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

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

Neuron 1288 in Layer 3

Load this data into an Interactive Neuroscope

See Documentation here

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

Text #0

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.6771. Min Act: -0.1700

Data Index: 905840 (C4 (Web Text))

Max Activating Token Index: 234

Text #1

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.7083. Min Act: -0.1700

Data Index: 8477 (C4 (Web Text))

Max Activating Token Index: 223

Text #2

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.6146. Min Act: -0.1700

Data Index: 838346 (C4 (Web Text))

Max Activating Token Index: 345

Text #3

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.6146. Min Act: -0.1699

Data Index: 1441992 (Python Code)

Max Activating Token Index: 202

Text #4

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5833. Min Act: -0.1700

Data Index: 607596 (C4 (Web Text))

Max Activating Token Index: 899

Text #5

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.6458. Min Act: -0.1700

Data Index: 1389420 (Python Code)

Max Activating Token Index: 935

Text #6

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.6771. Min Act: -0.1700

Data Index: 1423657 (Python Code)

Max Activating Token Index: 261

Text #7

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.6458. Min Act: -0.1700

Data Index: 518877 (C4 (Web Text))

Max Activating Token Index: 232

Text #8

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.4270. Min Act: -0.1700

Data Index: 1043361 (C4 (Web Text))

Max Activating Token Index: 964

Text #9

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5208. Min Act: -0.1700

Data Index: 1463116 (Python Code)

Max Activating Token Index: 410

Text #10

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5208. Min Act: -0.1700

Data Index: 1428057 (Python Code)

Max Activating Token Index: 291

Text #11

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5521. Min Act: -0.1700

Data Index: 532815 (C4 (Web Text))

Max Activating Token Index: 98

Text #12

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.4896. Min Act: -0.1700

Data Index: 1537266 (Python Code)

Max Activating Token Index: 343

Text #13

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5521. Min Act: -0.1700

Data Index: 1490633 (Python Code)

Max Activating Token Index: 816

Text #14

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.4583. Min Act: -0.1700

Data Index: 702393 (C4 (Web Text))

Max Activating Token Index: 155

Text #15

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5208. Min Act: -0.1700

Data Index: 1140436 (C4 (Web Text))

Max Activating Token Index: 92

Text #16

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.4896. Min Act: -0.1700

Data Index: 61308 (C4 (Web Text))

Max Activating Token Index: 924

Text #17

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.4896. Min Act: -0.1700

Data Index: 1458905 (Python Code)

Max Activating Token Index: 403

Text #18

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.4896. Min Act: -0.1700

Data Index: 678801 (C4 (Web Text))

Max Activating Token Index: 126

Text #19

Max Range: 4.7083. Min Range: -4.7083

Max Act: 4.5208. Min Act: -0.1700

Data Index: 487363 (C4 (Web Text))

Max Activating Token Index: 15