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

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

Neuron 1138 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: 4.3169. Min Range: -4.3169

Max Act: 4.3169. Min Act: -0.1700

Data Index: 925397 (C4 (Web Text))

Max Activating Token Index: 1023

Text #1

Max Range: 4.3169. Min Range: -4.3169

Max Act: 4.2544. Min Act: -0.1700

Data Index: 1493114 (Python Code)

Max Activating Token Index: 1023

Text #2

Max Range: 4.3169. Min Range: -4.3169

Max Act: 4.1294. Min Act: -0.1699

Data Index: 1653566 (Python Code)

Max Activating Token Index: 224

Text #3

Max Range: 4.3169. Min Range: -4.3169

Max Act: 4.0669. Min Act: -0.1699

Data Index: 1421889 (Python Code)

Max Activating Token Index: 1023

Text #4

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.9887. Min Act: -0.1700

Data Index: 1612729 (Python Code)

Max Activating Token Index: 1023

Text #5

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.8793. Min Act: -0.1700

Data Index: 599058 (C4 (Web Text))

Max Activating Token Index: 1023

Text #6

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.9105. Min Act: -0.1700

Data Index: 1498369 (Python Code)

Max Activating Token Index: 1023

Text #7

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.9262. Min Act: -0.1700

Data Index: 1653177 (Python Code)

Max Activating Token Index: 1023

Text #8

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.9105. Min Act: -0.1700

Data Index: 374212 (C4 (Web Text))

Max Activating Token Index: 1023

Text #9

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.8167. Min Act: -0.1700

Data Index: 384730 (C4 (Web Text))

Max Activating Token Index: 1023

Text #10

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.9105. Min Act: -0.1700

Data Index: 1519866 (Python Code)

Max Activating Token Index: 666

Text #11

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.8011. Min Act: -0.1700

Data Index: 1524412 (Python Code)

Max Activating Token Index: 1023

Text #12

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.6916. Min Act: -0.1700

Data Index: 1365514 (Python Code)

Max Activating Token Index: 1023

Text #13

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7698. Min Act: -0.1669

Data Index: 1502873 (Python Code)

Max Activating Token Index: 1023

Text #14

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7541. Min Act: -0.1678

Data Index: 1474020 (Python Code)

Max Activating Token Index: 1023

Text #15

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7854. Min Act: -0.1700

Data Index: 1415440 (Python Code)

Max Activating Token Index: 234

Text #16

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7541. Min Act: -0.1700

Data Index: 1650783 (Python Code)

Max Activating Token Index: 36

Text #17

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7229. Min Act: -0.1700

Data Index: 1534012 (Python Code)

Max Activating Token Index: 1023

Text #18

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7385. Min Act: -0.1700

Data Index: 1603385 (Python Code)

Max Activating Token Index: 1023

Text #19

Max Range: 4.3169. Min Range: -4.3169

Max Act: 3.7229. Min Act: -0.1700

Data Index: 1480781 (Python Code)

Max Activating Token Index: 247