Hugging Face's logo

TheBloke
/
minotaur-mpt-7B-GGML

Transformers
mpt
text-generation-inference
6 papers
Model card Files Community
1

mpt_model_load: tensor 'transformer.wte.weight' has wrong size in model file

#1
by vasilee - opened
Discussion
vasilee

code that leads to that error

from ctransformers import AutoModelForCausalLM

llm = AutoModelForCausalLM.from_pretrained("TheBloke/minotaur-mpt-7B-GGML",
                                           model_file="minotaur-mpt-7b.ggmlv3.q4_1.bin",
                                           model_type="mpt")
print(llm('AI is going to'))
TheBloke
 Owner

Please double check that the file is properly downloaded. This sounds like your download may have aborted part way through.

vasilee

I downloaded it locally and created a space that is downloading on docker the file and it is the same error for q4_0 and q4_1 files. Didn't check the others.
https://huggingface.co/spaces/vasilee/wizardcoder-ggml?logs=container

TheBloke
 Owner

OK in that case I am not sure

@marella do you have any idea what might be wrong?

marella

I'm getting the same error with the example in ggml repo.
Can you try re-converting and open an issue in ggml repo if it still doesn't work.

Sign up or log in to comment