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

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

Neuron 1666 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')

Text #0

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.8367. Min Act: -0.1700

Data Index: 600340 (C4 (Web Text))

Max Activating Token Index: 728

Text #1

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.8367. Min Act: -0.1700

Data Index: 916943 (C4 (Web Text))

Max Activating Token Index: 353

Text #2

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.7741. Min Act: -0.1700

Data Index: 1607729 (Python Code)

Max Activating Token Index: 614

Text #3

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.7429. Min Act: -0.1700

Data Index: 1493665 (Python Code)

Max Activating Token Index: 649

Text #4

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.7116. Min Act: -0.1700

Data Index: 936081 (C4 (Web Text))

Max Activating Token Index: 644

Text #5

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.7429. Min Act: -0.1700

Data Index: 1655783 (Python Code)

Max Activating Token Index: 828

Text #6

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.7116. Min Act: -0.1700

Data Index: 1412956 (Python Code)

Max Activating Token Index: 96

Text #7

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.7116. Min Act: -0.1700

Data Index: 629056 (C4 (Web Text))

Max Activating Token Index: 738

Text #8

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.6490. Min Act: -0.1700

Data Index: 287738 (C4 (Web Text))

Max Activating Token Index: 561

Text #9

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.6490. Min Act: -0.1700

Data Index: 699813 (C4 (Web Text))

Max Activating Token Index: 159

Text #10

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.6177. Min Act: -0.1700

Data Index: 1601504 (Python Code)

Max Activating Token Index: 514

Text #11

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5550. Min Act: -0.1700

Data Index: 1370250 (Python Code)

Max Activating Token Index: 332

Text #12

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5863. Min Act: -0.1700

Data Index: 126856 (C4 (Web Text))

Max Activating Token Index: 470

Text #13

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5863. Min Act: -0.1700

Data Index: 526224 (C4 (Web Text))

Max Activating Token Index: 60

Text #14

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5550. Min Act: -0.1700

Data Index: 425055 (C4 (Web Text))

Max Activating Token Index: 142

Text #15

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5550. Min Act: -0.1700

Data Index: 1180257 (C4 (Web Text))

Max Activating Token Index: 524

Text #16

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5863. Min Act: -0.1700

Data Index: 1474793 (Python Code)

Max Activating Token Index: 604

Text #17

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5550. Min Act: -0.1700

Data Index: 352812 (C4 (Web Text))

Max Activating Token Index: 851

Text #18

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.5550. Min Act: -0.1700

Data Index: 49392 (C4 (Web Text))

Max Activating Token Index: 853

Text #19

Max Range: 3.8367. Min Range: -3.8367

Max Act: 3.4924. Min Act: -0.1700

Data Index: 377486 (C4 (Web Text))

Max Activating Token Index: 252