Power User
  • Communities
  • Create Post
  • Create Community
  • heart
    Support Lemmy
  • search
    Search
  • Login
  • Sign Up
Delicious-Farmer-234@alien.topB to LocalLLaMAEnglish · 2 years ago

The overthinker

message-square
message-square
6
link
fedilink
1
message-square

The overthinker

Delicious-Farmer-234@alien.topB to LocalLLaMAEnglish · 2 years ago
message-square
6
link
fedilink

I overfitted the Phi 1.5 model on a riddle dataset found here:

https://huggingface.co/datasets/Ermarrero/riddles_v1

I just wanted to see how it behaves and I gotta say the output is interesting since it thinks everything is a riddle and tries to break it down logically.

It’s weird but it is kind of refreshing to see a model overthink it and dig too deep into things. I dunno, what do you guys think?

if you want to play around with the model I can upload it to hugginface.

https://preview.redd.it/poxet1xqdj3c1.png?width=1318&format=png&auto=webp&s=433ec4a7ce51e578182d35a97525a7b0f95f4745

https://preview.redd.it/au4617xqdj3c1.png?width=1279&format=png&auto=webp&s=76d00f416fdcf8785f84500c50a0115976dd3b97

https://preview.redd.it/ktyo25xqdj3c1.png?width=1305&format=png&auto=webp&s=2518c63ebf43b1b95955ef7846e464a76c175e7d

https://preview.redd.it/riygz6xqdj3c1.png?width=2130&format=png&auto=webp&s=b895fde3b7aff260268957dcdb1082f18ce55959

https://preview.redd.it/wjpha9xqdj3c1.png?width=1300&format=png&auto=webp&s=0523c754b17d05d09e194d10afe93236a5b3830f

alert-triangle
You must log in or register to comment.
  • Dry_Long3157@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    2 years ago

    May I know the LoRA parameters, if you used q/LoRA?

  • phree_radical@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    2 years ago

    It’s awesome! Please upload ♥

    RemindMe! 3 days

  • richinseattle@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    2 years ago

    This is hilarious on its own and may serve as a good phase of tuning a small model before feeding it more domain specific data, love it.

  • FPham@alien.topB
    link
    fedilink
    English
    arrow-up
    0
    ·
    2 years ago

    The ridle json + sydney actually makes the model far more lucid than normally. I applied it on Mythomax and the answers are really good.

    ​

    https://preview.redd.it/gmisu6ac0l3c1.png?width=925&format=png&auto=webp&s=4d6fcf5bd5b89a6e6bd8ec86def35b19514473b1

    • Feztopia@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      2 years ago

      To bad that the last sentence is incorrect. For example, Singapore is the capital of Singapore.

    • liquiddandruff@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      2 years ago

      dang, that’s actually very impressive

LocalLLaMA

localllama

Subscribe from Remote Instance

Create a post
You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: !localllama@poweruser.forum

Community to discuss about Llama, the family of large language models created by Meta AI.

Visibility: Public
globe

This community can be federated to other instances and be posted/commented in by their users.

  • 4 users / day
  • 4 users / week
  • 4 users / month
  • 4 users / 6 months
  • 1 local subscriber
  • 11 subscribers
  • 1.02K Posts
  • 5.82K Comments
  • Modlog
  • mods:
  • communick
  • BE: 0.19.11
  • Modlog
  • Instances
  • Docs
  • Code
  • join-lemmy.org