technical question about the model

#15
by arwwarr - opened

what steps would be necessary if I wanted to translate this model into my language?

by the way, you're doing a great job :)

Cognitive Computations org

Basically, rebuild WizardLM just like I did.

I have a guide on https://erichartford.com/uncensored-models

Step one is to download the dataset. You have a choice to download the censored dataset or the original dataset.

https://huggingface.co/datasets/ehartford/WizardLM_alpaca_evol_instruct_70k_unfiltered

https://huggingface.co/datasets/victor123/evol_instruct_70k

Next you need to translate the dataset to your language.

Finally you need to fine-tune llama with your translate dataset.

The official procedure is here:
https://github.com/nlpxucan/WizardLM

I followed it exactly.

thank you very much for your reply :)

https://huggingface.co/datasets/ehartford/WizardLM_alpaca_evol_instruct_70k_unfiltered

is it recommended to run wizardlm_clean.py here?

Cognitive Computations org

You don't have to run the script if you are using my filtered dataset,

That's the script to get to my dataset from the original dataset

But you can change it if you like and generate your own custom dataset that way

Sign up or log in to comment