# Copyright 2022 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
"""Zero Shot Image Classification Dataset."""
import os
from mindformers.version_control import get_dataset_map
from .dataloader import build_dataset_loader
from .transforms import build_transforms
from .base_dataset import BaseDataset
from ..tools import logger
from ..tools.register import MindFormerRegister, MindFormerModuleType
from ..models.build_tokenizer import build_tokenizer
[文档]@MindFormerRegister.register(MindFormerModuleType.DATASET)
class ZeroShotImageClassificationDataset(BaseDataset):
r"""
Zero Shot Image Classification Dataset API.
output image, text, and label columns
Args:
dataset_config (dict): Config for dataset.
Returns:
A dataset for ZeroShotImageClassificationTrainer.
Examples:
>>> import os
>>> from mindformers import MindFormerBook, MindFormerConfig, build_dataset
>>> project_path = MindFormerBook.get_project_path()
>>> config_path = os.path.join(project_path, "configs", "clip",
>>> "run_clip_vit_b_32_zero_shot_image_classification_cifar100.yaml")
>>> config = MindFormerConfig(config_path)
Note:
Put cifar100 dataset to ./
The detailed data setting could refer to ./configs/clip/clip.md
>>> config.eval_dataset_task.dataset_config.batch_size = 1
>>> dataset = build_dataset(config.eval_dataset_task)
>>> for item in dataset:
>>> print(item)
>>> break
[Tensor(shape=[1, 3, 224, 224], dtype=Float32, value=
[[[[1.11282456e+000, 1.11282456e+000, ... 1.47778523e+000, 1.47778523e+000],
[1.11282456e+000, 1.11282456e+000, ... 1.47778523e+000, 1.47778523e+000],
[1.11282456e+000, 1.11282456e+000, ... 1.47778523e+000, 1.47778523e+000],
...
[1.97748125e-001, 1.97748125e-001, ... 1.12205243e+000, 1.12205243e+000],
[1.97748125e-001, 1.97748125e-001, ... 1.12205243e+000, 1.12205243e+000],
[1.97748125e-001, 1.97748125e-001, ... 1.12205243e+000, 1.12205243e+000]]]]),
Tensor(shape=[1, 100, 77], dtype=Int32, value=
[[[49406, 320, 1674 ... 0, 0, 0],
[49406, 320, 1674 ... 0, 0, 0],
[49406, 320, 1674 ... 0, 0, 0],
...
[49406, 320, 1674 ... 0, 0, 0],
[49406, 320, 1674 ... 0, 0, 0],
[49406, 320, 1674 ... 0, 0, 0]]]),
Tensor(shape=[1], dtype=Int32, value= [49])]
"""
def __new__(cls, dataset_config: dict = None):
"""New method"""
logger.info("Now Create Zero Shot Image Classification Dataset.")
cls.init_dataset_config(dataset_config)
rank_id = int(os.getenv("RANK_ID", "0"))
device_num = int(os.getenv("RANK_SIZE", "1"))
dataset = build_dataset_loader(
dataset_config.data_loader, default_args={'num_shards': device_num, 'shard_id': rank_id})
transforms = build_transforms(dataset_config.transforms)
tokenizer = build_tokenizer(dataset_config.tokenizer)
text_transforms = build_transforms(dataset_config.text_transforms,
default_args={"tokenizer": tokenizer})
if transforms is not None:
dataset = get_dataset_map(dataset,
input_columns="image",
operations=transforms,
num_parallel_workers=dataset_config.num_parallel_workers,
python_multiprocessing=dataset_config.python_multiprocessing)
if text_transforms is not None:
dataset = get_dataset_map(dataset,
input_columns="text",
operations=text_transforms,
num_parallel_workers=dataset_config.num_parallel_workers,
python_multiprocessing=dataset_config.python_multiprocessing)
dataset = dataset.batch(dataset_config.batch_size,
drop_remainder=dataset_config.drop_remainder,
num_parallel_workers=dataset_config.num_parallel_workers)
dataset = dataset.repeat(dataset_config.repeat)
return dataset