lmsys

/

vicuna-13b-delta-v0

like 453

Follow

Large Model Systems Organization 513

I have the same problem, and I don't know how to solve it. someone help me?

not yet.

This is model v1.0. There is already a newer model, v1.1. And one of the things it fixes is the added_tokens, which might be the problem you're having here.

So unless there's a particular reason why you want 1.0 instead of 1.1, I would try: https://huggingface.co/lmsys/vicuna-13b-delta-v1.1

Or I have already merged deltas for v1.1 and uploaded them in HF format, so you could use those instead. Then you wouldn't need to merge the deltas yourselves: https://huggingface.co/TheBloke/vicuna-13B-1.1-HF

I have this problem。
OSError: Can't load tokenizer for 'pretrain_models/vicuna-7b-delta-v1.1/'. 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 'pretrain_models/vicuna-7b-delta-v1.1/' is the correct path to a directory containing all relevant files for a LlamaTokenizer
tokenizer.
I don't know how to solve it.

Again, you could just use https://huggingface.co/TheBloke/vicuna-13B-1.1-HF where it is already merged and you don't need to merge the deltas yourself.

I don't know why you're getting those errors. It worked fine when I ran it a few days ago. I expect something isn't set up right. Did you download Llama-13B-HF to merge the deltas on to?

It seems you were using a newer version of fschat with these old weights.
Please checkout the version compatibility here https://github.com/lm-sys/FastChat/blob/main/docs/weights_version.md
We suggest you use the newer v1.1 weights.
Close this for now. Feel free to reopen.

Thank you. I updated this FastChat and it worked.
Then I get the MiniGPT-4 checkpoint. Then when I run demo.py, it tells me:
RuntimeError: Error(s) in loading state_dict for MiniGPT4:
size mismatch for llama_proj.weight: copying a param with shape
torch.Size([5120, 768]) from checkpoint, the shape in current model is
torch.Size([4096, 768]).
size mismatch for llama_proj.bias: copying a param with shape
torch.Size([5120]) from checkpoint, the shape in current model is
torch.Size([4096]).
I wonder where the dimension error occurred.

This is model v1.0. There is already a newer model, v1.1. And one of the things it fixes is the added_tokens, which might be the problem you're having here.

So unless there's a particular reason why you want 1.0 instead of 1.1, I would try: https://huggingface.co/lmsys/vicuna-13b-delta-v1.1

Or I have already merged deltas for v1.1 and uploaded them in HF format, so you could use those instead. Then you wouldn't need to merge the deltas yourselves: https://huggingface.co/TheBloke/vicuna-13B-1.1-HF

thank you dude,i you're right we'd better use v1.1,once again,thank you

It seems you were using a newer version of fschat with these old weights.
Please checkout the version compatibility here https://github.com/lm-sys/FastChat/blob/main/docs/weights_version.md
We suggest you use the newer v1.1 weights.
Close this for now. Feel free to reopen.

yes, i had made it by your advice. and i would say vicuna is a masterpiece,thanks for sharing.