Langchain embedding with llama.cpp does not work

[frank-onspecta]

Using langchain llama.cpp embedding generates wrong results. See the examples in langchain subfolder. Either chromadb or faiss have wrong results for document query. Here are relevant discussions on GGML llama.cpp repo

**abetlen/llama-cpp-python#503

Related:

ggerganov/llama.cpp#3643
ggerganov/llama.cpp#3625
ggerganov/llama.cpp#3243
ggerganov/llama.cpp#2180
ggerganov/llama.cpp#1013