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

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

Neuron 70 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: 5.2617. Min Range: -5.2617

Max Act: 5.2461. Min Act: -0.1700

Data Index: 1200538 (C4 (Web Text))

Max Activating Token Index: 75

Text #1

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.2617. Min Act: -0.1700

Data Index: 1368746 (Python Code)

Max Activating Token Index: 807

Text #2

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0430. Min Act: -0.1700

Data Index: 1168463 (C4 (Web Text))

Max Activating Token Index: 183

Text #3

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0742. Min Act: -0.1700

Data Index: 1059931 (C4 (Web Text))

Max Activating Token Index: 299

Text #4

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0742. Min Act: -0.1700

Data Index: 1187033 (C4 (Web Text))

Max Activating Token Index: 223

Text #5

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0430. Min Act: -0.1700

Data Index: 82355 (C4 (Web Text))

Max Activating Token Index: 333

Text #6

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0274. Min Act: -0.1700

Data Index: 494816 (C4 (Web Text))

Max Activating Token Index: 883

Text #7

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.1055. Min Act: -0.1700

Data Index: 716726 (C4 (Web Text))

Max Activating Token Index: 311

Text #8

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0117. Min Act: -0.1700

Data Index: 35449 (C4 (Web Text))

Max Activating Token Index: 198

Text #9

Max Range: 5.2617. Min Range: -5.2617

Max Act: 5.0274. Min Act: -0.1700

Data Index: 1129920 (C4 (Web Text))

Max Activating Token Index: 852

Text #10

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.9805. Min Act: -0.1700

Data Index: 1386589 (Python Code)

Max Activating Token Index: 932

Text #11

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.9961. Min Act: -0.1700

Data Index: 782125 (C4 (Web Text))

Max Activating Token Index: 340

Text #12

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.9336. Min Act: -0.1700

Data Index: 99704 (C4 (Web Text))

Max Activating Token Index: 739

Text #13

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.9180. Min Act: -0.1700

Data Index: 1649648 (Python Code)

Max Activating Token Index: 954

Text #14

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.8867. Min Act: -0.1700

Data Index: 9215 (C4 (Web Text))

Max Activating Token Index: 590

Text #15

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.8242. Min Act: -0.1700

Data Index: 72571 (C4 (Web Text))

Max Activating Token Index: 862

Text #16

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.8399. Min Act: -0.1700

Data Index: 1278026 (C4 (Web Text))

Max Activating Token Index: 719

Text #17

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.8242. Min Act: -0.1700

Data Index: 348368 (C4 (Web Text))

Max Activating Token Index: 933

Text #18

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.8242. Min Act: -0.1700

Data Index: 1034562 (C4 (Web Text))

Max Activating Token Index: 908

Text #19

Max Range: 5.2617. Min Range: -5.2617

Max Act: 4.7617. Min Act: -0.1700

Data Index: 790983 (C4 (Web Text))

Max Activating Token Index: 290