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

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

Neuron 1304 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: 3.1900. Min Range: -3.1900

Max Act: 3.1900. Min Act: -0.1700

Data Index: 499085 (C4 (Web Text))

Max Activating Token Index: 703

Text #1

Max Range: 3.1900. Min Range: -3.1900

Max Act: 3.1113. Min Act: -0.1700

Data Index: 818958 (C4 (Web Text))

Max Activating Token Index: 581

Text #2

Max Range: 3.1900. Min Range: -3.1900

Max Act: 3.0955. Min Act: -0.1700

Data Index: 1431075 (Python Code)

Max Activating Token Index: 654

Text #3

Max Range: 3.1900. Min Range: -3.1900

Max Act: 3.0482. Min Act: -0.1700

Data Index: 191506 (C4 (Web Text))

Max Activating Token Index: 924

Text #4

Max Range: 3.1900. Min Range: -3.1900

Max Act: 3.0166. Min Act: -0.1700

Data Index: 1078108 (C4 (Web Text))

Max Activating Token Index: 200

Text #5

Max Range: 3.1900. Min Range: -3.1900

Max Act: 3.0166. Min Act: -0.1700

Data Index: 1184169 (C4 (Web Text))

Max Activating Token Index: 496

Text #6

Max Range: 3.1900. Min Range: -3.1900

Max Act: 3.0008. Min Act: -0.1700

Data Index: 650730 (C4 (Web Text))

Max Activating Token Index: 662

Text #7

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9850. Min Act: -0.1700

Data Index: 358241 (C4 (Web Text))

Max Activating Token Index: 920

Text #8

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9533. Min Act: -0.1700

Data Index: 1648480 (Python Code)

Max Activating Token Index: 335

Text #9

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9533. Min Act: -0.1700

Data Index: 1028964 (C4 (Web Text))

Max Activating Token Index: 598

Text #10

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9217. Min Act: -0.1700

Data Index: 1298159 (C4 (Web Text))

Max Activating Token Index: 352

Text #11

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9850. Min Act: -0.1700

Data Index: 652377 (C4 (Web Text))

Max Activating Token Index: 730

Text #12

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.8741. Min Act: -0.1700

Data Index: 141555 (C4 (Web Text))

Max Activating Token Index: 421

Text #13

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9533. Min Act: -0.1700

Data Index: 954021 (C4 (Web Text))

Max Activating Token Index: 348

Text #14

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.8741. Min Act: -0.1700

Data Index: 196677 (C4 (Web Text))

Max Activating Token Index: 447

Text #15

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9533. Min Act: -0.1700

Data Index: 601637 (C4 (Web Text))

Max Activating Token Index: 116

Text #16

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9692. Min Act: -0.1700

Data Index: 584405 (C4 (Web Text))

Max Activating Token Index: 595

Text #17

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9533. Min Act: -0.1700

Data Index: 269743 (C4 (Web Text))

Max Activating Token Index: 640

Text #18

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9375. Min Act: -0.1700

Data Index: 1480429 (Python Code)

Max Activating Token Index: 282

Text #19

Max Range: 3.1900. Min Range: -3.1900

Max Act: 2.9533. Min Act: -0.1700

Data Index: 654174 (C4 (Web Text))

Max Activating Token Index: 337