Ollama Embeddings
If youβre opening this Notebook on colab, you will probably need to install LlamaIndex π¦.
%pip install llama-index-embeddings-ollamafrom 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},)You can generate embeddings using one of several methods:
get_text_embedding_batchget_text_embeddingget_query_embedding
As well as async versions:
aget_text_embedding_batchaget_text_embeddingaget_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%|ββββββββββ| 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]