• Stars
    star
    123
  • Rank 290,145 (Top 6 %)
  • Language
    Python
  • License
    Apache License 2.0
  • Created about 3 years ago
  • Updated 2 months ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Megvii FILE Library - Working with Files in Python same as the standard library

megfile - Megvii FILE library

Build Documents Codecov Latest version Support python versions License CII Best Practices

megfile provides a silky operation experience with different backends (currently including local file system and s3), which enable you to focus more on the logic of your own project instead of the question of "Which backend is used for this file?"

megfile provides:

  • Almost unified file system operation experience. Target path can be easily moved from local file system to s3.
  • Complete boundary case handling. Even the most difficult (or even you can't even think of) boundary conditions, megfile can help you easily handle it.
  • Perfect type hints and built-in documentation. You can enjoy the IDE's auto-completion and static checking.
  • Semantic version and upgrade guide, which allows you enjoy the latest features easily.

megfile's advantages are:

  • smart_open can open resources that use various protocols. Especially, reader / writer of s3 in megfile is implemented with multi-thread, which is faster than known competitors.
  • smart_glob is available on majority protocols. And it supports zsh extended pattern syntax of [], e.g. s3://bucket/video.{mp4,avi}.
  • All-inclusive functions like smart_exists / smart_stat / smart_sync. If you don't find the functions you want, submit an issue.
  • Compatible with pathlib.Path interface, referring to SmartPath and other protocol classes like S3Path.

Support Protocols

  • fs(local filesystem)
  • s3
  • sftp
  • http
  • stdio
  • hdfs: pip install 'megfile[hdfs]'

Quick Start

Path string in megfile almost is protocol://path/to/file, for example s3://bucketA/key. But sftp path is a little different, format is sftp://[username[:password]@]hostname[:port]//file_path, and relative path is replace //file_path to /file_path. Here's an example of writing a file to s3 / fs, syncing to local, reading and finally deleting it.

Functional Interface

from megfile import smart_open, smart_exists, smart_sync, smart_remove, smart_glob

# open a file in s3 bucket
with smart_open('s3://playground/megfile-test', 'w') as fp:
    fp.write('megfile is not silver bullet')

# test if file in s3 bucket exist
smart_exists('s3://playground/megfile-test')

# or in local file system
smart_exists('/tmp/playground/megfile-test')

# copy files or directories
smart_sync('s3://playground/megfile-test', '/tmp/playground/megfile-test')

# remove files or directories
smart_remove('s3://playground/megfile-test')

# glob files or directories in s3 bucket
smart_glob('s3://playground/megfile-?.{mp4,avi}')

SmartPath Interface

SmartPath has a similar interface with pathlib.Path.

from megfile.smart_path import SmartPath

path = SmartPath('s3://playground/megfile-test')
if path.exists():
    with path.open() as f:
        result = f.read(7)
        assert result == b'megfile'

Command Line Interface

$ megfile --help  # see what you can do

$ megfile ls s3://playground/
$ megfile ls -l -h s3://playground/

$ megfile cat s3://playground/megfile-test

$ megfile cp s3://playground/megfile-test /tmp/playground/megfile-test

Installation

PyPI

pip3 install megfile

You can specify megfile version as well

pip3 install "megfile~=0.0"

Build from Source

megfile can be installed from source

git clone [email protected]:megvii-research/megfile.git
cd megfile
pip3 install -U .

Development Environment

git clone [email protected]:megvii-research/megfile.git
cd megfile
pip3 install -r requirements.txt -r requirements-dev.txt

Configuration

Using s3 as an example, the following describes the configuration methods. For more details, please refer to Configuration.

You can use environments and configuration file for configuration, and priority is that environment variables take precedence over configuration file.

Use environments

You can use environments to setup authentication credentials for your s3 account:

  • AWS_ACCESS_KEY_ID: access key
  • AWS_SECRET_ACCESS_KEY: secret key
  • OSS_ENDPOINT: endpoint url of s3
  • AWS_S3_ADDRESSING_STYLE: addressing style

Use command

You can update config file with megfile command easyly: megfile config s3 [OPTIONS] AWS_ACCESS_KEY_ID AWS_SECRET_ACCESS_KEY

$ megfile config s3 accesskey secretkey

# for aliyun
$ megfile config s3 accesskey secretkey \
--addressing-style virtual \
--endpoint-url http://oss-cn-hangzhou.aliyuncs.com \

You can get the configuration from ~/.aws/credentials, like:

[default]
aws_secret_access_key = accesskey
aws_access_key_id = secretkey

s3 =
    addressing_style = virtual
    endpoint_url = http://oss-cn-hangzhou.aliyuncs.com

How to Contribute

  • We welcome everyone to contribute code to the megfile project, but the contributed code needs to meet the following conditions as much as possible:

    You can submit code even if the code doesn't meet conditions. The project members will evaluate and assist you in making code changes

    • Code format: Your code needs to pass code format check. megfile uses yapf as lint tool and the version is locked at 0.27.0. The version lock may be removed in the future

    • Static check: Your code needs complete type hint. megfile uses pytype as static check tool. If pytype failed in static check, use # pytype: disable=XXX to disable the error and please tell us why you disable it.

      Note : Because pytype doesn't support variable type annation, the variable type hint format introduced by py36 cannot be used.

      i.e. variable: int is invalid, replace it with variable # type: int

    • Test: Your code needs complete unit test coverage. megfile uses pyfakefs and moto as local file system and s3 virtual environment in unit tests. The newly added code should have a complete unit test to ensure the correctness

  • You can help to improve megfile in many ways:

    • Write code.
    • Improve documentation.
    • Report or investigate bugs and issues.
    • If you find any problem or have any improving suggestion, submit a new issuse as well. We will reply as soon as possible and evaluate whether to adopt.
    • Review pull requests.
    • Star megfile repo.
    • Recommend megfile to your friends.
    • Any other form of contribution is welcomed.

More Repositories

1

NAFNet

The state-of-the-art image restoration model without nonlinear activation functions.
Python
2,195
star
2

ML-GCN

PyTorch implementation of Multi-Label Image Recognition with Graph Convolutional Networks, CVPR 2019.
Python
1,408
star
3

PETR

[ECCV2022] PETR: Position Embedding Transformation for Multi-View 3D Object Detection & [ICCV2023] PETRv2: A Unified Framework for 3D Perception from Multi-Camera Images
Python
862
star
4

video_analyst

A series of basic algorithms that are useful for video understanding, including Single Object Tracking (SOT), Video Object Segmentation (VOS) and so on.
Python
829
star
5

mdistiller

The official implementation of [CVPR2022] Decoupled Knowledge Distillation https://arxiv.org/abs/2203.08679 and [ICCV2023] DOT: A Distillation-Oriented Trainer https://openaccess.thecvf.com/content/ICCV2023/papers/Zhao_DOT_A_Distillation-Oriented_Trainer_ICCV_2023_paper.pdf
Python
801
star
6

IJCAI2023-CoNR

IJCAI2023 - Collaborative Neural Rendering using Anime Character Sheets
Jupyter Notebook
797
star
7

HiDiffusion

[ECCV 2024] HiDiffusion: Increases the resolution and speed of your diffusion model by only adding a single line of code!
Jupyter Notebook
752
star
8

megactor

Python
742
star
9

BBN

The official PyTorch implementation of paper BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognition
Python
659
star
10

MOTR

[ECCV2022] MOTR: End-to-End Multiple-Object Tracking with TRansformer
Python
614
star
11

neural-painter

Paint artistic patterns using random neural network.
Python
532
star
12

CREStereo

Official MegEngine implementation of CREStereo(CVPR 2022 Oral).
Python
483
star
13

megvii-pku-dl-course

Homepage for the joint course of Megvii Inc. and Peking University on Deep Learning.
Python
445
star
14

MOTRv2

[CVPR2023] MOTRv2: Bootstrapping End-to-End Multi-Object Tracking by Pretrained Object Detectors
Python
364
star
15

AnchorDETR

An official implementation of the Anchor DETR.
Python
335
star
16

MSPN

Multi-Stage Pose Network
Python
334
star
17

Sparsebit

A model compression and acceleration toolbox based on pytorch.
Python
325
star
18

FQ-ViT

[IJCAI 2022] FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer
Python
304
star
19

FSCE

Python
280
star
20

OccDepth

Maybe the first academic open work on stereo 3D SSC method with vision-only input.
Python
278
star
21

TransMVSNet

(CVPR 2022) TransMVSNet: Global Context-aware Multi-view Stereo Network with Transformers.
Python
268
star
22

RevCol

Official Code of Paper "Reversible Column Networks" "RevColv2"
Python
248
star
23

TLC

Test-time Local Converter
Python
229
star
24

DCLS-SR

Official PyTorch implementation of the paper "Deep Constrained Least Squares for Blind Image Super-Resolution", CVPR 2022.
Python
220
star
25

SOLQ

"SOLQ: Segmenting Objects by Learning Queries", SOLQ is an end-to-end instance segmentation framework with Transformer.
Python
198
star
26

introduction-neural-3d-reconstruction

Course materials for Introduction to Neural 3D Reconstruction
185
star
27

AAAI2023-PVD

Official Implementation of PVD and PVDAL: http://sk-fun.fun/PVD-AL/
Python
183
star
28

tf-tutorials

Tutorials for deep learning course here:
Jupyter Notebook
180
star
29

DPGN

[CVPR 2020] DPGN: Distribution Propagation Graph Network for Few-shot Learning.
Python
178
star
30

CADDM

Official implementation of ID-unaware Deepfake Detection Model
C++
146
star
31

Far3D

[AAAI2024] Far3D: Expanding the Horizon for Surround-view 3D Object Detection
Jupyter Notebook
140
star
32

PMN

[TPAMI 2023 / ACMMM 2022 Best Paper Runner-Up Award] Learnability Enhancement for Low-light Raw Denoising: Where Paired Real Data Meets Noise Modeling (a Data Perspective)
Python
131
star
33

CR-DA-DET

The official PyTorch implementation of paper Exploring Categorical Regularization for Domain Adaptive Object Detection (CR-DA-DET)
Python
115
star
34

CVPR2023-UniDistill

CVPR2023 (highlight) - UniDistill: A Universal Cross-Modality Knowledge Distillation Framework for 3D Object Detection in Bird's-Eye View
Python
103
star
35

TreeEnergyLoss

[CVPR2022] Tree Energy Loss: Towards Sparsely Annotated Semantic Segmentation
Python
102
star
36

hpman

A hyperparameter manager for deep learning experiments.
Python
95
star
37

RealFlow

The official implementation of the ECCV 2022 Oral paper: RealFlow: EM-based Realistic Optical Flow Dataset Generation from Videos
Python
93
star
38

HDR-Transformer

The official MegEngine implementation of the ECCV 2022 paper: Ghost-free High Dynamic Range Imaging with Context-aware Transformer
Python
90
star
39

Iter-E2EDET

Official implementation of the paper "Progressive End-to-End Object Detection in Crowded Scenes"
Python
88
star
40

cv-master-ex

torch version of instant-ngp, image rendering
C++
80
star
41

FSSD_OoD_Detection

[SafeAI'21] Feature Space Singularity for Out-of-Distribution Detection.
Python
80
star
42

SSQL-ECCV2022

PyTorch implementation of SSQL (Accepted to ECCV2022 oral presentation)
Python
75
star
43

expman

Shell
62
star
44

BasesHomo

The official PyTorch implementation of the paper "Motion Basis Learning for Unsupervised Deep Homography Estimation with Subspace Projection".
Python
61
star
45

megvii-tsinghua-dl-course

Slides with modifications for a course at Tsinghua University.
57
star
46

LGD

Official Implementation of the detection self-distillation framework LGD.
Python
53
star
47

protoclip

📍 Official pytorch implementation of paper "ProtoCLIP: Prototypical Contrastive Language Image Pretraining" (IEEE TNNLS)
Python
46
star
48

D2C-SR

Official MegEngine implementation of ECCV2022 "D2C-SR: A Divergence to Convergence Approach for Real-World Image Super-Resolution".
Python
44
star
49

HomoGAN

This is the official implementation of HomoGAN, CVPR2022
Python
44
star
50

FullMatch

Official implementation of FullMatch (CVPR2023)
Python
44
star
51

KD-MVS

Code for ECCV2022 paper 'KD-MVS: Knowledge Distillation Based Self-supervised Learning for Multi-view Stereo'
Python
44
star
52

AGFlow

Learning Optical Flow with Adaptive Graph Reasoning (AGFlow, AAAI-2022)
Python
42
star
53

pytorch-gym

Implementation of the Deep Deterministic Policy Gradient(DDPG) in bullet Gym using pytorch
Python
41
star
54

TPS-CVPR2023

Python
41
star
55

KPAFlow

PyTorch implementation of KPA-Flow. Learning Optical Flow with Kernel Patch Attention (CVPR-2022)
Python
38
star
56

PCB

Official code for CVPR 2022 paper "Relieving Long-tailed Instance Segmentation via Pairwise Class Balance".
Python
37
star
57

FST-Matching

Official implementation of the FST-Matching Model.
Python
37
star
58

basecls

A codebase & model zoo for pretrained backbone based on MegEngine.
Python
32
star
59

US3L-CVPR2023

PyTorch implementation of US3L (Accepted to CVPR2023)
Python
31
star
60

Sobolev_INRs

[ECCV 2022] The official experimental code of "Sobolev Training for Implicit Neural Representations with Approximated Image Derivatives"
Python
30
star
61

Portraits_Correction

Python
29
star
62

basedet

An object detection codebase based on MegEngine.
Python
28
star
63

Co-mining

Co-mining: Self-Supervised Learning for Sparsely Annotated Object Detection, AAAI 2021.
Python
27
star
64

zipfls

This repo is the official megengine implementation of the ECCV2022 paper: Efficient One Pass Self-distillation with Zipf's Label Smoothing.
Python
25
star
65

tf-cpn

Cascade Pyramid Netwrok
Python
24
star
66

Arch-Net

Arch-Net: Model Distillation for Architecture Agnostic Model Deployment
Python
22
star
67

juicefs-python

JuiceFS Python SDK
Python
21
star
68

ED-Net

PyTorch implementation of A Lightweight Encoder-Decoder Path for Deep Residual Networks.
Python
19
star
69

IntLLaMA

IntLLaMA: A fast and light quantization solution for LLaMA
Python
18
star
70

CasPL

17
star
71

MSCL

[ECCV2022] Motion Sensitive Contrastive Learning for Self-supervised Video Representation
Python
17
star
72

LBHomo

This is the official PyTorch implementation of Semi-supervised Deep Large-baseline Homography Estimation with Progressive Equivalence Constraint, AAAI 2023
Python
17
star
73

RG-SENet_SP-SENet

PyTorch implementation of Delving Deep into Spatial Pooling for Squeeze-and-Excitation Networks.
Python
17
star
74

hpargparse

argparse extension for hpman
Python
16
star
75

Sparse-Beats-Dense

[ECCV 2024] Sparse Beats Dense: Rethinking Supervision in Radar-Camera Depth Completion
Python
15
star
76

MCTrack

This is the offical implementation of the paper "MCTrack: A Unified 3D Multi-Object Tracking Framework for Autonomous Driving"
Python
14
star
77

MEMD

Megvii Electric Moped Detector (ONNX based inference)
Python
13
star
78

DVN

Python
13
star
79

Occ2net

Jupyter Notebook
13
star
80

revisitAIRL

[ECCV2022] Revisiting the Critical Factors of Augmentation-Invariant Representation Learning
Python
11
star
81

megengine-face-recognition

Python
9
star
82

SimpleDG

This is the training and test code for ECCV2022 workshop NICO challenge
Python
7
star
83

GeneGAN

Pytorch version of GeneGAN
Python
7
star
84

basecore

basecore is a simple repo that provides deep learning frame for MegEngine.
Python
7
star
85

hpnevergrad

A nevergrad extension for hpman
Python
5
star
86

DRConv

Python
4
star
87

.github

2
star