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

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

Neuron 3 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.2889. Min Act: -0.1700

Data Index: 254193 (C4 (Web Text))

Max Activating Token Index: 632

Text #1

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1789. Min Act: -0.1700

Data Index: 823261 (C4 (Web Text))

Max Activating Token Index: 451

Text #2

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1631. Min Act: -0.1700

Data Index: 179320 (C4 (Web Text))

Max Activating Token Index: 521

Text #3

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1316. Min Act: -0.1700

Data Index: 1329792 (C4 (Web Text))

Max Activating Token Index: 913

Text #4

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1474. Min Act: -0.1700

Data Index: 740580 (C4 (Web Text))

Max Activating Token Index: 504

Text #5

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1159. Min Act: -0.1700

Data Index: 730025 (C4 (Web Text))

Max Activating Token Index: 663

Text #6

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1316. Min Act: -0.1700

Data Index: 200544 (C4 (Web Text))

Max Activating Token Index: 319

Text #7

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0528. Min Act: -0.1700

Data Index: 288906 (C4 (Web Text))

Max Activating Token Index: 371

Text #8

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1001. Min Act: -0.1700

Data Index: 459230 (C4 (Web Text))

Max Activating Token Index: 175

Text #9

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1001. Min Act: -0.1700

Data Index: 24489 (C4 (Web Text))

Max Activating Token Index: 83

Text #10

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1001. Min Act: -0.1700

Data Index: 519164 (C4 (Web Text))

Max Activating Token Index: 989

Text #11

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0843. Min Act: -0.1700

Data Index: 1248439 (C4 (Web Text))

Max Activating Token Index: 607

Text #12

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0843. Min Act: -0.1700

Data Index: 329673 (C4 (Web Text))

Max Activating Token Index: 857

Text #13

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1001. Min Act: -0.1700

Data Index: 1405913 (Python Code)

Max Activating Token Index: 245

Text #14

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1001. Min Act: -0.1700

Data Index: 685241 (C4 (Web Text))

Max Activating Token Index: 277

Text #15

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.1001. Min Act: -0.1700

Data Index: 330850 (C4 (Web Text))

Max Activating Token Index: 819

Text #16

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0843. Min Act: -0.1700

Data Index: 1039192 (C4 (Web Text))

Max Activating Token Index: 675

Text #17

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0370. Min Act: -0.1700

Data Index: 260126 (C4 (Web Text))

Max Activating Token Index: 787

Text #18

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0054. Min Act: -0.1700

Data Index: 807385 (C4 (Web Text))

Max Activating Token Index: 415

Text #19

Max Range: 3.2889. Min Range: -3.2889

Max Act: 3.0686. Min Act: -0.1700

Data Index: 251736 (C4 (Web Text))

Max Activating Token Index: 255