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

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

Neuron 236 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: 3.6952. Min Range: -3.6952

Max Act: 3.6952. Min Act: -0.1700

Data Index: 1617625 (Python Code)

Max Activating Token Index: 219

Text #1

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.6483. Min Act: -0.1437

Data Index: 1431852 (Python Code)

Max Activating Token Index: 428

Text #2

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.6326. Min Act: -0.1700

Data Index: 441084 (C4 (Web Text))

Max Activating Token Index: 359

Text #3

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.6483. Min Act: -0.1696

Data Index: 1442316 (Python Code)

Max Activating Token Index: 423

Text #4

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5700. Min Act: -0.1644

Data Index: 487595 (C4 (Web Text))

Max Activating Token Index: 917

Text #5

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5856. Min Act: -0.1700

Data Index: 867952 (C4 (Web Text))

Max Activating Token Index: 381

Text #6

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5387. Min Act: -0.1700

Data Index: 82013 (C4 (Web Text))

Max Activating Token Index: 369

Text #7

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5230. Min Act: -0.1679

Data Index: 426269 (C4 (Web Text))

Max Activating Token Index: 556

Text #8

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5230. Min Act: -0.1700

Data Index: 1388201 (Python Code)

Max Activating Token Index: 654

Text #9

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5387. Min Act: -0.1700

Data Index: 92245 (C4 (Web Text))

Max Activating Token Index: 227

Text #10

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5387. Min Act: -0.1700

Data Index: 1031883 (C4 (Web Text))

Max Activating Token Index: 612

Text #11

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5230. Min Act: -0.1700

Data Index: 1640991 (Python Code)

Max Activating Token Index: 286

Text #12

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5073. Min Act: -0.0929

Data Index: 1616400 (Python Code)

Max Activating Token Index: 679

Text #13

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5230. Min Act: -0.1700

Data Index: 1215015 (C4 (Web Text))

Max Activating Token Index: 943

Text #14

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5073. Min Act: -0.1699

Data Index: 531414 (C4 (Web Text))

Max Activating Token Index: 950

Text #15

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5230. Min Act: -0.1699

Data Index: 1599589 (Python Code)

Max Activating Token Index: 660

Text #16

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.4917. Min Act: -0.0929

Data Index: 1481576 (Python Code)

Max Activating Token Index: 627

Text #17

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.4760. Min Act: -0.1700

Data Index: 1556895 (Python Code)

Max Activating Token Index: 735

Text #18

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.4917. Min Act: -0.1610

Data Index: 203751 (C4 (Web Text))

Max Activating Token Index: 535

Text #19

Max Range: 3.6952. Min Range: -3.6952

Max Act: 3.5073. Min Act: -0.1699

Data Index: 653443 (C4 (Web Text))

Max Activating Token Index: 883