Skip to content
mini-lab
Log in Sign up
Features of prelude-2

Feature #627

Fires on “.” (not labeled yet: read the examples)

Fires on
1.07% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 70%
  • python 20%
  • chat 5%
  • story 4%
  • worksheet 1%

Makes the model say

0.?One name7ptFix

Makes it not say

x off fell<|arg|> apples have birds help

Strongest examples

On text the dictionary never trained on. The underlined token is where it fires most.

  1. python 3.58
    example, `mean(1, 12)` returns `6.5`.⏎⏎
  2. python 3.55
    6])` returns `6.75`.⏎
  3. python 3.54
    _list([3, 8, 7])` returns `6.0`.⏎⏎
  4. python 3.53
    _list([8, 9, 8])` returns `8.3333
  5. python 3.5
    list(SCORES))⏎⏎$ python3 app.py⏎6.0⏎<|bos|>Once
  6. python 3.4
    1, 7, 2, 8])` returns `5.2`.⏎<|bos|>
  7. python 3.39
    average_list([9, 1, 5])⏎5.0⏎⏎`
  8. python 3.35
    8, 4, 6, 8])` returns `6.2`.⏎<|bos|>
  9. python 3.34
    For example, `mean(5, 1)` returns `3.0`.⏎<|bos|>
  10. python 3.34
    len(nums)⏎⏎$ python3 run.py⏎6.0⏎1⏎
  11. python 3.32
    mean_list([9, 9, 4])⏎7.3333
  12. python 3.3
    _list([2, 6, 0])` returns `2.6666