NickyNicky

/

gemma-2b-it_oasst2_chatML_Cluster2_aya_multilingual_10k_prompts_ranked_all_json_V1

like 3

Thank you very much and thanks to those who created the dataset.

My objectives would basically be to be able to have a prompt rating to be able to filter between the good ratings and the bad ones since there are many datasets and I can't see them all (no me da la vida), in addition to increasing the dataset and being able to train with more data.

If you see things to improve, here I am.

thank you so much.

I'm really excited to see this! I've also tried using training models for this, but so far, none have been very good! It would definitely be very powerful to be able to leverage a model to help with this task.

With the dataset you could reproduce the same model.

dataset:
https://huggingface.co/datasets/NickyNicky/10k_prompts_ranked_all_chatml_json_gemma

here I made a small 808 dataset and this is the code to run it (colab)

https://huggingface.co/datasets/NickyNicky/prompts_ranked_808