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

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

Neuron 452 in Layer 2

Load this data into an Interactive Neuroscope

See Documentation here

Transformer Lens Loading: HookedTransformer.from_pretrained('solu-4l')

Text #0

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.6067. Min Act: -0.0005

Data Index: 1459615 (Python Code)

Max Activating Token Index: 876

Text #1

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5355. Min Act: -0.0005

Data Index: 199649 (C4 (Web Text))

Max Activating Token Index: 951

Text #2

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5326. Min Act: -0.0005

Data Index: 1600771 (Python Code)

Max Activating Token Index: 898

Text #3

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5278. Min Act: -0.0005

Data Index: 270938 (C4 (Web Text))

Max Activating Token Index: 258

Text #4

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5103. Min Act: -0.0005

Data Index: 208693 (C4 (Web Text))

Max Activating Token Index: 314

Text #5

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5099. Min Act: -0.0005

Data Index: 1544706 (Python Code)

Max Activating Token Index: 228

Text #6

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5195. Min Act: -0.0005

Data Index: 520094 (C4 (Web Text))

Max Activating Token Index: 931

Text #7

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4906. Min Act: -0.0005

Data Index: 1161749 (C4 (Web Text))

Max Activating Token Index: 448

Text #8

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5383. Min Act: -0.0005

Data Index: 604842 (C4 (Web Text))

Max Activating Token Index: 556

Text #9

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4992. Min Act: -0.0005

Data Index: 501001 (C4 (Web Text))

Max Activating Token Index: 1004

Text #10

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4956. Min Act: -0.0005

Data Index: 292136 (C4 (Web Text))

Max Activating Token Index: 388

Text #11

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4394. Min Act: -0.0005

Data Index: 827675 (C4 (Web Text))

Max Activating Token Index: 488

Text #12

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4681. Min Act: -0.0005

Data Index: 968830 (C4 (Web Text))

Max Activating Token Index: 353

Text #13

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4810. Min Act: -0.0005

Data Index: 1077776 (C4 (Web Text))

Max Activating Token Index: 273

Text #14

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4649. Min Act: -0.0005

Data Index: 261295 (C4 (Web Text))

Max Activating Token Index: 104

Text #15

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4898. Min Act: -0.0005

Data Index: 931720 (C4 (Web Text))

Max Activating Token Index: 290

Text #16

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4920. Min Act: -0.0005

Data Index: 931804 (C4 (Web Text))

Max Activating Token Index: 503

Text #17

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4535. Min Act: -0.0005

Data Index: 1618926 (Python Code)

Max Activating Token Index: 373

Text #18

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.4394. Min Act: -0.0005

Data Index: 1564301 (Python Code)

Max Activating Token Index: 594

Text #19

Max Range: 0.6067. Min Range: -0.6067

Max Act: 0.5001. Min Act: -0.0005

Data Index: 597601 (C4 (Web Text))

Max Activating Token Index: 659