Model: SoLU Model: 12 Layers, 6144 Neurons per Layer

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

Neuron 4518 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-12l')

Text #0

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0474. Min Act: -0.0001

Data Index: 631746 (C4 (Web Text))

Max Activating Token Index: 809

Text #1

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0477. Min Act: -0.0001

Data Index: 1013356 (C4 (Web Text))

Max Activating Token Index: 447

Text #2

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0473. Min Act: -0.0001

Data Index: 1214134 (C4 (Web Text))

Max Activating Token Index: 11

Text #3

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0417. Min Act: -0.0001

Data Index: 388226 (C4 (Web Text))

Max Activating Token Index: 588

Text #4

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0420. Min Act: -0.0001

Data Index: 1396156 (Python Code)

Max Activating Token Index: 441

Text #5

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0447. Min Act: -0.0001

Data Index: 1194252 (C4 (Web Text))

Max Activating Token Index: 429

Text #6

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0417. Min Act: -0.0001

Data Index: 1404707 (Python Code)

Max Activating Token Index: 10

Text #7

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0352. Min Act: -0.0001

Data Index: 135492 (C4 (Web Text))

Max Activating Token Index: 404

Text #8

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0427. Min Act: -0.0001

Data Index: 924228 (C4 (Web Text))

Max Activating Token Index: 5

Text #9

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0379. Min Act: -0.0001

Data Index: 1199296 (C4 (Web Text))

Max Activating Token Index: 181

Text #10

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0381. Min Act: -0.0001

Data Index: 1579990 (Python Code)

Max Activating Token Index: 9

Text #11

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0365. Min Act: -0.0001

Data Index: 236457 (C4 (Web Text))

Max Activating Token Index: 544

Text #12

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0361. Min Act: -0.0001

Data Index: 1626841 (Python Code)

Max Activating Token Index: 740

Text #13

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0433. Min Act: -0.0001

Data Index: 1454136 (Python Code)

Max Activating Token Index: 413

Text #14

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0395. Min Act: -0.0001

Data Index: 1656021 (Python Code)

Max Activating Token Index: 437

Text #15

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0293. Min Act: -0.0001

Data Index: 19651 (C4 (Web Text))

Max Activating Token Index: 498

Text #16

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0342. Min Act: -0.0001

Data Index: 1428407 (Python Code)

Max Activating Token Index: 67

Text #17

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0344. Min Act: -0.0001

Data Index: 1463168 (Python Code)

Max Activating Token Index: 908

Text #18

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0382. Min Act: -0.0001

Data Index: 795554 (C4 (Web Text))

Max Activating Token Index: 492

Text #19

Max Range: 0.0477. Min Range: -0.0477

Max Act: 0.0384. Min Act: -0.0001

Data Index: 670058 (C4 (Web Text))

Max Activating Token Index: 325