FFRecord

The FFRecord format is a simple format for storing a sequence of binary records developed by HFAiLab, which supports random access and Linux Asynchronous Input/Output (AIO) read.

File Format

Storage Layout:

+-----------------------------------+---------------------------------------+
|         checksum                  |             N                         |
+-----------------------------------+---------------------------------------+
|         checksums                 |           offsets                     |
+---------------------+---------------------+--------+----------------------+
|      sample 1       |      sample 2       | ....   |      sample N        |
+---------------------+---------------------+--------+----------------------+

Fields:

field	size (bytes)	description
checksum	4	CRC32 checksum of metadata
N	8	number of samples
checksums	4 * N	CRC32 checksum of each sample
offsets	8 * N	byte offset of each sample
sample i	offsets[i + 1] - offsets[i]	data of the i-th sample

Get Started

Requirements

• OS: Linux
• Python >= 3.6
• Pytorch >= 1.6
• NumPy
• tqdm
• zlib: sudo apt install zliblg-dev
• cmake: pip install cmake
• pybind11 >= 2.8

Install

python3 setup.py install

Usage

We provide ffrecord.FileWriter and ffrecord.FileReader for reading and writing, respectively.

Write

To create a FileWriter object, you need to specify a file name and the total number of samples. And then you could call FileWriter.write_one() to write a sample to the FFRecord file. It accepts bytes or bytearray as input and appends the data to the end of the opened file.

from ffrecord import FileWriter


def serialize(sample):
    """ Serialize a sample to bytes or bytearray

    You could use anything you like to serialize the sample.
    Here we simply use pickle.dumps().
    """
    return pickle.dumps(sample)


samples = [i for i in range(100)]  # anything you would like to store
fname = 'test.ffr'
n = len(samples)  # number of samples to be written
writer = FileWriter(fname, n)

for i in range(n):
    data = serialize(samples[i])  # data should be bytes or bytearray
    writer.write_one(data)

writer.close()

Read

To create a FileReader object, you only need to specify the file name. And then you could call FileWriter.read() to read multiple samples from the FFReocrd file. It accepts a list of indices as input and outputs the corresponding samples data.

The reader would validate the checksum before returning the data if check_data = True.

from ffrecord import FileReader


def deserialize(data):
    """ deserialize bytes data

    The deserialize method should be paired with the serialize method above.
    """
    return pickle.loads(data)


fname = 'test.ffr'
reader = FileReader(fname, check_data=True)
print(f'Number of samples: {reader.n}')

indices = [3, 6, 0, 10]      # indices of each sample
data = reader.read(indices)  # return a list of bytes-like data

for i in range(n):
    sample = deserialize(data[i])
    # do what you want

reader.close()

Dataset and DataLoader for PyTorch

We also provide ffrecord.torch.Dataset and ffrecord.torch.DataLoader for PyTorch users to train models using FFRecord.

Different from torch.utils.data.Dataset which accepts an index as input and returns only one sample, ffrecord.torch.Dataset accepts a batch of indices as input and returns a batch of samples. One advantage of ffrecord.torch.Dataset is that it could read a batch of data at a time using Linux AIO.

Users need to inherit from ffrecord.torch.Dataset and define their custom __getitem__() and __len__() function.

For example:

class CustomDataset(ffrecord.torch.Dataset):

    def __init__(self, fname, check_data=True, transform=None):
        self.reader = FileReader(fname, check_data)
        self.transform = transform

    def __len__(self):
        return self.reader.n

    def __getitem__(self, indices):
        # we read a batch of samples at once
        assert isintance(indices, list)
        data = self.reader.read(indices)

        # deserialize data
        samples = [pickle.loads(b) for b in data]

        # transform data
        if self.transform:
            samples = [self.transform(s) for s in samples]
        return samples

dataset = CustomDataset('train.ffr')
indices = [3, 4, 1, 0]
samples = dataset[indices]

ffrecord.torch.DataLoader is a drop-in replacement for PyTorch's standard dataloader. ffrecord.torch.Dataset could be combined with it just like PyTorch. ffrecord.torch.DataLoader supports for skipping steps during training by set_step() method.

dataset = CustomDataset('train.ffr')
loader = ffrecord.torch.DataLoader(dataset,
                                   batch_size=16,
                                   shuffle=True,
                                   num_workers=8)

start_epoch = 5
start_step = 100  # resume from epoch 5, step 100
loader.set_step(start_step)

for epoch in range(start_epoch, epochs):
    for i, batch in enumerate(loader):
        # training model

    loader.set_step(0)  # remember to reset before the next epoch

Pack a folder into ffrecord

FFRecord could also be used to pack a folder into a single file, which could be accessed without unpacking.

For example:

Assume we have a folder named just_a_folder:

$ tree just_a_folder

just_a_folder/
├── 001.txt
├── 002.txt
├── 003.txt
├── just_a_figure.png
└── just_another_folder
    ├── 004.txt
    ├── jsonfile.json
    ├── npyfile.npy
    ├── npzfile.npz
    └── another_folder
        └── 005.txt

Now we pack this folder into a file named packed.ffr:

from ffrecord import pack_folder
pack_folder("just_a_folder", "packed.ffr", verbose=True)

And then we could access the packed folder by PackedFolder:

>>> import io
>>> from ffrecord import PackedFolder
>>>
>>> folder = PackedFolder("packed.ffr")
>>> folder.list()
['001.txt', '002.txt', '003.txt', 'just_a_figure.png', 'just_another_folder']
>>> folder.list('just_another_folder')
['004.txt','jsonfile.json','npyfile.npy','npzfile.npz','another_folder']
>>> folder.is_file("just_another_folder")
False
>>> folder.is_dir("just_another_folder")
True
>>> folder.exists("just_another_folder/another_folder")
True
>>> fp = io.BytesIO(folder.read('001.txt'))
>>> data = fp.read()  # binary data
>>> list_of_data = folder.read(["001.txt", "002.txt"])  # read multiple files by Linux AIO

Here are some samples for reading file formats that are frequently used. Just replace your original code blocks with follows and enjoy FFRecord.

Images：

import cv2
order = "RGB"
path = "just_a_figure.png"
fp = io.BytesIO(folder.read_one(path))
img = cv2.imdecode(np.frombuffer(fp.read(), np.uint8), cv2.IMREAD_COLOR)
if order == 'RGB':
    img = img[:, :, ::-1].copy()
cv2.imwrite("test.png", img)

Texts:

fp = io.BytesIO(folder.read_one("just_another_folder/another_folder/005.txt"))
bytestring = fp.read()
result_str = bytestring.decode("utf-8")
print(result_str)

JSON：

import json
fp = io.BytesIO(folder.read_one("just_another_folder/jsonfile.json"))
bytestring = fp.read()
result_str = bytestring.decode("utf-8")
annot = json.loads(result_str)
print(annot)

Ndarrays saved in .npy file：

import numpy as np
fp = io.BytesIO(folder.read_one("just_another_folder/npyfile.npy"))
result = np.load(fp,allow_pickle=True)
print(result)

Ndarrays saved in .npz file:

# .npz file is a zip file for ndarrays, generated by np.savez
import numpy as np
import zipfile
fp = io.BytesIO(folder.read_one("just_another_folder/npzfile.npz"))
test = zipfile.ZipFile(fp,allowZip64=True)
print(test.namelist())
# arr_0 is a key in the namelist
with test.open('arr_0.npy',"r") as myfile:
    result = np.load(myfile,allow_pickle=True)

Write a file directly:

fp = io.BytesIO(folder.read_one("just_another_folder/npyfile.npy"))
with open("just_a_name", 'wb') as f:
    f.write(fp.read())