Problem embedding with new-line
#3
by
planti
- opened
Hi vonjack,
This model does not seem to support the new-line character (\n).
I tried with LLM Studio, and if I enter text with the character \n, I get this error:
Unknown exception caused embedding to stop: unordered_map::at: key not found.
Sentences without newline characters are embedded successfully.
planti
I encountered the same issue using bge-m3-f16.gguf to embed a text containing \n.
I am running the model using Ollama.
This error message is:
terminate called after throwing an instance of 'std::out_of_range'
what(): _Map_base::at
which is basically the same error.
This conversion is using the wrong vocabulary, see https://github.com/ggerganov/llama.cpp/blob/017cc5f446863316d05522a87f25ec48713a9492/convert_hf_to_gguf.py#L3065.
@planti @ThetaLin , please give a shot with gpustack/bge-m3-GGUF .