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

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

Neuron 1375 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: 3.3240. Min Range: -3.3240

Max Act: 3.2769. Min Act: -0.1700

Data Index: 1150929 (C4 (Web Text))

Max Activating Token Index: 744

Text #1

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.3240. Min Act: -0.1700

Data Index: 1438 (C4 (Web Text))

Max Activating Token Index: 675

Text #2

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.2140. Min Act: -0.1700

Data Index: 234612 (C4 (Web Text))

Max Activating Token Index: 830

Text #3

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.1511. Min Act: -0.1700

Data Index: 1640615 (Python Code)

Max Activating Token Index: 325

Text #4

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 244798 (C4 (Web Text))

Max Activating Token Index: 944

Text #5

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.1038. Min Act: -0.1700

Data Index: 4172 (C4 (Web Text))

Max Activating Token Index: 258

Text #6

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0722. Min Act: -0.1700

Data Index: 656633 (C4 (Web Text))

Max Activating Token Index: 242

Text #7

Max Range: 3.3240. Min Range: -3.3240

Max Act: 2.9933. Min Act: -0.1700

Data Index: 804505 (C4 (Web Text))

Max Activating Token Index: 807

Text #8

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0722. Min Act: -0.1700

Data Index: 98621 (C4 (Web Text))

Max Activating Token Index: 307

Text #9

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 241443 (C4 (Web Text))

Max Activating Token Index: 515

Text #10

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0565. Min Act: -0.1700

Data Index: 1111453 (C4 (Web Text))

Max Activating Token Index: 844

Text #11

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 1137586 (C4 (Web Text))

Max Activating Token Index: 300

Text #12

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0565. Min Act: -0.1700

Data Index: 173422 (C4 (Web Text))

Max Activating Token Index: 260

Text #13

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.1038. Min Act: -0.1700

Data Index: 506214 (C4 (Web Text))

Max Activating Token Index: 572

Text #14

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.1038. Min Act: -0.1700

Data Index: 1357719 (C4 (Web Text))

Max Activating Token Index: 492

Text #15

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0407. Min Act: -0.1700

Data Index: 1377754 (Python Code)

Max Activating Token Index: 793

Text #16

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 1508664 (Python Code)

Max Activating Token Index: 756

Text #17

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 1342973 (C4 (Web Text))

Max Activating Token Index: 184

Text #18

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 147178 (C4 (Web Text))

Max Activating Token Index: 438

Text #19

Max Range: 3.3240. Min Range: -3.3240

Max Act: 3.0880. Min Act: -0.1700

Data Index: 385537 (C4 (Web Text))

Max Activating Token Index: 132