Model: GELU Model: 3 Layers, 2048 Neurons per Layer

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

Neuron 1539 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-3l')

Text #0

Max Range: 4.2617. Min Range: -4.2617

Max Act: 4.2617. Min Act: -0.1700

Data Index: 1244425 (C4 (Web Text))

Max Activating Token Index: 781

Text #1

Max Range: 4.2617. Min Range: -4.2617

Max Act: 4.1680. Min Act: -0.1700

Data Index: 193460 (C4 (Web Text))

Max Activating Token Index: 776

Text #2

Max Range: 4.2617. Min Range: -4.2617

Max Act: 4.1680. Min Act: -0.1700

Data Index: 633078 (C4 (Web Text))

Max Activating Token Index: 794

Text #3

Max Range: 4.2617. Min Range: -4.2617

Max Act: 4.0429. Min Act: -0.1694

Data Index: 1435115 (Python Code)

Max Activating Token Index: 995

Text #4

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.9491. Min Act: -0.1700

Data Index: 1573370 (Python Code)

Max Activating Token Index: 1015

Text #5

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.9179. Min Act: -0.1700

Data Index: 270564 (C4 (Web Text))

Max Activating Token Index: 833

Text #6

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.9179. Min Act: -0.1700

Data Index: 413888 (C4 (Web Text))

Max Activating Token Index: 941

Text #7

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.9179. Min Act: -0.1698

Data Index: 1393988 (Python Code)

Max Activating Token Index: 727

Text #8

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8866. Min Act: -0.1644

Data Index: 1156379 (C4 (Web Text))

Max Activating Token Index: 613

Text #9

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8866. Min Act: -0.1700

Data Index: 596061 (C4 (Web Text))

Max Activating Token Index: 1016

Text #10

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.9022. Min Act: -0.1700

Data Index: 1327235 (C4 (Web Text))

Max Activating Token Index: 444

Text #11

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8397. Min Act: -0.1700

Data Index: 1644535 (Python Code)

Max Activating Token Index: 396

Text #12

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8553. Min Act: -0.1700

Data Index: 114604 (C4 (Web Text))

Max Activating Token Index: 32

Text #13

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8553. Min Act: -0.1700

Data Index: 467823 (C4 (Web Text))

Max Activating Token Index: 413

Text #14

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8397. Min Act: -0.1700

Data Index: 1601337 (Python Code)

Max Activating Token Index: 787

Text #15

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8084. Min Act: -0.1700

Data Index: 535532 (C4 (Web Text))

Max Activating Token Index: 704

Text #16

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8240. Min Act: -0.1700

Data Index: 450069 (C4 (Web Text))

Max Activating Token Index: 801

Text #17

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8240. Min Act: -0.1700

Data Index: 691293 (C4 (Web Text))

Max Activating Token Index: 692

Text #18

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.7771. Min Act: -0.1700

Data Index: 735214 (C4 (Web Text))

Max Activating Token Index: 232

Text #19

Max Range: 4.2617. Min Range: -4.2617

Max Act: 3.8084. Min Act: -0.1700

Data Index: 821445 (C4 (Web Text))

Max Activating Token Index: 223