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

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

Neuron 2036 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.3502. Min Act: -0.1700

Data Index: 997891 (C4 (Web Text))

Max Activating Token Index: 859

Text #1

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2717. Min Act: -0.1700

Data Index: 160846 (C4 (Web Text))

Max Activating Token Index: 947

Text #2

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2403. Min Act: -0.1700

Data Index: 463469 (C4 (Web Text))

Max Activating Token Index: 682

Text #3

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2403. Min Act: -0.1700

Data Index: 489400 (C4 (Web Text))

Max Activating Token Index: 939

Text #4

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1931. Min Act: -0.1700

Data Index: 1537836 (Python Code)

Max Activating Token Index: 389

Text #5

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1931. Min Act: -0.1700

Data Index: 763598 (C4 (Web Text))

Max Activating Token Index: 831

Text #6

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2246. Min Act: -0.1700

Data Index: 1282581 (C4 (Web Text))

Max Activating Token Index: 338

Text #7

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2403. Min Act: -0.1700

Data Index: 1649478 (Python Code)

Max Activating Token Index: 486

Text #8

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2088. Min Act: -0.1700

Data Index: 1576639 (Python Code)

Max Activating Token Index: 525

Text #9

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1931. Min Act: -0.1700

Data Index: 1124491 (C4 (Web Text))

Max Activating Token Index: 342

Text #10

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1774. Min Act: -0.1700

Data Index: 1413934 (Python Code)

Max Activating Token Index: 161

Text #11

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.2088. Min Act: -0.1700

Data Index: 772968 (C4 (Web Text))

Max Activating Token Index: 985

Text #12

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1616. Min Act: -0.1700

Data Index: 820255 (C4 (Web Text))

Max Activating Token Index: 920

Text #13

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1774. Min Act: -0.1700

Data Index: 1214104 (C4 (Web Text))

Max Activating Token Index: 946

Text #14

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1774. Min Act: -0.1700

Data Index: 559726 (C4 (Web Text))

Max Activating Token Index: 559

Text #15

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1459. Min Act: -0.1700

Data Index: 862656 (C4 (Web Text))

Max Activating Token Index: 407

Text #16

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1616. Min Act: -0.1700

Data Index: 470700 (C4 (Web Text))

Max Activating Token Index: 789

Text #17

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1616. Min Act: -0.1700

Data Index: 1570986 (Python Code)

Max Activating Token Index: 422

Text #18

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1774. Min Act: -0.1700

Data Index: 476030 (C4 (Web Text))

Max Activating Token Index: 956

Text #19

Max Range: 3.3502. Min Range: -3.3502

Max Act: 3.1459. Min Act: -0.1700

Data Index: 1494657 (Python Code)

Max Activating Token Index: 368