Model: SoLU Model: 6 Layers, 3072 Neurons per Layer

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

Neuron 1067 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-6l')

Text #0

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.9238. Min Act: -0.0001

Data Index: 127568 (C4 (Web Text))

Max Activating Token Index: 145

Text #1

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.8247. Min Act: -0.0001

Data Index: 947985 (C4 (Web Text))

Max Activating Token Index: 872

Text #2

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.7986. Min Act: -0.0001

Data Index: 1473818 (Python Code)

Max Activating Token Index: 503

Text #3

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.7363. Min Act: -0.0001

Data Index: 81643 (C4 (Web Text))

Max Activating Token Index: 521

Text #4

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.7074. Min Act: -0.0001

Data Index: 1491750 (Python Code)

Max Activating Token Index: 1021

Text #5

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.7186. Min Act: -0.0001

Data Index: 1341841 (C4 (Web Text))

Max Activating Token Index: 839

Text #6

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.7018. Min Act: -0.0001

Data Index: 1478593 (Python Code)

Max Activating Token Index: 835

Text #7

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6848. Min Act: -0.0001

Data Index: 550281 (C4 (Web Text))

Max Activating Token Index: 560

Text #8

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.7195. Min Act: -0.0001

Data Index: 1136377 (C4 (Web Text))

Max Activating Token Index: 959

Text #9

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6684. Min Act: -0.0001

Data Index: 915592 (C4 (Web Text))

Max Activating Token Index: 894

Text #10

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6695. Min Act: -0.0001

Data Index: 1271927 (C4 (Web Text))

Max Activating Token Index: 888

Text #11

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6534. Min Act: -0.0001

Data Index: 982461 (C4 (Web Text))

Max Activating Token Index: 297

Text #12

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6693. Min Act: -0.0001

Data Index: 1510830 (Python Code)

Max Activating Token Index: 828

Text #13

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6612. Min Act: -0.0001

Data Index: 1631010 (Python Code)

Max Activating Token Index: 115

Text #14

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6463. Min Act: -0.0001

Data Index: 1614001 (Python Code)

Max Activating Token Index: 953

Text #15

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6470. Min Act: -0.0001

Data Index: 114956 (C4 (Web Text))

Max Activating Token Index: 553

Text #16

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6279. Min Act: -0.0001

Data Index: 1066043 (C4 (Web Text))

Max Activating Token Index: 652

Text #17

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6230. Min Act: -0.0001

Data Index: 885095 (C4 (Web Text))

Max Activating Token Index: 477

Text #18

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6374. Min Act: -0.0001

Data Index: 97714 (C4 (Web Text))

Max Activating Token Index: 603

Text #19

Max Range: 0.9238. Min Range: -0.9238

Max Act: 0.6120. Min Act: -0.0001

Data Index: 204708 (C4 (Web Text))

Max Activating Token Index: 806