Model: SoLU Model: 10 Layers, 5120 Neurons per Layer

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

Neuron 5102 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-10l')

Text #0

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4899. Min Act: -0.0001

Data Index: 113110 (C4 (Web Text))

Max Activating Token Index: 54

Text #1

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4916. Min Act: -0.0001

Data Index: 1457205 (Python Code)

Max Activating Token Index: 630

Text #2

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4967. Min Act: -0.0001

Data Index: 154565 (C4 (Web Text))

Max Activating Token Index: 697

Text #3

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4353. Min Act: -0.0001

Data Index: 1427920 (Python Code)

Max Activating Token Index: 85

Text #4

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4414. Min Act: -0.0001

Data Index: 1653678 (Python Code)

Max Activating Token Index: 312

Text #5

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4754. Min Act: -0.0001

Data Index: 95634 (C4 (Web Text))

Max Activating Token Index: 341

Text #6

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4028. Min Act: -0.0001

Data Index: 693696 (C4 (Web Text))

Max Activating Token Index: 347

Text #7

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4222. Min Act: -0.0001

Data Index: 739126 (C4 (Web Text))

Max Activating Token Index: 152

Text #8

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4584. Min Act: -0.0001

Data Index: 711610 (C4 (Web Text))

Max Activating Token Index: 340

Text #9

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4406. Min Act: -0.0001

Data Index: 1076530 (C4 (Web Text))

Max Activating Token Index: 24

Text #10

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4290. Min Act: -0.0001

Data Index: 1591514 (Python Code)

Max Activating Token Index: 815

Text #11

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4151. Min Act: -0.0001

Data Index: 987243 (C4 (Web Text))

Max Activating Token Index: 384

Text #12

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4265. Min Act: -0.0001

Data Index: 173899 (C4 (Web Text))

Max Activating Token Index: 868

Text #13

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4309. Min Act: -0.0001

Data Index: 1568621 (Python Code)

Max Activating Token Index: 327

Text #14

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.3991. Min Act: -0.0001

Data Index: 1189638 (C4 (Web Text))

Max Activating Token Index: 128

Text #15

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4109. Min Act: -0.0001

Data Index: 1390686 (Python Code)

Max Activating Token Index: 439

Text #16

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4093. Min Act: -0.0001

Data Index: 1087688 (C4 (Web Text))

Max Activating Token Index: 72

Text #17

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.3407. Min Act: -0.0001

Data Index: 119023 (C4 (Web Text))

Max Activating Token Index: 165

Text #18

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.4202. Min Act: -0.0001

Data Index: 1599629 (Python Code)

Max Activating Token Index: 726

Text #19

Max Range: 0.4967. Min Range: -0.4967

Max Act: 0.3882. Min Act: -0.0001

Data Index: 28995 (C4 (Web Text))

Max Activating Token Index: 849