add pooling_type for embedding llamacpp

[machengyan]

• PR title: "community: add pooling_type to llamacpp embedding"

• PR message: the existing llamacpp embedding model doesn't work

  • Description: the default pooling_type is none, so the output of embed method is token-wise embedding which is not compatible with langchain core's FAISS.from_documents. so I add this too the model, and call like :
  • embeddings = LlamaCppEmbeddings(model_path="/usr/share/ollama/.ollama/models/blobs/sha256-8eeb52dfb3bb9aefdf9d1ef24b3bdbcfbe82238798c4b918278320b6fcef18fe",pooling_type=1)
  • so it works

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
langchain	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Oct 31, 2024 6:58pm

[ccurme]

Thanks for this. Could you please:

1. Verify that existing integration tests pass
2. Format and lint the code following the instructions here.

[efriis]

closing without response to review :(

if you're interested in addressing the comments and reopening, feel free!