[BUG] 404 Client Error: Not Found for url: https://bpemb.h-its.org/multi/multi.wiki.bpe.vs100000.model

[ajndkr]

Describe the bug

AddressParser fails due to incorrect model path: https://bpemb.h-its.org/multi/multi.wiki.bpe.vs100000.model

To Reproduce

from deepparse.parser import AddressParser
address_parser = AddressParser(model_type="bpemb", device=0)

Expected behavior

I believe the class needs to download the model from the following url: https://bpemb.h-its.org/multi/multi/multi.wiki.bpe.vs1000000.model

[davebulaval]

Yeah, it appears they have changed the URL from https://bpemb.h-its.org/multi/multi.wiki.bpe.vs100000.model to https://bpemb.h-its.org/multi/multi/multi.wiki.bpe.vs1000000.model. The problem is not ours but BPEmb library.

It seems like an involuntary error on their side. I have opened an issue on the project repo here.

[akseljoonas]

Is there a fix or is the AddressParser with bpemb unusable for now?

[ajndkr]

@akseljoonas for the time being, until it is resolved on bpemb side, you can use this workaround.

mkdir -p ~/.cache/deepparse/multi/ && \
wget -O ~/.cache/deepparse/multi/multi.wiki.bpe.vs100000.model https://bpemb.h-its.org/multi/multi/multi.wiki.bpe.vs100000.model && \
wget -O multi.wiki.bpe.vs100000.d300.w2v.bin.tar.gz https://bpemb.h-its.org/multi/multi/multi.wiki.bpe.vs100000.d300.w2v.bin.tar.gz && \
tar -xzf multi.wiki.bpe.vs100000.d300.w2v.bin.tar.gz -C ~/.cache/deepparse/multi/ && rm multi.wiki.bpe.vs100000.d300.w2v.bin.tar.gz

I download these files into the cache folder before initialising AddressParser.

[davebulaval]

Is there a fix or is the AddressParser with bpemb unusable for now?

No, except for the @ajndkr solution.