4 (Web Text))

Max Activating Token Index: 5

Text #11

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3178. Min Act: -0.1700

Data Index: 1554334 (Python Code)

Max Activating Token Index: 1023

Text #12

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3335. Min Act: -0.1700

Data Index: 1547629 (Python Code)

Max Activating Token Index: 512

Text #13

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3178. Min Act: -0.1700

Data Index: 1452893 (Python Code)

Max Activating Token Index: 493

Text #14

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3178. Min Act: -0.1700

Data Index: 1569480 (Python Code)

Max Activating Token Index: 270

Text #15

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3021. Min Act: -0.1700

Data Index: 291276 (C4 (Web Text))

Max Activating Token Index: 700

Text #16

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3178. Min Act: -0.1700

Data Index: 932164 (C4 (Web Text))

Max Activating Token Index: 347

Text #17

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3178. Min Act: -0.1700

Data Index: 32144 (C4 (Web Text))

Max Activating Token Index: 265

Text #18

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3021. Min Act: -0.1700

Data Index: 998357 (C4 (Web Text))

Max Activating Token Index: 286

Text #19

Max Range: 3.4277. Min Range: -3.4277

Max Act: 3.3021. Min Act: -0.1700

Data Index: 740316 (C4 (Web Text))

Max Activating Token Index: 482