Skip to content

Latest commit

 

History

History
88 lines (75 loc) · 3.46 KB

README.md

File metadata and controls

88 lines (75 loc) · 3.46 KB
Raw

made-with-python


Logo
FastColabCopy

Python3 script to transfer files in Google Colab 10-50x faster.

About The Project • How To Use • Examples • Best Practice • Credits • More Examples

screenshot

About The Project

FastColabCopy is a Python script for parallel (multi-threading) copying of files between two locations. Currently developed for Google-Drive to Google-Drive transfers using Google-Colab. This script frequently achieves 10-50x speed improvements when copying numerous small files.

Importing

Import from GitHub:

!wget https://raw.githubusercontent.com/L0garithmic/fastcolabcopy/main/fastcopy.py
import fastcopy

Import from Google Drive:

!cp /gdrive/MyDrive/fastcopy.py .
import fastcopy

Usage

usage: fast-copy.py [-h HELP] source destination [-d DELETE] [-s SYNC] [-r REPLACE]

optional arguments:
  -h --help            show this help message and exit
  source                the drive you are copying from
  destination           the drive you are copying to
  -d --delete           delete the source files after copy
  -s --sync             delete files in destination if not found in source (do not use, if using with rsync)
  -r --replace          replace files if they exist
  -t --thread           set the amount of parallel threads used
  -l --size-limit       set max size of files copied (supports gb, mb, kb) eg 1.5gb

The source and destination fields are required. Everything else is optional.

Examples

from google.colab import drive
drive.mount('/gdrive', force_remount=False)
import os
!wget -q https://raw.githubusercontent.com/L0garithmic/fastcolabcopy/main/fastcopy.py
import fastcopy
!python fastcopy.py /gdrive/Shareddrives/Source/. /gdrive/Shareddrives/Destination --thread 20 --size-limit 400mb

If you want to see copy execution time:

!pip install -q ipython-autotime
%load_ext autotime

Check out examples.md for some more examples.

Best Practice

Colab has wildly varying transfer speeds, because of this, the best we can offer are suggestions:

  • For large groups of medium/small files, 15-40 threads seems to work best.
  • For 50+ files with significantly varying sizes, try 2 sequentially copies. -t 15 -l 400 then -t 2
  • For files that are 100MB+, it is best to use 2 threads. It is still faster then rsync.
  • Currently --sync breaks if rsync is ran after. If you are mirroring drives. Disable --sync and use the rsync's --delete function.

Credits

  • Credit to ikonikon for the base multi-threading code.
  • Thanks to @Ostokhoon for ALL argument and folder hierarchy functionality.