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

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

Neuron 3065 in Layer 4

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.6168. Min Act: -0.0001

Data Index: 927121 (C4 (Web Text))

Max Activating Token Index: 120

Text #1

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.5434. Min Act: -0.0001

Data Index: 1144820 (C4 (Web Text))

Max Activating Token Index: 424

Text #2

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.5615. Min Act: -0.0001

Data Index: 911920 (C4 (Web Text))

Max Activating Token Index: 848

Text #3

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.5520. Min Act: -0.0001

Data Index: 870762 (C4 (Web Text))

Max Activating Token Index: 11

Text #4

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.5438. Min Act: -0.0001

Data Index: 495859 (C4 (Web Text))

Max Activating Token Index: 762

Text #5

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.5285. Min Act: -0.0001

Data Index: 955431 (C4 (Web Text))

Max Activating Token Index: 101

Text #6

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4546. Min Act: -0.0001

Data Index: 1445453 (Python Code)

Max Activating Token Index: 945

Text #7

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4964. Min Act: -0.0001

Data Index: 511938 (C4 (Web Text))

Max Activating Token Index: 155

Text #8

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4683. Min Act: -0.0001

Data Index: 1439213 (Python Code)

Max Activating Token Index: 937

Text #9

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4351. Min Act: -0.0001

Data Index: 1633227 (Python Code)

Max Activating Token Index: 954

Text #10

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4638. Min Act: -0.0001

Data Index: 1288375 (C4 (Web Text))

Max Activating Token Index: 5

Text #11

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4268. Min Act: -0.0001

Data Index: 1580288 (Python Code)

Max Activating Token Index: 888

Text #12

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4534. Min Act: -0.0001

Data Index: 655910 (C4 (Web Text))

Max Activating Token Index: 231

Text #13

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4725. Min Act: -0.0001

Data Index: 1611202 (Python Code)

Max Activating Token Index: 973

Text #14

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4378. Min Act: -0.0001

Data Index: 1354190 (C4 (Web Text))

Max Activating Token Index: 7

Text #15

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4363. Min Act: -0.0001

Data Index: 550774 (C4 (Web Text))

Max Activating Token Index: 4

Text #16

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4529. Min Act: -0.0001

Data Index: 1430083 (Python Code)

Max Activating Token Index: 921

Text #17

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4493. Min Act: -0.0001

Data Index: 1544284 (Python Code)

Max Activating Token Index: 983

Text #18

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4406. Min Act: -0.0001

Data Index: 1232961 (C4 (Web Text))

Max Activating Token Index: 881

Text #19

Max Range: 0.6168. Min Range: -0.6168

Max Act: 0.4687. Min Act: -0.0001

Data Index: 1475141 (Python Code)

Max Activating Token Index: 935