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

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

Neuron 1616 in Layer 1

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-3l')

Text #0

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.2061. Min Act: -0.1700

Data Index: 110475 (C4 (Web Text))

Max Activating Token Index: 536

Text #1

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.2532. Min Act: -0.1700

Data Index: 1371120 (Python Code)

Max Activating Token Index: 913

Text #2

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.1746. Min Act: -0.1700

Data Index: 549956 (C4 (Web Text))

Max Activating Token Index: 792

Text #3

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.1589. Min Act: -0.1700

Data Index: 1585858 (Python Code)

Max Activating Token Index: 990

Text #4

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.1274. Min Act: -0.1700

Data Index: 1574708 (Python Code)

Max Activating Token Index: 904

Text #5

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.1274. Min Act: -0.1700

Data Index: 325438 (C4 (Web Text))

Max Activating Token Index: 741

Text #6

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0958. Min Act: -0.1700

Data Index: 1513089 (Python Code)

Max Activating Token Index: 299

Text #7

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0643. Min Act: -0.1700

Data Index: 363328 (C4 (Web Text))

Max Activating Token Index: 883

Text #8

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0327. Min Act: -0.1700

Data Index: 1156387 (C4 (Web Text))

Max Activating Token Index: 939

Text #9

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0485. Min Act: -0.1700

Data Index: 131985 (C4 (Web Text))

Max Activating Token Index: 955

Text #10

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0485. Min Act: -0.1700

Data Index: 306040 (C4 (Web Text))

Max Activating Token Index: 163

Text #11

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0643. Min Act: -0.1700

Data Index: 922753 (C4 (Web Text))

Max Activating Token Index: 492

Text #12

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0011. Min Act: -0.1700

Data Index: 20128 (C4 (Web Text))

Max Activating Token Index: 745

Text #13

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0327. Min Act: -0.1700

Data Index: 1544146 (Python Code)

Max Activating Token Index: 605

Text #14

Max Range: 3.2532. Min Range: -3.2532

Max Act: 2.9853. Min Act: -0.1700

Data Index: 80791 (C4 (Web Text))

Max Activating Token Index: 887

Text #15

Max Range: 3.2532. Min Range: -3.2532

Max Act: 2.9536. Min Act: -0.1700

Data Index: 1643924 (Python Code)

Max Activating Token Index: 550

Text #16

Max Range: 3.2532. Min Range: -3.2532

Max Act: 3.0011. Min Act: -0.1700

Data Index: 721266 (C4 (Web Text))

Max Activating Token Index: 777

Text #17

Max Range: 3.2532. Min Range: -3.2532

Max Act: 2.9536. Min Act: -0.1700

Data Index: 152077 (C4 (Web Text))

Max Activating Token Index: 990

Text #18

Max Range: 3.2532. Min Range: -3.2532

Max Act: 2.9536. Min Act: -0.1700

Data Index: 1096815 (C4 (Web Text))

Max Activating Token Index: 681

Text #19

Max Range: 3.2532. Min Range: -3.2532

Max Act: 2.9378. Min Act: -0.1700

Data Index: 1020726 (C4 (Web Text))

Max Activating Token Index: 705