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

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

Neuron 1608 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: 3.5948. Min Range: -3.5948

Max Act: 3.5948. Min Act: -0.1700

Data Index: 51884 (C4 (Web Text))

Max Activating Token Index: 983

Text #1

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4381. Min Act: -0.1700

Data Index: 1271755 (C4 (Web Text))

Max Activating Token Index: 813

Text #2

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4068. Min Act: -0.1700

Data Index: 1412471 (Python Code)

Max Activating Token Index: 615

Text #3

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4695. Min Act: -0.1700

Data Index: 1169850 (C4 (Web Text))

Max Activating Token Index: 232

Text #4

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4695. Min Act: -0.1700

Data Index: 1394506 (Python Code)

Max Activating Token Index: 904

Text #5

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4381. Min Act: -0.1700

Data Index: 443442 (C4 (Web Text))

Max Activating Token Index: 389

Text #6

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4068. Min Act: -0.1700

Data Index: 1345801 (C4 (Web Text))

Max Activating Token Index: 41

Text #7

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4068. Min Act: -0.1700

Data Index: 1541921 (Python Code)

Max Activating Token Index: 410

Text #8

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3283. Min Act: -0.1700

Data Index: 1616323 (Python Code)

Max Activating Token Index: 818

Text #9

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3597. Min Act: -0.1700

Data Index: 1019474 (C4 (Web Text))

Max Activating Token Index: 105

Text #10

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.2969. Min Act: -0.1700

Data Index: 824652 (C4 (Web Text))

Max Activating Token Index: 981

Text #11

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3440. Min Act: -0.1700

Data Index: 1180810 (C4 (Web Text))

Max Activating Token Index: 120

Text #12

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.4068. Min Act: -0.1700

Data Index: 585064 (C4 (Web Text))

Max Activating Token Index: 380

Text #13

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3597. Min Act: -0.1700

Data Index: 1590399 (Python Code)

Max Activating Token Index: 814

Text #14

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3283. Min Act: -0.1700

Data Index: 378648 (C4 (Web Text))

Max Activating Token Index: 686

Text #15

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3440. Min Act: -0.1700

Data Index: 164568 (C4 (Web Text))

Max Activating Token Index: 702

Text #16

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3283. Min Act: -0.1700

Data Index: 1368240 (Python Code)

Max Activating Token Index: 669

Text #17

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.3126. Min Act: -0.1700

Data Index: 1020542 (C4 (Web Text))

Max Activating Token Index: 356

Text #18

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.2655. Min Act: -0.1700

Data Index: 586180 (C4 (Web Text))

Max Activating Token Index: 956

Text #19

Max Range: 3.5948. Min Range: -3.5948

Max Act: 3.2969. Min Act: -0.1700

Data Index: 1599841 (Python Code)

Max Activating Token Index: 233