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

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

Neuron 1720 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: 5.3475. Min Range: -5.3475

Max Act: 5.3475. Min Act: -0.1700

Data Index: 793257 (C4 (Web Text))

Max Activating Token Index: 908

Text #1

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.9413. Min Act: -0.1700

Data Index: 346316 (C4 (Web Text))

Max Activating Token Index: 968

Text #2

Max Range: 5.3475. Min Range: -5.3475

Max Act: 5.0038. Min Act: -0.1700

Data Index: 1638314 (Python Code)

Max Activating Token Index: 971

Text #3

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.9100. Min Act: -0.1700

Data Index: 1325500 (C4 (Web Text))

Max Activating Token Index: 936

Text #4

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.8788. Min Act: -0.1700

Data Index: 1025312 (C4 (Web Text))

Max Activating Token Index: 899

Text #5

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.9100. Min Act: -0.1700

Data Index: 440074 (C4 (Web Text))

Max Activating Token Index: 619

Text #6

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.8475. Min Act: -0.1700

Data Index: 235448 (C4 (Web Text))

Max Activating Token Index: 676

Text #7

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.8163. Min Act: -0.1699

Data Index: 1390169 (Python Code)

Max Activating Token Index: 954

Text #8

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7850. Min Act: -0.1700

Data Index: 457574 (C4 (Web Text))

Max Activating Token Index: 476

Text #9

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7850. Min Act: -0.1700

Data Index: 1521497 (Python Code)

Max Activating Token Index: 87

Text #10

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7850. Min Act: -0.1700

Data Index: 39498 (C4 (Web Text))

Max Activating Token Index: 507

Text #11

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7538. Min Act: -0.1700

Data Index: 916149 (C4 (Web Text))

Max Activating Token Index: 875

Text #12

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.6913. Min Act: -0.1700

Data Index: 142405 (C4 (Web Text))

Max Activating Token Index: 283

Text #13

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7538. Min Act: -0.1699

Data Index: 1624102 (Python Code)

Max Activating Token Index: 676

Text #14

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7850. Min Act: -0.1700

Data Index: 1399482 (Python Code)

Max Activating Token Index: 577

Text #15

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.6913. Min Act: -0.1700

Data Index: 1483966 (Python Code)

Max Activating Token Index: 613

Text #16

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.7225. Min Act: -0.1700

Data Index: 717255 (C4 (Web Text))

Max Activating Token Index: 240

Text #17

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.6600. Min Act: -0.1700

Data Index: 791445 (C4 (Web Text))

Max Activating Token Index: 363

Text #18

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.6131. Min Act: -0.1700

Data Index: 1002259 (C4 (Web Text))

Max Activating Token Index: 1001

Text #19

Max Range: 5.3475. Min Range: -5.3475

Max Act: 4.6600. Min Act: -0.1700

Data Index: 1524885 (Python Code)

Max Activating Token Index: 785