Model: GELU Model: 4 Layers, 2048 Neurons per Layer
Dataset: 80% C4 (Web Text) and 20% Python Code
Neuron 2006 in Layer 3
Transformer Lens Loading: HookedTransformer.from_pretrained('gelu-4l')
Text #0
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.7248. Min Act: -0.1700
Data Index: 682503 (C4 (Web Text))
Max Activating Token Index: 224
Click toggle to see full text
Truncated
Full Text #0
Text #1
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.4123. Min Act: -0.1700
Data Index: 220784 (C4 (Web Text))
Max Activating Token Index: 839
Click toggle to see full text
Truncated
Full Text #1
Text #2
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.4435. Min Act: -0.1700
Data Index: 887843 (C4 (Web Text))
Max Activating Token Index: 919
Click toggle to see full text
Truncated
Full Text #2
Text #3
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.3810. Min Act: -0.1700
Data Index: 915913 (C4 (Web Text))
Max Activating Token Index: 107
Click toggle to see full text
Truncated
Full Text #3
Text #4
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.3654. Min Act: -0.1700
Data Index: 1075352 (C4 (Web Text))
Max Activating Token Index: 392
Click toggle to see full text
Truncated
Full Text #4
Text #5
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.3029. Min Act: -0.1700
Data Index: 796529 (C4 (Web Text))
Max Activating Token Index: 907
Click toggle to see full text
Truncated
Full Text #5
Text #6
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.2717. Min Act: -0.1700
Data Index: 99829 (C4 (Web Text))
Max Activating Token Index: 928
Click toggle to see full text
Truncated
Full Text #6
Text #7
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.3185. Min Act: -0.1700
Data Index: 245831 (C4 (Web Text))
Max Activating Token Index: 1005
Click toggle to see full text
Truncated
Full Text #7
Text #8
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.2717. Min Act: -0.1700
Data Index: 681120 (C4 (Web Text))
Max Activating Token Index: 320
Click toggle to see full text
Truncated
Full Text #8
Text #9
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.2717. Min Act: -0.1700
Data Index: 515522 (C4 (Web Text))
Max Activating Token Index: 441
Click toggle to see full text
Truncated
Full Text #9
Text #10
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.3029. Min Act: -0.1700
Data Index: 805999 (C4 (Web Text))
Max Activating Token Index: 221
Click toggle to see full text
Truncated
Full Text #10
Text #11
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.2717. Min Act: -0.1700
Data Index: 289848 (C4 (Web Text))
Max Activating Token Index: 592
Click toggle to see full text
Truncated
Full Text #11
Text #12
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.2248. Min Act: -0.1700
Data Index: 641895 (C4 (Web Text))
Max Activating Token Index: 865
Click toggle to see full text
Truncated
Full Text #12
Text #13
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.1935. Min Act: -0.1700
Data Index: 261317 (C4 (Web Text))
Max Activating Token Index: 643
Click toggle to see full text
Truncated
Full Text #13
Text #14
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.1935. Min Act: -0.1700
Data Index: 560545 (C4 (Web Text))
Max Activating Token Index: 947
Click toggle to see full text
Truncated
Full Text #14
Text #15
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.2560. Min Act: -0.1700
Data Index: 315659 (C4 (Web Text))
Max Activating Token Index: 659
Click toggle to see full text
Truncated
Full Text #15
Text #16
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.1779. Min Act: -0.1700
Data Index: 935622 (C4 (Web Text))
Max Activating Token Index: 823
Click toggle to see full text
Truncated
Full Text #16
Text #17
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.1466. Min Act: -0.1700
Data Index: 1463314 (Python Code)
Max Activating Token Index: 806
Click toggle to see full text
Truncated
Full Text #17
Text #18
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.1935. Min Act: -0.1700
Data Index: 757668 (C4 (Web Text))
Max Activating Token Index: 699
Click toggle to see full text
Truncated
Full Text #18
Text #19
Max Range: 4.7248. Min Range: -4.7248
Max Act: 4.1779. Min Act: -0.1700
Data Index: 1464932 (Python Code)
Max Activating Token Index: 661
Click toggle to see full text
Truncated
Full Text #19