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

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

Neuron 1460 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.6630. Min Act: -0.1700

Data Index: 784277 (C4 (Web Text))

Max Activating Token Index: 383

Text #1

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.6317. Min Act: -0.1700

Data Index: 1419543 (Python Code)

Max Activating Token Index: 811

Text #2

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.4907. Min Act: -0.1700

Data Index: 1114268 (C4 (Web Text))

Max Activating Token Index: 260

Text #3

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.4280. Min Act: -0.1700

Data Index: 1323958 (C4 (Web Text))

Max Activating Token Index: 876

Text #4

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.3967. Min Act: -0.1700

Data Index: 1277312 (C4 (Web Text))

Max Activating Token Index: 129

Text #5

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.3653. Min Act: -0.1700

Data Index: 93679 (C4 (Web Text))

Max Activating Token Index: 970

Text #6

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.3496. Min Act: -0.1700

Data Index: 640141 (C4 (Web Text))

Max Activating Token Index: 631

Text #7

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.3653. Min Act: -0.1700

Data Index: 113507 (C4 (Web Text))

Max Activating Token Index: 286

Text #8

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2711. Min Act: -0.1700

Data Index: 27678 (C4 (Web Text))

Max Activating Token Index: 582

Text #9

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.3339. Min Act: -0.1700

Data Index: 926495 (C4 (Web Text))

Max Activating Token Index: 432

Text #10

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2868. Min Act: -0.1700

Data Index: 510505 (C4 (Web Text))

Max Activating Token Index: 652

Text #11

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2396. Min Act: -0.1700

Data Index: 980379 (C4 (Web Text))

Max Activating Token Index: 725

Text #12

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2554. Min Act: -0.1700

Data Index: 493616 (C4 (Web Text))

Max Activating Token Index: 744

Text #13

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2239. Min Act: -0.1700

Data Index: 1540448 (Python Code)

Max Activating Token Index: 885

Text #14

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2554. Min Act: -0.1700

Data Index: 1571411 (Python Code)

Max Activating Token Index: 579

Text #15

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2554. Min Act: -0.1700

Data Index: 725271 (C4 (Web Text))

Max Activating Token Index: 868

Text #16

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2396. Min Act: -0.1700

Data Index: 526753 (C4 (Web Text))

Max Activating Token Index: 741

Text #17

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2239. Min Act: -0.1700

Data Index: 822054 (C4 (Web Text))

Max Activating Token Index: 759

Text #18

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2396. Min Act: -0.1700

Data Index: 1101252 (C4 (Web Text))

Max Activating Token Index: 128

Text #19

Max Range: 3.6630. Min Range: -3.6630

Max Act: 3.2554. Min Act: -0.1700

Data Index: 307330 (C4 (Web Text))

Max Activating Token Index: 286