ONNX-Demo / pyserini /resources /index-metadata /faiss-flat.wikipedia.dkrr-dpr-tqa-retriever.20220217.25ed1f.cc91b2.README.md
ArthurChen189's picture
upload pyserini
62977bb
|
raw
history blame
1.11 kB
# wikipedia-dpr-dkrr-tqa
Faiss FlatIP index of Wikipedia DPR encoded by the retriever model from [Distilling Knowledge from Reader to Retriever for Question Answering](https://arxiv.org/abs/2012.04584) trained on TriviaQA.
This index was generated on 2022/02/17 on `orca` at commits:
+ Pyserini commit [`cc91b2`](https://github.com/castorini/pyserini/commit/cc91b22f549702068cea1283f91b31d28d127b2f) (2022/02/17)
+ [FiD](https://github.com/facebookresearch/FiD) commit [`25ed1f`](https://github.com/facebookresearch/FiD/commit/25ed1ff0fe0288b80fb5e9e5de8d6346b94b8d48) (2022/02/17)
with the following command to generate the embeddings (from FiD repo):
```bash
python generate_passage_embeddings.py \
--model_path tqa_retriever \
--passages passages.tsv \
--output_path wikipedia_embeddings_tqa \
--shard_id 0 \
--num_shards 1 \
--per_gpu_batch_size 500
```
and the following command to convert the embeddings to faiss IndexFlatIP form:
```bash
python convert_dkrr_embeddings_to_faiss.py \
--embeddings wikipedia_embeddings_tqa \
--output faiss-flat.wikipedia.dkrr-dpr-tqa-retriever
```