Open thewh1teagle opened 3 months ago
I understand the issue, but I think it shoud be simlper than wget (not all users have wget installed). Also, how can the model be fetched automatically upon pip install
?
I understand the issue, but I think it shoud be simlper than wget
wget is installed by default in most systems. Windows (powershell), Linux and macOS. The thing is that git does not fetch efficiently large files, while wget does.
Also, how can the model be fetched automatically upon
pip install
I don't recommend doing that too since it's error prune. most ML libraries today simply instruct to fetch the model before executing the script.
The imporant part is to make sure the user will find it easy to fetch to the correct path, even if they are not very well-versed in Python. If you know an idiomatic way to do that, one that won't be a hurdle to the user (e.g. lazily fetching upon first execution), then I would really appreciate a PR. Otherwise I will try to implement a better solution when I'll get to it.
Simple wget
from PowerShell does not create the file.
Simple
wget
from PowerShell does not create the file.
Turns out that in Windows
the -O
flag is required and then the same command works in macOS/Linux/Windows.
wget https://github.com/elazarg/nakdimon/blob/master/nakdimon/Nakdimon.h5 -O Nakdimon.h5
In addition wget
in Windows can be installed with
winget install -e --id JernejSimoncic.Wget
It's faster and more reliable than PowerShell's wget handler. Use wget.exe
to use it from PowerShell then.
It takes a lot of time to do
git clone
with this repository since the model is uploaded here as well. I suggest remove it and upload it to the release https://github.com/elazarg/nakdimon/releases/tag/v0.1.2Then you can simply instruct the users to execute
wget
with the model path before running.