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

Feature #2508

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

Fires on
0.25% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 48%
  • worksheet 25%
  • story 18%
  • agent 8%
  • python 1%

Makes the model say

didn wearid Oneil Heb tiny

Makes it not say

, end =)` more import together time

Strongest examples

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

  1. chat 2.75
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>
  2. chat 2.74
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|user_start|>
  3. chat 2.69
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>
  4. chat 2.67
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|user_start|>
  5. chat 2.65
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|user_start|>
  6. chat 2.65
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>
  7. chat 2.64
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>
  8. chat 2.64
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>
  9. chat 2.63
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|user_start|>
  10. chat 2.62
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>
  11. chat 2.62
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|user_start|>
  12. chat 2.62
    , a very small language model trained from scratch on a laptop by mini-lab.<|assistant_end|><|bos|>