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

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

Neuron 805 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: 3.6917. Min Range: -3.6917

Max Act: 3.6917. Min Act: -0.1700

Data Index: 386150 (C4 (Web Text))

Max Activating Token Index: 294

Text #1

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.5665. Min Act: -0.1700

Data Index: 801155 (C4 (Web Text))

Max Activating Token Index: 196

Text #2

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.5665. Min Act: -0.1700

Data Index: 348954 (C4 (Web Text))

Max Activating Token Index: 480

Text #3

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.4881. Min Act: -0.1700

Data Index: 836250 (C4 (Web Text))

Max Activating Token Index: 12

Text #4

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.4568. Min Act: -0.1700

Data Index: 342857 (C4 (Web Text))

Max Activating Token Index: 82

Text #5

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.4098. Min Act: -0.1700

Data Index: 52425 (C4 (Web Text))

Max Activating Token Index: 463

Text #6

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.4411. Min Act: -0.1700

Data Index: 323097 (C4 (Web Text))

Max Activating Token Index: 15

Text #7

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3784. Min Act: -0.1700

Data Index: 1519474 (Python Code)

Max Activating Token Index: 43

Text #8

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.4098. Min Act: -0.1700

Data Index: 1093132 (C4 (Web Text))

Max Activating Token Index: 599

Text #9

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.4098. Min Act: -0.1700

Data Index: 202056 (C4 (Web Text))

Max Activating Token Index: 253

Text #10

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3627. Min Act: -0.1700

Data Index: 419209 (C4 (Web Text))

Max Activating Token Index: 190

Text #11

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3470. Min Act: -0.1700

Data Index: 1383952 (Python Code)

Max Activating Token Index: 621

Text #12

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3627. Min Act: -0.1700

Data Index: 1523063 (Python Code)

Max Activating Token Index: 66

Text #13

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3313. Min Act: -0.1700

Data Index: 1381969 (Python Code)

Max Activating Token Index: 487

Text #14

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3470. Min Act: -0.1700

Data Index: 1468107 (Python Code)

Max Activating Token Index: 308

Text #15

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.2999. Min Act: -0.1700

Data Index: 392556 (C4 (Web Text))

Max Activating Token Index: 334

Text #16

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3784. Min Act: -0.1700

Data Index: 1105072 (C4 (Web Text))

Max Activating Token Index: 240

Text #17

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3313. Min Act: -0.1700

Data Index: 403048 (C4 (Web Text))

Max Activating Token Index: 252

Text #18

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3313. Min Act: -0.1700

Data Index: 476103 (C4 (Web Text))

Max Activating Token Index: 858

Text #19

Max Range: 3.6917. Min Range: -3.6917

Max Act: 3.3156. Min Act: -0.1700

Data Index: 1411106 (Python Code)

Max Activating Token Index: 406