All Tools
W
RAGFreeOpen Source
WORDLLAMA
Cheap embeddings from LLM tokenizer weights
MIT
ABOUT
Many teams need decent text similarity but cannot afford a GPU embedding model on every request. WordLlama reuses the token embedding table of an LLM as a small, CPU-friendly embedding model for clustering, reranking, and lightweight retrieval.
INSTALL
pip install wordllamaINTEGRATION GUIDE
1. Embed queries and docs on CPU for a small RAG index
2. Cluster support tickets or papers without a sentence-transformer GPU
3. Prototype similarity search before paying for a hosted embedding API
4. Deduplicate or fuzzy-match strings using Llama tokenizer embeddings
TAGS
pythonembeddingsragllamatokenizeropen-source