I was able to run the full pegasus pipeline that uses the docker image produced from this Dockerfile.
So I think it can be merged. There is a question on the python_tap_amr_parsing step where it
took maybe an hour for 5 documents. The whole pipeline on 5 docs too 132m. It's possible
amr parsing was not properly using a gpu, but that's just speculation at this point.
I was able to run the full pegasus pipeline that uses the docker image produced from this Dockerfile. So I think it can be merged. There is a question on the
python_tap_amr_parsingstep where it took maybe an hour for 5 documents. The whole pipeline on 5 docs too 132m. It's possible amr parsing was not properly using a gpu, but that's just speculation at this point.