jhyuklee/paragraphranker

EMNLP'2018: Ranking Paragraphs for Improving Answer Recall in Open-Domain Question Answering

paragraphranker's Issues

when evaluation, it shows no module named 'drqa'

Below is the error output, even though I have developed the according to the DrQA. I think the reason is pytorch version? Right now I am using pytorch=0.4.1, while DrQA is not compatible with this version.
Hope someone can help me thanks.
By the way, can i do the interactive retriever with the help of Paragraph Ranker?

12/13/2018 05:26:03 PM: [ COMMAND: entityqa/pipeline/predict.py --query-type SQuAD --ranker-type default --reader-type default ]
12/13/2018 05:26:03 PM: [ ---------------------------------------------------------------------------------------------------- ]
12/13/2018 05:26:03 PM: [ CONFIG:
{
    "candidate_file": null,
    "cuda": true,
    "data_workers": 5,
    "db_path": "/home/jiangkunhaha/ParagraphRanker/data/wikipedia/docs.db",
    "embed_dir": "/home/jiangkunhaha/datasets",
    "embedding_dim": 300,
    "embedding_file": "/home/jiangkunhaha/datasets/glove/glove.840B.300d.txt",
    "fasttext": false,
    "gpu": -1,
    "log_file": "/home/jiangkunhaha/ParagraphRanker/entityqa/pipeline/results/20181213-d59542b8.txt",
    "match": "string",
    "model_dir": "/home/jiangkunhaha/ParagraphRanker/entityqa/pipeline/results",
    "model_name": "20181213-d59542b8",
    "module_batch_size": 64,
    "module_max_loaders": 5,
    "n_docs": 20,
    "n_pars": 200,
    "no_cuda": false,
    "parallel": false,
    "pred_file": "/home/jiangkunhaha/ParagraphRanker/entityqa/pipeline/results/pred_20181213-d59542b8.json",
    "predict_batch_size": 100,
    "query_data": "/home/jiangkunhaha/ParagraphRanker/data/datasets/SQuAD-v1.1-dev.txt",
    "query_split": "dev",
    "query_type": "SQuAD",
    "random_seed": 1013,
    "ranker_file": "/home/jiangkunhaha/ParagraphRanker/entityqa/pipeline/results/20181213-d59542b8.ranker",
    "ranker_path": "/home/jiangkunhaha/ParagraphRanker/entityqa/ranker/results/20181212-a0db0070.mdl",
    "ranker_type": "default",
    "reader_path": "/home/jiangkunhaha/ParagraphRanker/entityqa/reader/results/single.mdl",
    "reader_type": "default",
    "retriever_path": "/home/jiangkunhaha/ParagraphRanker/data/wikipedia/docs-tfidf-ngram=2-hash=16777216-tokenizer=simple.npz",
    "tokenizer": "regexp",
    "train": false,
    "train_epoch": 30
} ]
12/13/2018 05:26:03 PM: [ Reading query data /home/jiangkunhaha/ParagraphRanker/data/datasets/SQuAD-v1.1-dev.txt ]
12/13/2018 05:26:03 PM: [ Initializing pipeline... ]
12/13/2018 05:26:03 PM: [ Loading retriever /home/jiangkunhaha/ParagraphRanker/data/wikipedia/docs-tfidf-ngram=2-hash=16777216-tokenizer=simple.npz ]
12/13/2018 05:26:35 PM: [ Loading ranker /home/jiangkunhaha/ParagraphRanker/entityqa/ranker/results/20181212-a0db0070.mdl ]
12/13/2018 05:26:37 PM: [ Loading reader /home/jiangkunhaha/ParagraphRanker/entityqa/reader/results/single.mdl ]
Traceback (most recent call last):
  File "entityqa/pipeline/predict.py", line 367, in <module>
    main(args)
  File "entityqa/pipeline/predict.py", line 257, in main
    fixed_candidates=candidates)
  File "/home/jiangkunhaha/ParagraphRanker/entityqa/pipeline/root/pipeline/pipeline.py", line 74, in __init__
    self.reader = DocReader.load(reader_path, normalize=False)
  File "/home/jiangkunhaha/ParagraphRanker/entityqa/pipeline/root/reader/model.py", line 438, in load
    filename, map_location=lambda storage, loc: storage
  File "/home/jiangkunhaha/anaconda3/envs/ranker/lib/python3.6/site-packages/torch/serialization.py", line 358, in load
    return _load(f, map_location, pickle_module)
  File "/home/jiangkunhaha/anaconda3/envs/ranker/lib/python3.6/site-packages/torch/serialization.py", line 542, in _load
    result = unpickler.load()
ModuleNotFoundError: No module named 'drqa'

Lack of details about tdidf ranker

Need more information on these lines in entityqa/pipeline/predict.py,

RET_PATH = os.path.join(PATHS['WIKI'],
'docs-tfidf-ngram=2-hash=16777216-tokenizer=simple.npz')
DOC_DB_PATH = os.path.join(PATHS['WIKI'], 'docs.db')

Modification required in download.sh

In line 17 in download.sh,

wget -O "$DOWNLOAD_PATH_TAR" "https://s3.amazonaws.com/fair-data/drqa/data.tar.gz"

has to be replaced with,

wget -O "$DOWNLOAD_PATH_TAR" "https://dl.fbaipublicfiles.com/drqa/data.tar.gz"

Since the file location has been changed. :)