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

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

Neuron 64 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: 3.0376. Min Range: -3.0376

Max Act: 3.0376. Min Act: -0.1700

Data Index: 126559 (C4 (Web Text))

Max Activating Token Index: 794

Text #1

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.9586. Min Act: -0.1700

Data Index: 491583 (C4 (Web Text))

Max Activating Token Index: 809

Text #2

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 1421193 (Python Code)

Max Activating Token Index: 439

Text #3

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.9269. Min Act: -0.1700

Data Index: 52263 (C4 (Web Text))

Max Activating Token Index: 509

Text #4

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 1654760 (Python Code)

Max Activating Token Index: 665

Text #5

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 407014 (C4 (Web Text))

Max Activating Token Index: 661

Text #6

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.9111. Min Act: -0.1700

Data Index: 926522 (C4 (Web Text))

Max Activating Token Index: 486

Text #7

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 334503 (C4 (Web Text))

Max Activating Token Index: 333

Text #8

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8793. Min Act: -0.1700

Data Index: 857749 (C4 (Web Text))

Max Activating Token Index: 595

Text #9

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8952. Min Act: -0.1700

Data Index: 274020 (C4 (Web Text))

Max Activating Token Index: 106

Text #10

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 442818 (C4 (Web Text))

Max Activating Token Index: 505

Text #11

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8793. Min Act: -0.1700

Data Index: 64129 (C4 (Web Text))

Max Activating Token Index: 825

Text #12

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 517724 (C4 (Web Text))

Max Activating Token Index: 977

Text #13

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8317. Min Act: -0.1700

Data Index: 174258 (C4 (Web Text))

Max Activating Token Index: 838

Text #14

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8476. Min Act: -0.1700

Data Index: 824788 (C4 (Web Text))

Max Activating Token Index: 331

Text #15

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8635. Min Act: -0.1700

Data Index: 799828 (C4 (Web Text))

Max Activating Token Index: 481

Text #16

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8952. Min Act: -0.1700

Data Index: 1479381 (Python Code)

Max Activating Token Index: 413

Text #17

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8476. Min Act: -0.1700

Data Index: 788219 (C4 (Web Text))

Max Activating Token Index: 868

Text #18

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8952. Min Act: -0.1700

Data Index: 1057105 (C4 (Web Text))

Max Activating Token Index: 736

Text #19

Max Range: 3.0376. Min Range: -3.0376

Max Act: 2.8476. Min Act: -0.1700

Data Index: 618416 (C4 (Web Text))

Max Activating Token Index: 908