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

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

Neuron 135 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: 5.0485. Min Range: -5.0485

Max Act: 5.0485. Min Act: -0.1700

Data Index: 536217 (C4 (Web Text))

Max Activating Token Index: 753

Text #1

Max Range: 5.0485. Min Range: -5.0485

Max Act: 5.0485. Min Act: -0.1700

Data Index: 1525563 (Python Code)

Max Activating Token Index: 504

Text #2

Max Range: 5.0485. Min Range: -5.0485

Max Act: 5.0485. Min Act: -0.1700

Data Index: 1488955 (Python Code)

Max Activating Token Index: 343

Text #3

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.9235. Min Act: -0.1700

Data Index: 173922 (C4 (Web Text))

Max Activating Token Index: 103

Text #4

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.8922. Min Act: -0.1700

Data Index: 1430230 (Python Code)

Max Activating Token Index: 310

Text #5

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.9860. Min Act: -0.1700

Data Index: 233275 (C4 (Web Text))

Max Activating Token Index: 427

Text #6

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.8610. Min Act: -0.1700

Data Index: 971253 (C4 (Web Text))

Max Activating Token Index: 832

Text #7

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.8297. Min Act: -0.1700

Data Index: 1654044 (Python Code)

Max Activating Token Index: 401

Text #8

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7985. Min Act: -0.1700

Data Index: 1502657 (Python Code)

Max Activating Token Index: 169

Text #9

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7985. Min Act: -0.1700

Data Index: 1005447 (C4 (Web Text))

Max Activating Token Index: 171

Text #10

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7360. Min Act: -0.1700

Data Index: 430906 (C4 (Web Text))

Max Activating Token Index: 78

Text #11

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7360. Min Act: -0.1700

Data Index: 1412923 (Python Code)

Max Activating Token Index: 770

Text #12

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.6735. Min Act: -0.1700

Data Index: 962681 (C4 (Web Text))

Max Activating Token Index: 197

Text #13

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7360. Min Act: -0.1700

Data Index: 624291 (C4 (Web Text))

Max Activating Token Index: 76

Text #14

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.6735. Min Act: -0.1700

Data Index: 1163081 (C4 (Web Text))

Max Activating Token Index: 168

Text #15

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.6422. Min Act: -0.1700

Data Index: 607389 (C4 (Web Text))

Max Activating Token Index: 282

Text #16

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.6735. Min Act: -0.1700

Data Index: 1366198 (Python Code)

Max Activating Token Index: 213

Text #17

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7047. Min Act: -0.1700

Data Index: 255436 (C4 (Web Text))

Max Activating Token Index: 120

Text #18

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.7047. Min Act: -0.1700

Data Index: 1306232 (C4 (Web Text))

Max Activating Token Index: 51

Text #19

Max Range: 5.0485. Min Range: -5.0485

Max Act: 4.6735. Min Act: -0.1700

Data Index: 878824 (C4 (Web Text))

Max Activating Token Index: 404