feat: support embedding mode in WASI-NN ggml plugin
hydai opened this issue · comments
Summary
llama.cpp provides APIs for generating embedding vectors from a given model and prompts.
We would like to enable this feature so users using our WASI-NN ggml plugin can retrieve the embedding vector from the metadata.
Details
Check https://github.com/ggerganov/llama.cpp/blob/master/examples/embedding/embedding.cpp
Provide an option in metadata to enable the embedding mode and return the embedding vector inside the metadata field.
Appendix
No response