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

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

Neuron 347 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: 5.1716. Min Range: -5.1716

Max Act: 5.1716. Min Act: -0.1700

Data Index: 1147836 (C4 (Web Text))

Max Activating Token Index: 355

Text #1

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.8825. Min Act: -0.1700

Data Index: 618993 (C4 (Web Text))

Max Activating Token Index: 506

Text #2

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.8552. Min Act: -0.1700

Data Index: 1595229 (Python Code)

Max Activating Token Index: 987

Text #3

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.8005. Min Act: -0.1700

Data Index: 135839 (C4 (Web Text))

Max Activating Token Index: 192

Text #4

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.8239. Min Act: -0.1700

Data Index: 1105758 (C4 (Web Text))

Max Activating Token Index: 657

Text #5

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.8161. Min Act: -0.1700

Data Index: 270397 (C4 (Web Text))

Max Activating Token Index: 962

Text #6

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.7028. Min Act: -0.1700

Data Index: 1479000 (Python Code)

Max Activating Token Index: 42

Text #7

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.7106. Min Act: -0.1700

Data Index: 1590510 (Python Code)

Max Activating Token Index: 905

Text #8

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6657. Min Act: -0.1700

Data Index: 1436498 (Python Code)

Max Activating Token Index: 979

Text #9

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6501. Min Act: -0.1700

Data Index: 997720 (C4 (Web Text))

Max Activating Token Index: 265

Text #10

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.7028. Min Act: -0.1700

Data Index: 1527709 (Python Code)

Max Activating Token Index: 764

Text #11

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6852. Min Act: -0.1700

Data Index: 1614360 (Python Code)

Max Activating Token Index: 777

Text #12

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6618. Min Act: -0.1700

Data Index: 834228 (C4 (Web Text))

Max Activating Token Index: 614

Text #13

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6188. Min Act: -0.1700

Data Index: 241241 (C4 (Web Text))

Max Activating Token Index: 409

Text #14

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6110. Min Act: -0.1700

Data Index: 620424 (C4 (Web Text))

Max Activating Token Index: 635

Text #15

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.5329. Min Act: -0.1700

Data Index: 512308 (C4 (Web Text))

Max Activating Token Index: 1012

Text #16

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6383. Min Act: -0.1700

Data Index: 175550 (C4 (Web Text))

Max Activating Token Index: 902

Text #17

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6169. Min Act: -0.1700

Data Index: 456064 (C4 (Web Text))

Max Activating Token Index: 925

Text #18

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6012. Min Act: -0.1700

Data Index: 1295201 (C4 (Web Text))

Max Activating Token Index: 726

Text #19

Max Range: 5.1716. Min Range: -5.1716

Max Act: 4.6071. Min Act: -0.1700

Data Index: 1527615 (Python Code)

Max Activating Token Index: 387