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

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

Neuron 1394 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: 5.1083. Min Range: -5.1083

Max Act: 5.1083. Min Act: -0.1700

Data Index: 378200 (C4 (Web Text))

Max Activating Token Index: 124

Text #1

Max Range: 5.1083. Min Range: -5.1083

Max Act: 5.0771. Min Act: -0.1699

Data Index: 1623364 (Python Code)

Max Activating Token Index: 853

Text #2

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.8896. Min Act: -0.1700

Data Index: 517193 (C4 (Web Text))

Max Activating Token Index: 397

Text #3

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.8583. Min Act: -0.1700

Data Index: 165024 (C4 (Web Text))

Max Activating Token Index: 810

Text #4

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.8896. Min Act: -0.1700

Data Index: 451071 (C4 (Web Text))

Max Activating Token Index: 53

Text #5

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.8271. Min Act: -0.1700

Data Index: 948735 (C4 (Web Text))

Max Activating Token Index: 847

Text #6

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.8271. Min Act: -0.1700

Data Index: 607170 (C4 (Web Text))

Max Activating Token Index: 26

Text #7

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7958. Min Act: -0.1700

Data Index: 1345960 (C4 (Web Text))

Max Activating Token Index: 367

Text #8

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7958. Min Act: -0.1700

Data Index: 776642 (C4 (Web Text))

Max Activating Token Index: 421

Text #9

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7646. Min Act: -0.1700

Data Index: 522432 (C4 (Web Text))

Max Activating Token Index: 156

Text #10

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7646. Min Act: -0.1700

Data Index: 983749 (C4 (Web Text))

Max Activating Token Index: 55

Text #11

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7021. Min Act: -0.1700

Data Index: 1471946 (Python Code)

Max Activating Token Index: 1008

Text #12

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7333. Min Act: -0.1700

Data Index: 800974 (C4 (Web Text))

Max Activating Token Index: 66

Text #13

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.6708. Min Act: -0.1700

Data Index: 1355020 (C4 (Web Text))

Max Activating Token Index: 297

Text #14

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.6708. Min Act: -0.1700

Data Index: 1298871 (C4 (Web Text))

Max Activating Token Index: 749

Text #15

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7021. Min Act: -0.1700

Data Index: 1557380 (Python Code)

Max Activating Token Index: 524

Text #16

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7021. Min Act: -0.1700

Data Index: 1329324 (C4 (Web Text))

Max Activating Token Index: 144

Text #17

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.6708. Min Act: -0.1700

Data Index: 431103 (C4 (Web Text))

Max Activating Token Index: 45

Text #18

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.6708. Min Act: -0.1700

Data Index: 187604 (C4 (Web Text))

Max Activating Token Index: 706

Text #19

Max Range: 5.1083. Min Range: -5.1083

Max Act: 4.7021. Min Act: -0.1700

Data Index: 805974 (C4 (Web Text))

Max Activating Token Index: 580