Session crashes when loading RerankingParser.from_unified_model_dir()

[pythonista0801]

I have installed NegBio, bllipparser with pip and imported the required modules including NegBioParser and RerankingParser.
However, when I try to load parser = NegBioParser() the session crashes on the line rrp = RerankingParser.from_unified_model_dir(self.model_dir), the reason in logs was first NO CUDA detected. I fixed it and switched to GPU runtime. After loading all dependencies and parsers, the session crashed again... it gives no message what's wrong, I can just see some logs such as :
gzip: stdout: Broken pipe.
OR
2022-07-13 17:54:43.937715: W tensorflow/core/common_runtime/gpu/gpu_bfc_allocator.cc:39] Overriding allow_growth setting because the TF_FORCE_GPU_ALLOW_GROWTH environment variable is set. Original config value was 0.
and then
Kernel restarted

[dmcc]

Interesting! (I wasn't aware NegBio was using BLLIP Parser, but that's cool to hear)

Looks like this is the same bug as ncbi-nlp/NegBio#52 -- my guess is that it's due to some interaction between the two projects and unfortunately I don't know much about the internals of NegBio. (BLLIP Parser was built way before there were GPU accelerators so there could easily be some issues with that? That could also be a red herring.)

Unless BLLIP Parser is busted (I just did a fresh install and it seems to work?) I suspect this issue is more on the NegBio side, I'm afraid, but either way more investigation will be needed.