Vicuna 13b model, could not parse ModelProto from /home/ken/git-repos/vicuna/config.json

[KennethThompson]

I am using the 13b checkpoint and vicuna model. Followed the steps and am stuck at this error message.

Not sure what I should do. I am downloading the 7b model to see if that one works.

In the readme, it does not indicate which exact model file should be loaded. When I try the tokenizer.model, it gives an error about invalid json. When I try the config.json, I get the error below.

Any assistance is much appreciated!

(minigptv) ken@aibox:~/git-repos/MiniGPT-4$ python demo.py --cfg-path eval_configs/minigpt4_eval.yaml --gpu-id 0

===================================BUG REPORT===================================
Welcome to bitsandbytes. For bug reports, please submit your error trace to: https://github.com/TimDettmers/bitsandbytes/issues

Initializing Chat
/home/ken/anaconda3/envs/minigptv/lib/python3.9/site-packages/transformers/tokenization_utils_base.py:1727: FutureWarning: Calling LlamaTokenizer.from_pretrained() with the path to a single file or url is deprecated and won't be possible anymore in v5. Use a model identifier or the path to a directory instead.
warnings.warn(
Traceback (most recent call last):
File "/home/ken/git-repos/MiniGPT-4/demo.py", line 65, in
model = model_cls.from_config(model_config).to('cuda:{}'.format(args.gpu_id))
File "/home/ken/git-repos/MiniGPT-4/minigpt4/models/minigpt4.py", line 169, in from_config
model = cls(
File "/home/ken/git-repos/MiniGPT-4/minigpt4/models/minigpt4.py", line 45, in init
super().init(
File "/home/ken/git-repos/MiniGPT-4/minigpt4/models/minigpt_base.py", line 41, in init
self.llama_model, self.llama_tokenizer = self.init_llm(
File "/home/ken/git-repos/MiniGPT-4/minigpt4/models/base_model.py", line 174, in init_llm
llama_tokenizer = LlamaTokenizer.from_pretrained(llama_model_path, use_fast=False)
File "/home/ken/anaconda3/envs/minigptv/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 1825, in from_pretrained
return cls._from_pretrained(
File "/home/ken/anaconda3/envs/minigptv/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 1988, in _from_pretrained
tokenizer = cls(*init_inputs, **init_kwargs)
File "/home/ken/anaconda3/envs/minigptv/lib/python3.9/site-packages/transformers/models/llama/tokenization_llama.py", line 96, in init
self.sp_model.Load(vocab_file)
File "/home/ken/anaconda3/envs/minigptv/lib/python3.9/site-packages/sentencepiece/init.py", line 961, in Load
return self.LoadFromFile(model_file)
File "/home/ken/anaconda3/envs/minigptv/lib/python3.9/site-packages/sentencepiece/init.py", line 316, in LoadFromFile
return _sentencepiece.SentencePieceProcessor_LoadFromFile(self, arg)
RuntimeError: Internal: could not parse ModelProto from /home/ken/git-repos/vicuna/config.json

[KennethThompson]

confirmed same issue is happening with 7b model. When I change the model path to the directory, I get an error about how the location is a directory. Really stuck here. What can I do to fix it?

[KennethThompson]

Nevermind. Fixed it. Issue was the checkpoint path in the config must be the full path to the file and not just the directory. Once fixed, it runs. Thanks!

[tuandattt]

Hi, I still cannot fix it, can you share your code?