What's the reason for deleting the llama_convert_and_add_eot_token script?

[l3utterfly]

I feel it's a very valuable script to help people get started training their own models using the openchat method.

[imoneoi]

Because of some compatibility issues with Llama 2. I'm working on a new version. Besides, all converted models are uploaded here:

https://huggingface.co/imone/LLaMA2_13B_with_EOT_token
https://huggingface.co/imone/LLaMA_13B_with_EOT_token

[l3utterfly]

I would like to convert a llama2_7b model and use the training method here. Do you mind sharing your new converter code?

[imoneoi]

Use the following script. It only supports 7B & 13B models

convert_llama_weights_to_hf_add_tokens

[l3utterfly]

Thank you, I got it working.

https://huggingface.co/l3utterfly/llama-2-7b_with-EOT-token (if you are interested)