codebert/codebert.py

# Copyright 2021 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import numpy
import os
import torch
from pathlib import Path
from transformers import AutoTokenizer, AutoModel

from towhee.operator import NNOperator
from towhee import register

import warnings
import logging

warnings.filterwarnings('ignore')
logging.getLogger('transformers').setLevel(logging.ERROR)
log = logging.getLogger()


@register(output_schema=['vec'])
class CodeBert(NNOperator):
    """
    An operator generates an embedding for code or natural language text
    using a pretrained codebert model gathered by huggingface.

    Args:
        model_name (`str`):
            Which model to use for the embeddings.
        device (`str`):
            Device to run model inference. Defaults to None, enable GPU when it is available.
    """

    def __init__(self, model_name: str = 'huggingface/CodeBERTa-small-v1', device: str = None):
        super().__init__()
        self.model_name = model_name
        assert modality in ['nlp', 'code'], 'Invalid modality value. Accept only "nlp" or "code".'

        if device is None:
            device = 'cuda' if torch.cuda.is_available() else 'cpu'
        self.device = device

        try:
            self.model = AutoModel.from_pretrained(model_name).to(self.device)
            self.model.eval()
        except Exception as e:
            model_list = self.supported_model_names()
            if model_name not in model_list:
                log.error(f'Invalid model name: {model_name}. Supported model names: {model_list}')
            else:
                log.error(f'Fail to load model by name: {self.model_name}')
            raise e
        try:
            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
        except Exception as e:
            log.error(f'Fail to load tokenizer by name: {self.model_name}')
            raise e

    def __call__(self, txt: str) -> numpy.ndarray:
        try:
            inputs = self.tokenizer.encode(txt, return_tensors='pt').to(self.device)
        except Exception as e:
            log.error(f'Invalid input for the tokenizer: {self.model_name}')
            raise e
        try:
            outs = self.model(inputs)
        except Exception as e:
            log.error(f'Invalid input for the model: {self.model_name}')
            raise e
        try:
            features = outs['pooler_output'].squeeze(0)
        except Exception as e:
            log.error(f'Fail to extract features by model: {self.model_name}')
            raise e
        vec = features.cpu().detach().numpy()
        return vec

    def save_model(self, format: str = 'pytorch', path: str = 'default'):
        if path == 'default':
            path = str(Path(__file__).parent)
            path = os.path.join(path, 'saved', format)
            os.makedirs(path, exist_ok=True)
            name = self.model_name.replace('/', '-')
            path = os.path.join(path, name)
        inputs = self.tokenizer.encode('test', return_tensors='pt').to(self.device)  # return a tensor of token ids
        if format == 'pytorch':
            path = path + '.pt'
            torch.save(self.model, path)
        elif format == 'torchscript':
            path = path + '.pt'
            inputs = list(inputs)
            try:
                try:
                    jit_model = torch.jit.script(self.model)
                except Exception:
                    jit_model = torch.jit.trace(self.model, inputs, strict=False)
                torch.jit.save(jit_model, path)
            except Exception as e:
                log.error(f'Fail to save as torchscript: {e}.')
                raise RuntimeError(f'Fail to save as torchscript: {e}.')
        elif format == 'onnx':
            path = path + '.onnx'
            try:
                torch.onnx.export(self.model,
                                  tuple(inputs),
                                  path,
                                  input_names=['input_ids'],  # list(inputs.keys())
                                  output_names=['last_hidden_state', 'pooler_output'],
                                  opset_version=12,
                                  dynamic_axes={
                                      'input_ids': {0: 'batch_size', 1: 'input_length'},
                                      'last_hidden_state': {0: 'batch_size'},
                                      'pooler_output': {0: 'batch_size', 1: 'output_dim'},
                                  })
            except Exception:
                torch.onnx.export(self.model,
                                  tuple(inputs.values()),
                                  path,
                                  input_names=['input_ids'],  # list(inputs.keys())
                                  output_names=['last_hidden_state'],
                                  opset_version=12,
                                  dynamic_axes={
                                      'input_ids': {0: 'batch_size', 1: 'input_length'},
                                      'last_hidden_state': {0: 'batch_size'},
                                  })
        # todo: elif format == 'tensorrt':
        else:
            log.error(f'Unsupported format "{format}".')

    @staticmethod
    def supported_model_names(format: str = None):
        full_list = [
            'huggingface/CodeBERTa-small-v1',
            'microsoft/codebert-base',
            'microsoft/codebert-base-mlm',
            'mrm8488/codebert-base-finetuned-stackoverflow-ner'
        ]
        full_list.sort()
        if format is None:
            model_list = full_list
        elif format == 'pytorch':
            to_remove = []
            assert set(to_remove).issubset(set(full_list))
            model_list = list(set(full_list) - set(to_remove))
        # todo: elif format == 'torchscript':
        #     to_remove = [
        #     ]
        #     assert set(to_remove).issubset(set(full_list))
        #     model_list = list(set(full_list) - set(to_remove))
        # todo: elif format == 'onnx':
        #     to_remove = []
        #     assert set(to_remove).issubset(set(full_list))
        #     model_list = list(set(full_list) - set(to_remove))
        # todo: elif format == 'tensorrt':
        else:
            log.error(f'Invalid format "{format}". Currently supported formats: "pytorch".')
        return model_list