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

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

Neuron 1606 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: 5.0300. Min Range: -5.0300

Max Act: 5.0300. Min Act: -0.1700

Data Index: 376178 (C4 (Web Text))

Max Activating Token Index: 6

Text #1

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.7800. Min Act: -0.1700

Data Index: 450753 (C4 (Web Text))

Max Activating Token Index: 269

Text #2

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.7488. Min Act: -0.1700

Data Index: 294721 (C4 (Web Text))

Max Activating Token Index: 4

Text #3

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.7488. Min Act: -0.1700

Data Index: 1337307 (C4 (Web Text))

Max Activating Token Index: 248

Text #4

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.7800. Min Act: -0.1700

Data Index: 336459 (C4 (Web Text))

Max Activating Token Index: 380

Text #5

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.8113. Min Act: -0.1700

Data Index: 1095786 (C4 (Web Text))

Max Activating Token Index: 479

Text #6

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.6863. Min Act: -0.1700

Data Index: 885417 (C4 (Web Text))

Max Activating Token Index: 113

Text #7

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.6550. Min Act: -0.1700

Data Index: 7822 (C4 (Web Text))

Max Activating Token Index: 947

Text #8

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5925. Min Act: -0.1700

Data Index: 1014310 (C4 (Web Text))

Max Activating Token Index: 370

Text #9

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5300. Min Act: -0.1700

Data Index: 16391 (C4 (Web Text))

Max Activating Token Index: 366

Text #10

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5612. Min Act: -0.1700

Data Index: 1432495 (Python Code)

Max Activating Token Index: 527

Text #11

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5612. Min Act: -0.1700

Data Index: 1310721 (C4 (Web Text))

Max Activating Token Index: 519

Text #12

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5612. Min Act: -0.1700

Data Index: 747162 (C4 (Web Text))

Max Activating Token Index: 14

Text #13

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5300. Min Act: -0.1700

Data Index: 1028918 (C4 (Web Text))

Max Activating Token Index: 417

Text #14

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5612. Min Act: -0.1700

Data Index: 725113 (C4 (Web Text))

Max Activating Token Index: 469

Text #15

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5925. Min Act: -0.1700

Data Index: 1183902 (C4 (Web Text))

Max Activating Token Index: 13

Text #16

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5612. Min Act: -0.1700

Data Index: 978979 (C4 (Web Text))

Max Activating Token Index: 7

Text #17

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5300. Min Act: -0.1700

Data Index: 1237713 (C4 (Web Text))

Max Activating Token Index: 156

Text #18

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.5300. Min Act: -0.1700

Data Index: 973182 (C4 (Web Text))

Max Activating Token Index: 867

Text #19

Max Range: 5.0300. Min Range: -5.0300

Max Act: 4.4987. Min Act: -0.1700

Data Index: 1024812 (C4 (Web Text))

Max Activating Token Index: 17