-
download model, then run bert-as-service
pip install bert-serving-server bert-serving-start -model_dir ./model_dir -num_worker=4
-
in julia
include("BertClient.jl") BertClient.bert_encode(str)
bert_encode( [str1,str2,...] ) is also okay
bert_init() is optional,
default: tcp_server_address="127.0.0.1", port_client_input=5555, port_server_output=5556
remember to install tensorflow-gpu for (server side) better performance
no need to pack into a package I guess?