Ollama 임베딩

Colab에서 이 노트북을 여는 경우라면 LlamaIndex 🦙 설치가 필요할 거예요.

%pip install llama-index-embeddings-ollama

from llama_index.embeddings.ollama import OllamaEmbedding


ollama_embedding = OllamaEmbedding(
    model_name="embeddinggemma",
    base_url="http://localhost:11434",
    # Can optionally pass additional kwargs to ollama
    # ollama_additional_kwargs={"mirostat": 0},
)

다음과 같은 여러 메서드 중 하나로 임베딩을 생성할 수 있어요.

get_text_embedding_batch
get_text_embedding
get_query_embedding

그리고 비동기 버전도 있어요.

aget_text_embedding_batch
aget_text_embedding
aget_query_embedding

embeddings = ollama_embedding.get_text_embedding_batch(
    ["This is a passage!", "This is another passage"], show_progress=True
)
print(f"Got vectors of length {len(embeddings[0])}")
print(embeddings[0][:10])

Generating embeddings: 100%|██████████| 2/2 [00:00<00:00,  3.66it/s]


Got vectors of length 768
[-0.19284482, -0.0048683924, 0.011490762, -0.035292886, 0.0018508184, 0.013227936, -0.045588765, 0.027076142, 0.03387062, -0.030585105]

embedding = ollama_embedding.get_text_embedding(
    "This is a piece of text!",
)
print(f"Got vectors of length {len(embedding)}")
print(embedding[:10])

Got vectors of length 768
[-0.18305846, -0.009758809, 0.022796445, -0.038445882, -0.00894579, 0.023117013, -0.05166001, 0.037556227, 0.03699912, -0.017603736]

embedding = ollama_embedding.get_query_embedding(
    "This is a query!",
)
print(f"Got vectors of length {len(embedding)}")
print(embedding[:10])

Got vectors of length 768
[-0.19484262, -0.014648143, 0.02743501, -0.015000358, 0.0027351314, 0.019096522, -0.071097225, 0.033618074, 0.05173764, -0.024861954]

출처: 문서

더 알아보기 (Learn more)

Ollama