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

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

Neuron 1933 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: 8.5062. Min Range: -8.5062

Max Act: 8.5062. Min Act: -0.1700

Data Index: 1472161 (Python Code)

Max Activating Token Index: 415

Text #1

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.1937. Min Act: -0.1700

Data Index: 259343 (C4 (Web Text))

Max Activating Token Index: 342

Text #2

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.1312. Min Act: -0.1700

Data Index: 1373349 (Python Code)

Max Activating Token Index: 802

Text #3

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.0062. Min Act: -0.1700

Data Index: 1569996 (Python Code)

Max Activating Token Index: 787

Text #4

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.0687. Min Act: -0.1700

Data Index: 1476892 (Python Code)

Max Activating Token Index: 685

Text #5

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.0062. Min Act: -0.1699

Data Index: 1038160 (C4 (Web Text))

Max Activating Token Index: 74

Text #6

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.0062. Min Act: -0.1700

Data Index: 1520933 (Python Code)

Max Activating Token Index: 589

Text #7

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.0062. Min Act: -0.1700

Data Index: 68772 (C4 (Web Text))

Max Activating Token Index: 800

Text #8

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.8812. Min Act: -0.1700

Data Index: 744911 (C4 (Web Text))

Max Activating Token Index: 146

Text #9

Max Range: 8.5062. Min Range: -8.5062

Max Act: 8.0062. Min Act: -0.1700

Data Index: 1380875 (Python Code)

Max Activating Token Index: 414

Text #10

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.9437. Min Act: -0.1700

Data Index: 1507597 (Python Code)

Max Activating Token Index: 366

Text #11

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.9437. Min Act: -0.1700

Data Index: 159021 (C4 (Web Text))

Max Activating Token Index: 320

Text #12

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.8812. Min Act: -0.1700

Data Index: 394348 (C4 (Web Text))

Max Activating Token Index: 783

Text #13

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.8812. Min Act: -0.1700

Data Index: 1111398 (C4 (Web Text))

Max Activating Token Index: 882

Text #14

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.9437. Min Act: -0.1700

Data Index: 1091866 (C4 (Web Text))

Max Activating Token Index: 978

Text #15

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.9437. Min Act: -0.1700

Data Index: 1319510 (C4 (Web Text))

Max Activating Token Index: 720

Text #16

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.8812. Min Act: -0.1700

Data Index: 1475980 (Python Code)

Max Activating Token Index: 336

Text #17

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.8187. Min Act: -0.1700

Data Index: 1456447 (Python Code)

Max Activating Token Index: 387

Text #18

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.7562. Min Act: -0.1700

Data Index: 1243580 (C4 (Web Text))

Max Activating Token Index: 733

Text #19

Max Range: 8.5062. Min Range: -8.5062

Max Act: 7.6937. Min Act: -0.1700

Data Index: 1506343 (Python Code)

Max Activating Token Index: 376