I am unable to directly load this model?

#2
by hrituraj - opened

Thanks for this work guys - I have been trying to download and setup this model locally but have not been able to do so.

Directly using Yukang/Llama-2-70b-chat-longlora-32k-sft threw the following error

OSError: Can't load tokenizer for 'Yukang/Llama-2-70b-chat-longlora-32k-sft'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure 'Yukang/Llama-2-70b-chat-longlora-32k-sft' is the correct path to a directory containing all relevant files for a LlamaTokenizer tokenizer.

So, I pointed the tokenizer towards Yukang/Llama-13b-chat-longlora-32k-sft assuming same tokenizer must have been used. I, then started receiving the following error though -

OSError: Can't load the model for '/dataset/pretrained-models/Llama-2-70b-chat-hf'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure '/dataset/pretrained-models/Llama-2-70b-chat-hf' is the correct path to a directory containing a file named pytorch_model.bin, tf_model.h5, model.ckpt or flax_model.msgpack

Coul you please look into it?

Hi,

Because this is a lora weight, you need to merge it into the base model first to get the pytorch_model.bin and tokenizer files. Please refer to the step in the Github README.
https://github.com/dvlab-research/LongLoRA#merge-lora-weight

Regards,
Yukang Chen

Sign up or log in to comment