Skip to content

Latest commit

 

History

History
221 lines (176 loc) · 8.91 KB

README.md

File metadata and controls

221 lines (176 loc) · 8.91 KB
Raw

Dataflux Dataset for PyTorch with Google Cloud Storage

The Dataflux Dataset for PyTorch lets you connect directly to a GCS bucket as a PyTorch dataset, without pre-loading the data to local disk, and with optimizations to make training up to 3X faster when the dataset consists of many small files (e.g., 100 - 500 KB).

The Dataflux Dataset for PyTorch implements PyTorch’s dataset primitive that can be used to efficiently load training data from GCS. The library currently supports map-style datasets for random data access patterns.

Furthermore, the Dataflux Dataset for PyTorch provides a checkpointing interface to conveniently save and load checkpoints directly to and from a Google Cloud Storage (GCS) bucket.

Note that the Dataflux Dataset for PyTorch library is in an early preview stage and the team is consistently working on improvements and support for new features.

Getting started

Prerequisites

  • Python 3.8 or greater is installed (Note: Using 3.12+ is not recommended as PyTorch does not support).

Installation

git clone --recurse-submodules https://github.com/GoogleCloudPlatform/dataflux-pytorch
cd dataflux-pytorch
pip install .

Configuration

Authentication must be provided to use the Dataflux Dataset for PyTorch via Application Default Credentials through one of the following methods:

  1. While running this library on a GCE VM, Application Default Credentials will automatically use the VM’s attached service account by default. More details can be found here.
  2. Application Default Credentials can also be configured manually as described here. The quickest way is to log in directly using the gcloud CLI:
gcloud auth application-default login

Examples

Before getting started, please make sure you have installed the library and configured authentication following the instructions above.

Data Loading

Dataflux Dataset for PyTorch can be constructed by specifying the project name, bucket name and an optional prefix.

from dataflux_pytorch import dataflux_mapstyle_dataset

# Please update these fields.
PROJECT_NAME = "<PROJECT_NAME>"
BUCKET_NAME = "<BUCKET_NAME>"
PREFIX = "<PREFIX>"

dataset = dataflux_mapstyle_dataset.DataFluxMapStyleDataset(
  project_name=PROJECT_NAME,
  bucket_name=BUCKET_NAME,
  prefix=PREFIX,
)

# Random access to an object.
sample_object = dataset.objects[0]

# Learn about the name and the size (in bytes) of the object.
name = sample_object[0]
size = sample_object[1]

# Iterate over the datasets.
for each_object in dataset:
  # Raw bytes of the object.
  print(each_object)

Dataflux Dataset for PyTorch offers the flexibility to transform the downloaded raw bytes of data into any format of choice.

# Assume that you have a bucket with image files and you want
# to load them into Numpy arrays for training.
import io
import numpy
from PIL import Image

transform = lambda img_in_bytes : numpy.asarray(Image.open(io.BytesIO(img_in_bytes)))

dataset = dataflux_mapstyle_dataset.DataFluxMapStyleDataset(
  project_name=PROJECT_NAME,
  bucket_name=BUCKET_NAME,
  prefix=PREFIX,
  data_format_fn=transform,
)

for each_object in dataset:
  # each_object is now a Numpy array.
  print(each_object)

Checkpointing

The Dataflux Dataset for PyTorch supports fast data loading and allows the user to save and load model checkpoints directly to/from a Google Cloud Storage (GCS) bucket.

import torch
import torchvision

from dataflux_pytorch import dataflux_checkpoint

ckpt = dataflux_checkpoint.DatafluxCheckpoint(
  project_name=PROJECT_NAME, bucket_name=BUCKET_NAME
)
CKPT_PATH = "checkpoints/epoch0.ckpt"

model = torchvision.models.resnet50()

with ckpt.writer(CKPT_PATH) as writer:
  torch.save(model.state_dict(), writer)

with ckpt.reader(CKPT_PATH) as reader:
  read_state_dict = torch.load(reader)

model.load_state_dict(read_state_dict)

Performance

We tested Dataflux's early performance using DLIO benchmark simulations with standard mean file-sizes and dataset sizes. A total of 5 training epochs were simulated. For small files (100KB, 500KB), Dataflux can be 2-3x faster than using GCS native APIs.

File size / count Tool Training time (s)
100 KiB / 500000 files Direct GCS API calls 2,459
Dataflux 757
500 KiB / 2.2m files Direct GCS API calls 7,472
Dataflux 2,696
3 MiB / 50000 files Direct GCS API calls 463
Dataflux 318
150 MiB / 5000 files Direct GCS API calls 1,228
Dataflux 1,288

Note: Within each experiment, all training parameters such as batch size and parallelism are consistent. The team is working on publishing a detailed analysis soon.

Limitations

Billing

To optimize listing performance, the Dataflux Dataset for PyTorch library utilizes a “fast listing” algorithm developed to balance the listing workload among parallelized GCS object listing processes. Therefore, the library will cause more listing API calls to be made than a regular sequential listing, which are charged as Class A operations.

Composite Objects

To optimize the download performance of small files, the Dataflux Dataset for PyTorch library utilizes the GCS Compose API to concatenate a set of smaller objects into a new and larger one in the same bucket under a folder named “dataflux-composed-objects”. The new composite objects will be removed at the end of your training loop but in rare cases that they don’t, you can run this command to clean the composed files.

gcloud storage rm --recursive gs://<my-bucket>/dataflux-composed-objects/

You can also turn off this behavior by setting the “max_composite_object_size” parameter to 0 when constructing the dataset. Example:

dataset = dataflux_mapstyle_dataset.DataFluxMapStyleDataset(
  project_name=PROJECT_NAME,
  bucket_name=BUCKET_NAME,
  prefix=PREFIX,
  config=dataflux_mapstyle_dataset.Config(max_composite_object_size=0),
)

Note that turning off this behavior will cause the training loop to take significantly longer to complete when working with small files.

Google Cloud Storage Class

Due to the quick creation and deletion of composite objects, we recommend that only the Standard storage class is applied to your bucket to minimize cost and maximize performance.

Contributing

We welcome your feedback, issues, and bug fixes. We have a tight roadmap at this time so if you have a major feature or change in functionality you'd like to contribute, please open a GitHub Issue for discussion prior to sending a pull request. Please see CONTRIBUTING for more information on how to report bugs or submit pull requests.

Code of Conduct

This project has adopted the Google Open Source Code of Conduct. Please see code-of-conduct.md for more information.

License

The Dataflux Dataset for PyTorch library has an Apache License 2.0. Please see the LICENSE file for more information.