Datasets:

togethercomputer

/

RedPajama-Data-1T

like 1.06k

Follow

Together 434

Thanks! Does it download the files into different folders? (see #9 - there are duplicate file names across CC)

No it doesn't my bad, didn't think about that, it was very late.

should also probably have setup multi-download, this will take a long time. but hey I've got the time.

I'll use this when I restart or something similar, on the assumption that the file names stay in the same order.
Get-Content urls.txt | Foreach-Object {
$originalFilename = Split-Path -Path $_ -Leaf
$filename = $originalFilename
$counter = 1

while (Test-Path $filename) {
    $fileExtension = [System.IO.Path]::GetExtension($originalFilename)
    $fileNameWithoutExtension = [System.IO.Path]::GetFileNameWithoutExtension($originalFilename)
    $filename = $fileNameWithoutExtension + "_" + $counter + $fileExtension
    $counter++
}

Invoke-WebRequest -Uri $_ -OutFile $filename

}

I'll look into hash file comparison later.

Cheers

tom

Thanks! I don't have a Windows machine to test on, so it's very helpful here :)

If it helps, this is a download script for *nix based machines:

wget 'https://data.together.xyz/redpajama-data-1T/v1.0.0/urls.txt'
while read line; do
    dload_loc=${line#https://data.together.xyz/redpajama-data-1T/v1.0.0/}
    mkdir -p $(dirname $dload_loc)
    wget "$line" -O "$dload_loc"
done < urls.txt

the download works fine, the problem is it takes a looooong time, I think I worked out it would take about a week if not longer, my connection is slow. Working on another solution at the moment.

I will try out your download script for the first couple of files in the sets, and let you know.

wget 'https://data.together.xyz/redpajama-data-1T/v1.0.0/urls.txt'
cat urls.txt | xargs -n 1 -P 8 wget --no-cache --no-cookies --timeout=5

for a little more parallel action....but be kind to the server(s) This will put all the files in the same directory as where you're running it. (should work on most *nix)

Thanks Emma, problem is I live in the bush in Darwin, top of Australia, my internet is as slow as a wet weekend especially in the wet season. I just don't have university or Researcher bandwidth. I'll give it a try when I get the Ubuntu server up and running again, for the moment it's the win box that needs to do the work.