PaddleX/paddlex/inference/predictors/base.py

# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
#
# 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 yaml
import codecs
from pathlib import Path
from abc import abstractmethod

import GPUtil

from ...utils.subclass_register import AutoRegisterABCMetaClass
from ..utils.device import constr_device
from ...utils import logging
from ..components.base import BaseComponent, ComponentsEngine
from ..utils.pp_option import PaddlePredictorOption
from ..utils.process_hook import generatorable_method


def _get_default_device():
    avail_gpus = GPUtil.getAvailable()
    if not avail_gpus:
        return "cpu"
    else:
        return constr_device("gpu", [avail_gpus[0]])


class BasePredictor(BaseComponent):
    KEEP_INPUT = False
    YIELD_BATCH = False

    INPUT_KEYS = "x"
    DEAULT_INPUTS = {"x": "x"}
    OUTPUT_KEYS = "result"
    DEAULT_OUTPUTS = {"result": "result"}

    MODEL_FILE_PREFIX = "inference"

    def __init__(self, model_dir, config=None, device=None, **kwargs):
        super().__init__()
        self.model_dir = Path(model_dir)
        self.config = config if config else self.load_config(self.model_dir)
        self.device = device if device else _get_default_device()
        self.kwargs = self._check_args(kwargs)
        # alias predict() to the __call__()
        self.predict = self.__call__

    def __call__(self, *args, **kwargs):
        for res in super().__call__(*args, **kwargs):
            yield res["result"]

    @property
    def config_path(self):
        return self.get_config_path(self.model_dir)

    @property
    def model_name(self) -> str:
        return self.config["Global"]["model_name"]

    @abstractmethod
    def apply(self, x):
        raise NotImplementedError

    @classmethod
    def get_config_path(cls, model_dir):
        return model_dir / f"{cls.MODEL_FILE_PREFIX}.yml"

    @classmethod
    def load_config(cls, model_dir):
        config_path = cls.get_config_path(model_dir)
        with codecs.open(config_path, "r", "utf-8") as file:
            dic = yaml.load(file, Loader=yaml.FullLoader)
        return dic

    def _check_args(self, kwargs):
        return kwargs


class BasicPredictor(BasePredictor, metaclass=AutoRegisterABCMetaClass):

    __is_base = True

    def __init__(self, model_dir, config=None, device=None, pp_option=None, **kwargs):
        super().__init__(model_dir=model_dir, config=config, device=device, **kwargs)
        self.pp_option = PaddlePredictorOption() if pp_option is None else pp_option
        self.pp_option.set_device(self.device)
        self.components = self._build_components()
        self.engine = ComponentsEngine(self.components)
        logging.debug(
            f"-------------------- {self.__class__.__name__} --------------------\nModel: {self.model_dir}\nEnv: {self.pp_option}"
        )

    def apply(self, x):
        """predict"""
        yield from self._generate_res(self.engine(x))

    @generatorable_method
    def _generate_res(self, batch_data):
        return [{"result": self._pack_res(data)} for data in batch_data]

    @abstractmethod
    def _build_components(self):
        raise NotImplementedError

    @abstractmethod
    def _pack_res(self, data):
        raise NotImplementedError