Sincere Appreciation!!!

#1
by Michael34234 - opened

Thank you so much for your great efforts of bringing this dataset to Huggingface, it really means so much for our work!!!
Could you please upload the following training splits onto this repo?
Once again, great gratitude for your efforts!

Hi @Michael34234 , I am still figuring out how HF works.
Can you please tell me whether you can download the TRAIN_0 split with the following line of code:

pip install huggingface_hub["cli"]

from huggingface_hub import snapshot_download

# Download TRAIN_0 split
snapshot_download(repo_id="SilvioGiancola/TrackingNet", repo_type="dataset", revision="main", allow_patterns="*TRAIN_0/*", local_dir="TrackingNet_HF")

I am uploading the other splits, but it will take a few days to commit, push and upload 1.2TB of data.

Delighted to hear from you this soon!
The answer is a definite yes! And the downloading process works perfectly fine!

image.png
Waiting for your precious data~~ Thanks~ Love U~

BTW, my code is :
snapshot_download(repo_id='SilvioGiancola/TrackingNet',
repo_type='dataset',
local_dir='/home/liyou/opensource_datasets/TrackingNet',
resume_download=True,
local_dir_use_symlinks=False)

Maybe directly uploading the .zip file will be faster? if the floder consists of too many small files

Great, thanks for sharing your line of code, I like that option: local_dir_use_symlinks=False better.
I considered the zip, but HF has a limit of 50GB per file and the zip versions of the split are ~90GB each.

Actually, I would not advise to use local_dir_use_symlinks=False as it will overwrite the file every time you run the line of code, creating an overhead on huggingface servers. Please prefer that line of code:

snapshot_download(repo_id="SilvioGiancola/TrackingNet", 
                  repo_type="dataset", revision="main",
                  local_dir="TrackingNet_HF", 
                  allow_patterns="*TRAIN_0/*")

Actually, I would not advise to use local_dir_use_symlinks=False as it will overwrite the file every time you run the line of code, creating an overhead on huggingface servers. Please prefer that line of code:

snapshot_download(repo_id="SilvioGiancola/TrackingNet", 
                  repo_type="dataset", revision="main",
                  local_dir="TrackingNet_HF", 
                  allow_patterns="*TRAIN_0/*")

Got it! It really helps~

SilvioGiancola changed discussion status to closed

Sign up or log in to comment