danieljunior/bert-ner

Pytorch-Named-Entity-Recognition-with-BERT

BERT NER PT-BR

Requirements

• python3
• pip3 install -r requirements.txt

Run

• Edit tags.txt according your dataset.

Using DGX

• docker build -t bert_ner .

• NV_GPU=1 nvidia-docker run -itd --rm --shm-size=32g --ulimit memlock=-1 -v ${PWD}:/app bert_ner bash

• python run_ner.py --data_dir=data/ --bert_model=bert-base-cased-pt-br --task_name=ner --output_dir=out_base --max_seq_length=128 --do_train --num_train_epochs 5 --do_eval --warmup_proportion=0.1 --eval_on=test

Otherwise

python run_ner.py --data_dir=data/ --bert_model=bert-base-cased-pt-br --task_name=ner --output_dir=out_base --max_seq_length=128 --do_train --num_train_epochs 5 --do_eval --warmup_proportion=0.1

Limiting batch size

python run_ner.py --data_dir=data/ --bert_model=bert-base-cased-pt-br --train_batch_size=8 --task_name=ner --output_dir=out_base --max_seq_length=128 --do_train --num_train_epochs 5 --do_eval --warmup_proportion=0.1

BERT-BASE Pretrained model download from here

** To use pytorch with notebook: conda install pytorch torchvision cuda90 -c pytorch

Inference

from bert import Ner

model = Ner("out_base/")

output = model.predict("Steve went to Paris")

print(output)
'''
    [
        {
            "confidence": 0.9981840252876282,
            "tag": "B-PER",
            "word": "Steve"
        },
        {
            "confidence": 0.9998939037322998,
            "tag": "O",
            "word": "went"
        },
        {
            "confidence": 0.999891996383667,
            "tag": "O",
            "word": "to"
        },
        {
            "confidence": 0.9991968274116516,
            "tag": "B-LOC",
            "word": "Paris"
        }
    ]
'''

Inference C++

Pretrained and converted bert-base model download from here

Download libtorch from here

• install cmake, tested with cmake version 3.10.2

• unzip downloaded model and libtorch in BERT-NER

• Compile C++ App

    cd cpp-app/
    cmake -DCMAKE_PREFIX_PATH=../libtorch

  

  make

  

• Runing APP

     ./app ../base

  

NB: Bert-Base C++ model is split in to two parts.

• Bert Feature extractor and NER classifier.
• This is done because jit trace don't support input depended for loop or if conditions inside forword function of model.

Deploy REST-API

BERT NER model deployed as rest api

python api.py

API will be live at 0.0.0.0:8000 endpoint predict

cURL request

curl -X POST http://0.0.0.0:8000/predict -H 'Content-Type: application/json' -d '{ "text": "Steve went to Paris" }'

Output

{
    "result": [
        {
            "confidence": 0.9981840252876282,
            "tag": "B-PER",
            "word": "Steve"
        },
        {
            "confidence": 0.9998939037322998,
            "tag": "O",
            "word": "went"
        },
        {
            "confidence": 0.999891996383667,
            "tag": "O",
            "word": "to"
        },
        {
            "confidence": 0.9991968274116516,
            "tag": "B-LOC",
            "word": "Paris"
        }
    ]
}

cURL

Postman

C++ unicode support

• http://github.com/ufal/unilib

Tensorflow version

• https://github.com/kyzhouhzau/BERT-NER