TypeError: expected str, bytes or os.PathLike object, not NoneType

[XY2323819551]

Hello, I tried to reproduce the code of “ance-prf” followed by "experiments-ance-prf.md"， but the following error occurred. I the "encoders/ance-msmarco-passage" is downloaded from huggingface. Could you help me if it's convenient for you. Thanks very much!!

(ance-prf-inference) zhangxy@s2:~/QA/pyserini_for_ance-prf/pyserini$ python -m pyserini.search.faiss --topics msmarco-passage-dev-subset --index indexes/dindex-msmarco-passage-ance-bf-20210224-060cef --encoder encoders/ance-msmarco-passage --batch-size 32 --output runs/run.marco-passagev1.ance-prf3.tsv --prf-depth 3 --prf-method ance-prf --threads 12 --sparse-index indexes/index-msmarco-passage-20201117-f87c94 --ance-prf-encoder ckpt/ance_prf_k3_checkpoint --output-format msmarco
Using pre-defined topic order for msmarco-passage-dev-subset
Traceback (most recent call last):
  File "/home/zhangxy/anaconda3/envs/ance-prf-inference/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/zhangxy/anaconda3/envs/ance-prf-inference/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/zhangxy/QA/pyserini_for_ance-prf/pyserini/pyserini/search/faiss/__main__.py", line 189, in <module>
    prf_query_encoder = AnceQueryEncoder(encoder_dir=args.ance_prf_encoder, tokenizer_name=args.tokenizer,
  File "/home/zhangxy/QA/pyserini_for_ance-prf/pyserini/pyserini/search/faiss/_searcher.py", line 228, in __init__
    self.tokenizer = RobertaTokenizer.from_pretrained(tokenizer_name or encoder_dir)
  File "/home/zhangxy/anaconda3/envs/ance-prf-inference/lib/python3.8/site-packages/transformers/tokenization_utils_base.py", line 1780, in from_pretrained
    return cls._from_pretrained(
  File "/home/zhangxy/anaconda3/envs/ance-prf-inference/lib/python3.8/site-packages/transformers/tokenization_utils_base.py", line 1915, in _from_pretrained
    tokenizer = cls(*init_inputs, **init_kwargs)
  File "/home/zhangxy/anaconda3/envs/ance-prf-inference/lib/python3.8/site-packages/transformers/models/roberta/tokenization_roberta.py", line 228, in __init__
    with open(merges_file, encoding="utf-8") as merges_handle:
TypeError: expected str, bytes or os.PathLike object, not NoneType
(ance-prf-inference) zhangxy@s2:~/QA/pyserini_for_ance-prf/pyserini$ ^C

[XY2323819551]

I got it! The checkpoint provided by the original author was missing the merges.txt file, I added this file to the checkpoint directory and then the code ran successfully!

[lintool]

Would you mind sending a PR to update the documentation?

[XY2323819551]

Would you mind sending a PR to update the documentation?

OK. I will do it now！