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

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

Neuron 1426 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: 3.8536. Min Range: -3.8536

Max Act: 3.8536. Min Act: -0.1700

Data Index: 1511958 (Python Code)

Max Activating Token Index: 953

Text #1

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.7598. Min Act: -0.1700

Data Index: 1209586 (C4 (Web Text))

Max Activating Token Index: 141

Text #2

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.7285. Min Act: -0.1700

Data Index: 1566973 (Python Code)

Max Activating Token Index: 990

Text #3

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.6659. Min Act: -0.1699

Data Index: 846392 (C4 (Web Text))

Max Activating Token Index: 742

Text #4

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.6033. Min Act: -0.1700

Data Index: 1526025 (Python Code)

Max Activating Token Index: 880

Text #5

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.6033. Min Act: -0.1700

Data Index: 1629295 (Python Code)

Max Activating Token Index: 541

Text #6

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5720. Min Act: -0.1700

Data Index: 1424944 (Python Code)

Max Activating Token Index: 895

Text #7

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.6033. Min Act: -0.1699

Data Index: 1618146 (Python Code)

Max Activating Token Index: 856

Text #8

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5720. Min Act: -0.1700

Data Index: 997908 (C4 (Web Text))

Max Activating Token Index: 50

Text #9

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.6033. Min Act: -0.1700

Data Index: 86441 (C4 (Web Text))

Max Activating Token Index: 725

Text #10

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5093. Min Act: -0.1700

Data Index: 351242 (C4 (Web Text))

Max Activating Token Index: 519

Text #11

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5406. Min Act: -0.1700

Data Index: 1428960 (Python Code)

Max Activating Token Index: 825

Text #12

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5093. Min Act: -0.1700

Data Index: 749214 (C4 (Web Text))

Max Activating Token Index: 388

Text #13

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.4466. Min Act: -0.1700

Data Index: 416377 (C4 (Web Text))

Max Activating Token Index: 377

Text #14

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5093. Min Act: -0.1700

Data Index: 1616670 (Python Code)

Max Activating Token Index: 605

Text #15

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.4780. Min Act: -0.1700

Data Index: 1604218 (Python Code)

Max Activating Token Index: 343

Text #16

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5406. Min Act: -0.1700

Data Index: 1588037 (Python Code)

Max Activating Token Index: 877

Text #17

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5406. Min Act: -0.1700

Data Index: 470793 (C4 (Web Text))

Max Activating Token Index: 1007

Text #18

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5093. Min Act: -0.1700

Data Index: 1011176 (C4 (Web Text))

Max Activating Token Index: 753

Text #19

Max Range: 3.8536. Min Range: -3.8536

Max Act: 3.5406. Min Act: -0.1700

Data Index: 285073 (C4 (Web Text))

Max Activating Token Index: 556