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

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

Neuron 14 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: 5.3920. Min Range: -5.3920

Max Act: 5.3920. Min Act: -0.1700

Data Index: 425305 (C4 (Web Text))

Max Activating Token Index: 574

Text #1

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.2670. Min Act: -0.1700

Data Index: 619304 (C4 (Web Text))

Max Activating Token Index: 327

Text #2

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.2670. Min Act: -0.1700

Data Index: 739356 (C4 (Web Text))

Max Activating Token Index: 823

Text #3

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.2045. Min Act: -0.1700

Data Index: 1238430 (C4 (Web Text))

Max Activating Token Index: 887

Text #4

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.2670. Min Act: -0.1700

Data Index: 1188681 (C4 (Web Text))

Max Activating Token Index: 440

Text #5

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0951. Min Act: -0.1700

Data Index: 374473 (C4 (Web Text))

Max Activating Token Index: 646

Text #6

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.1263. Min Act: -0.1700

Data Index: 1493490 (Python Code)

Max Activating Token Index: 169

Text #7

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.1263. Min Act: -0.1700

Data Index: 1242470 (C4 (Web Text))

Max Activating Token Index: 428

Text #8

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0795. Min Act: -0.1700

Data Index: 562394 (C4 (Web Text))

Max Activating Token Index: 505

Text #9

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.1576. Min Act: -0.1700

Data Index: 770437 (C4 (Web Text))

Max Activating Token Index: 636

Text #10

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.1263. Min Act: -0.1700

Data Index: 440834 (C4 (Web Text))

Max Activating Token Index: 200

Text #11

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.1107. Min Act: -0.1700

Data Index: 496804 (C4 (Web Text))

Max Activating Token Index: 261

Text #12

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0638. Min Act: -0.1700

Data Index: 592441 (C4 (Web Text))

Max Activating Token Index: 199

Text #13

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0326. Min Act: -0.1700

Data Index: 796377 (C4 (Web Text))

Max Activating Token Index: 366

Text #14

Max Range: 5.3920. Min Range: -5.3920

Max Act: 4.9545. Min Act: -0.1700

Data Index: 1323803 (C4 (Web Text))

Max Activating Token Index: 521

Text #15

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0013. Min Act: -0.1700

Data Index: 1547568 (Python Code)

Max Activating Token Index: 690

Text #16

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0170. Min Act: -0.1700

Data Index: 1250703 (C4 (Web Text))

Max Activating Token Index: 228

Text #17

Max Range: 5.3920. Min Range: -5.3920

Max Act: 5.0326. Min Act: -0.1700

Data Index: 809360 (C4 (Web Text))

Max Activating Token Index: 970

Text #18

Max Range: 5.3920. Min Range: -5.3920

Max Act: 4.9701. Min Act: -0.1700

Data Index: 268730 (C4 (Web Text))

Max Activating Token Index: 373

Text #19

Max Range: 5.3920. Min Range: -5.3920

Max Act: 4.9701. Min Act: -0.1700

Data Index: 132628 (C4 (Web Text))

Max Activating Token Index: 408