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

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

Neuron 1198 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: 5.6348. Min Range: -5.6348

Max Act: 5.6348. Min Act: -0.1700

Data Index: 1511257 (Python Code)

Max Activating Token Index: 1013

Text #1

Max Range: 5.6348. Min Range: -5.6348

Max Act: 5.3067. Min Act: -0.1699

Data Index: 573537 (C4 (Web Text))

Max Activating Token Index: 349

Text #2

Max Range: 5.6348. Min Range: -5.6348

Max Act: 5.1348. Min Act: -0.1697

Data Index: 133094 (C4 (Web Text))

Max Activating Token Index: 1013

Text #3

Max Range: 5.6348. Min Range: -5.6348

Max Act: 5.1348. Min Act: -0.1700

Data Index: 71665 (C4 (Web Text))

Max Activating Token Index: 789

Text #4

Max Range: 5.6348. Min Range: -5.6348

Max Act: 5.0254. Min Act: -0.1700

Data Index: 405900 (C4 (Web Text))

Max Activating Token Index: 408

Text #5

Max Range: 5.6348. Min Range: -5.6348

Max Act: 5.0567. Min Act: -0.1700

Data Index: 706971 (C4 (Web Text))

Max Activating Token Index: 954

Text #6

Max Range: 5.6348. Min Range: -5.6348

Max Act: 5.0723. Min Act: -0.1700

Data Index: 201093 (C4 (Web Text))

Max Activating Token Index: 840

Text #7

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9473. Min Act: -0.1700

Data Index: 427566 (C4 (Web Text))

Max Activating Token Index: 577

Text #8

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9317. Min Act: -0.1700

Data Index: 408321 (C4 (Web Text))

Max Activating Token Index: 971

Text #9

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9004. Min Act: -0.1700

Data Index: 1421811 (Python Code)

Max Activating Token Index: 952

Text #10

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9317. Min Act: -0.1700

Data Index: 688589 (C4 (Web Text))

Max Activating Token Index: 885

Text #11

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9004. Min Act: -0.1700

Data Index: 600945 (C4 (Web Text))

Max Activating Token Index: 19

Text #12

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.8223. Min Act: -0.1700

Data Index: 1338042 (C4 (Web Text))

Max Activating Token Index: 495

Text #13

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9629. Min Act: -0.1700

Data Index: 188275 (C4 (Web Text))

Max Activating Token Index: 345

Text #14

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.8613. Min Act: -0.1700

Data Index: 1519590 (Python Code)

Max Activating Token Index: 952

Text #15

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.9160. Min Act: -0.1700

Data Index: 142718 (C4 (Web Text))

Max Activating Token Index: 1014

Text #16

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.8692. Min Act: -0.1700

Data Index: 657309 (C4 (Web Text))

Max Activating Token Index: 608

Text #17

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.8926. Min Act: -0.1700

Data Index: 264230 (C4 (Web Text))

Max Activating Token Index: 69

Text #18

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.8848. Min Act: -0.1700

Data Index: 596914 (C4 (Web Text))

Max Activating Token Index: 487

Text #19

Max Range: 5.6348. Min Range: -5.6348

Max Act: 4.8692. Min Act: -0.1700

Data Index: 349895 (C4 (Web Text))

Max Activating Token Index: 999