llama-index packs infer retrieve rerank integration
Project description
Infer-Retrieve-Rerank LlamaPack
This is our implementation of the paper "In-Context Learning for Extreme Multi-Label Classification by Oosterlinck et al.
The paper proposes "infer-retrieve-rerank", a simple paradigm using frozen LLM/retriever models that can do "extreme"-label classification (the label space is huge).
- Given a user query, use an LLM to predict an initial set of labels.
- For each prediction, retrieve the actual label from the corpus.
- Given the final set of labels, rerank them using an LLM.
All of these can be implemented as LlamaIndex abstractions.
A full notebook guide can be found here.
CLI Usage
You can download llamapacks directly using llamaindex-cli
, which comes installed with the llama-index
python package:
llamaindex-cli download-llamapack InferRetrieveRerankPack --download-dir ./infer_retrieve_rerank_pack
You can then inspect the files at ./infer_retrieve_rerank_pack
and use them as a template for your own project!
Code Usage
You can download the pack to a ./infer_retrieve_rerank_pack
directory:
from llama_index.core.llama_pack import download_llama_pack
# download and install dependencies
InferRetrieveRerankPack = download_llama_pack(
"InferRetrieveRerankPack", "./infer_retrieve_rerank_pack"
)
From here, you can use the pack, or inspect and modify the pack in ./infer_retrieve_rerank_pack
.
Then, you can set up the pack like so:
# create the pack
pack = InferRetrieveRerankPack(
labels, # list of all label strings
llm=llm,
pred_context="<pred_context>",
reranker_top_n=3,
verbose=True,
)
The run()
function runs predictions.
pred_reactions = pack.run(inputs=[s["text"] for s in samples])
You can also use modules individually.
# call the llm.complete()
llm = pack.llm
label_retriever = pack.label_retriever
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for llama_index_packs_infer_retrieve_rerank-0.0.1.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | dfddf7523e94f5e4cb8afdfd50d06bd6134ed115501d352eb522107a410bc518 |
|
MD5 | 7a467fca8971d8de93c06d457e85cc0b |
|
BLAKE2b-256 | 9de81dd80650db2197f5036e09fd3f860bd43d25386ac0a71b595d3b98733afd |
Hashes for llama_index_packs_infer_retrieve_rerank-0.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 42711d2ccb203686d352d26460c3bce2920921d81db0ee0f5b4a5e4954ad5336 |
|
MD5 | 7e4268d42b855706814df02eb8de4c52 |
|
BLAKE2b-256 | d536727ac4ee436739ce61e2fb6ecf4a73ee878eaeda249fc33d102690faa7b7 |