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

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

Neuron 1524 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: 5.3835. Min Range: -5.3835

Max Act: 5.3835. Min Act: -0.1700

Data Index: 214086 (C4 (Web Text))

Max Activating Token Index: 566

Text #1

Max Range: 5.3835. Min Range: -5.3835

Max Act: 5.3679. Min Act: -0.1700

Data Index: 1629488 (Python Code)

Max Activating Token Index: 940

Text #2

Max Range: 5.3835. Min Range: -5.3835

Max Act: 5.1179. Min Act: -0.1700

Data Index: 1289149 (C4 (Web Text))

Max Activating Token Index: 996

Text #3

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.9616. Min Act: -0.1700

Data Index: 776735 (C4 (Web Text))

Max Activating Token Index: 416

Text #4

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.9772. Min Act: -0.1700

Data Index: 416387 (C4 (Web Text))

Max Activating Token Index: 344

Text #5

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.9147. Min Act: -0.1682

Data Index: 1422288 (Python Code)

Max Activating Token Index: 677

Text #6

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.9304. Min Act: -0.1700

Data Index: 1004988 (C4 (Web Text))

Max Activating Token Index: 639

Text #7

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.9460. Min Act: -0.1700

Data Index: 573584 (C4 (Web Text))

Max Activating Token Index: 68

Text #8

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8835. Min Act: -0.1485

Data Index: 1463510 (Python Code)

Max Activating Token Index: 263

Text #9

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8835. Min Act: -0.1700

Data Index: 800216 (C4 (Web Text))

Max Activating Token Index: 400

Text #10

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8991. Min Act: -0.1682

Data Index: 1471366 (Python Code)

Max Activating Token Index: 862

Text #11

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8678. Min Act: -0.1700

Data Index: 1055064 (C4 (Web Text))

Max Activating Token Index: 477

Text #12

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.9147. Min Act: -0.1698

Data Index: 1480304 (Python Code)

Max Activating Token Index: 350

Text #13

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8678. Min Act: -0.1700

Data Index: 755278 (C4 (Web Text))

Max Activating Token Index: 337

Text #14

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8678. Min Act: -0.1700

Data Index: 1013043 (C4 (Web Text))

Max Activating Token Index: 220

Text #15

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8366. Min Act: -0.1700

Data Index: 378542 (C4 (Web Text))

Max Activating Token Index: 727

Text #16

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8835. Min Act: -0.1700

Data Index: 886745 (C4 (Web Text))

Max Activating Token Index: 462

Text #17

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.7741. Min Act: -0.1699

Data Index: 1442662 (Python Code)

Max Activating Token Index: 423

Text #18

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.8522. Min Act: -0.1699

Data Index: 1537384 (Python Code)

Max Activating Token Index: 511

Text #19

Max Range: 5.3835. Min Range: -5.3835

Max Act: 4.7897. Min Act: -0.1699

Data Index: 1635465 (Python Code)

Max Activating Token Index: 402