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

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

Neuron 896 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: 8.0596. Min Range: -8.0596

Max Act: 8.0596. Min Act: -0.1700

Data Index: 623869 (C4 (Web Text))

Max Activating Token Index: 614

Text #1

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.6221. Min Act: -0.1700

Data Index: 211920 (C4 (Web Text))

Max Activating Token Index: 319

Text #2

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.5283. Min Act: -0.1700

Data Index: 557485 (C4 (Web Text))

Max Activating Token Index: 740

Text #3

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.5283. Min Act: -0.1700

Data Index: 842281 (C4 (Web Text))

Max Activating Token Index: 162

Text #4

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.3096. Min Act: -0.1700

Data Index: 173756 (C4 (Web Text))

Max Activating Token Index: 782

Text #5

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.2158. Min Act: -0.1700

Data Index: 1546539 (Python Code)

Max Activating Token Index: 813

Text #6

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.1533. Min Act: -0.1700

Data Index: 441859 (C4 (Web Text))

Max Activating Token Index: 801

Text #7

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.1846. Min Act: -0.1700

Data Index: 669075 (C4 (Web Text))

Max Activating Token Index: 928

Text #8

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.1846. Min Act: -0.1700

Data Index: 687901 (C4 (Web Text))

Max Activating Token Index: 72

Text #9

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.0908. Min Act: -0.1700

Data Index: 699962 (C4 (Web Text))

Max Activating Token Index: 874

Text #10

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.0283. Min Act: -0.1700

Data Index: 78297 (C4 (Web Text))

Max Activating Token Index: 327

Text #11

Max Range: 8.0596. Min Range: -8.0596

Max Act: 7.0596. Min Act: -0.1700

Data Index: 709690 (C4 (Web Text))

Max Activating Token Index: 925

Text #12

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.9658. Min Act: -0.1700

Data Index: 227764 (C4 (Web Text))

Max Activating Token Index: 255

Text #13

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.9346. Min Act: -0.1700

Data Index: 1244508 (C4 (Web Text))

Max Activating Token Index: 69

Text #14

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.8721. Min Act: -0.1700

Data Index: 1466797 (Python Code)

Max Activating Token Index: 599

Text #15

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.8096. Min Act: -0.1699

Data Index: 1146371 (C4 (Web Text))

Max Activating Token Index: 625

Text #16

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.8408. Min Act: -0.1700

Data Index: 1421649 (Python Code)

Max Activating Token Index: 721

Text #17

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.8096. Min Act: -0.1700

Data Index: 299577 (C4 (Web Text))

Max Activating Token Index: 472

Text #18

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.8721. Min Act: -0.1700

Data Index: 895339 (C4 (Web Text))

Max Activating Token Index: 652

Text #19

Max Range: 8.0596. Min Range: -8.0596

Max Act: 6.7471. Min Act: -0.1700

Data Index: 1160291 (C4 (Web Text))

Max Activating Token Index: 902