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

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

Neuron 1444 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')

Text #0

Max Range: 5.0168. Min Range: -5.0168

Max Act: 5.0168. Min Act: -0.1700

Data Index: 506120 (C4 (Web Text))

Max Activating Token Index: 434

Text #1

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.6418. Min Act: -0.1699

Data Index: 934806 (C4 (Web Text))

Max Activating Token Index: 838

Text #2

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.7043. Min Act: -0.1700

Data Index: 1304685 (C4 (Web Text))

Max Activating Token Index: 668

Text #3

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.6106. Min Act: -0.1700

Data Index: 1474469 (Python Code)

Max Activating Token Index: 645

Text #4

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.6418. Min Act: -0.1699

Data Index: 1187449 (C4 (Web Text))

Max Activating Token Index: 317

Text #5

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.6418. Min Act: -0.1700

Data Index: 526617 (C4 (Web Text))

Max Activating Token Index: 707

Text #6

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.5793. Min Act: -0.1700

Data Index: 1230297 (C4 (Web Text))

Max Activating Token Index: 614

Text #7

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.5793. Min Act: -0.1700

Data Index: 1513148 (Python Code)

Max Activating Token Index: 223

Text #8

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4855. Min Act: -0.1700

Data Index: 1543634 (Python Code)

Max Activating Token Index: 665

Text #9

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.5168. Min Act: -0.1698

Data Index: 708175 (C4 (Web Text))

Max Activating Token Index: 637

Text #10

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4855. Min Act: -0.1698

Data Index: 50363 (C4 (Web Text))

Max Activating Token Index: 785

Text #11

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.5168. Min Act: -0.1699

Data Index: 691932 (C4 (Web Text))

Max Activating Token Index: 861

Text #12

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4855. Min Act: -0.1696

Data Index: 396211 (C4 (Web Text))

Max Activating Token Index: 741

Text #13

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4543. Min Act: -0.1700

Data Index: 39167 (C4 (Web Text))

Max Activating Token Index: 177

Text #14

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4230. Min Act: -0.1700

Data Index: 1486666 (Python Code)

Max Activating Token Index: 342

Text #15

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4230. Min Act: -0.1698

Data Index: 1410404 (Python Code)

Max Activating Token Index: 315

Text #16

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4230. Min Act: -0.1700

Data Index: 510904 (C4 (Web Text))

Max Activating Token Index: 464

Text #17

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4855. Min Act: -0.1700

Data Index: 385728 (C4 (Web Text))

Max Activating Token Index: 674

Text #18

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.3918. Min Act: -0.1700

Data Index: 486021 (C4 (Web Text))

Max Activating Token Index: 594

Text #19

Max Range: 5.0168. Min Range: -5.0168

Max Act: 4.4230. Min Act: -0.1700

Data Index: 530701 (C4 (Web Text))

Max Activating Token Index: 605