Closed IbtihalFerwana closed 1 year ago
I got the same pproblem
are you referring to canonical-trn-dev-data.tar that was being downloaded from https://storage.yandexcloud.net/yandex-research/shifts/weather/canonical-trn-dev-data.tar ? It is still missing.
Does anyone have an old copy they can share so we can load it some place?
Hi! I have checked the download link and it appears to work fine for me.
Also, the link has both train and dev-in - so what is supposedly missing?
Unless something has changed since I last gave it a poke in November, the canonical-trn-dev-data.tar is no longer on the https://storage.yandexcloud.net site. If you run the tutorial in the weathers folder, you will see a comment that the data is located at "https://storage.yandexcloud.net/yandex-research/shifts/weather/canonical-trn-dev-data.tar". So I was not able to run the tutorial examples. Other parts of shifts works, but I was hoping to specifically verify the examples before writing a bunch of front end stuff to scrape and ready EPA and other weather data for Shifts. If you got a copy of canonical-trn-dev-data.ta some place, please work with me to get that so I can poke this just a bit.
ps: sorry for the delay in response. I am on holiday travel.
Hi @ebo I understand. The original link was used to provide the data during the running of the challenge. After finishing the challenge, all the canonical data is made available together at: https://storage.yandexcloud.net/yandex-research/shifts/weather/canonical-partitioned-dataset.tar This includes the files train, dev-in and dev-out. Hence, you can just select these 3 files to verify the examples in the tutorial. The tutorial itself has been updated to indicate this. Hope that helps!
Woo Hoo! It is downloading. Thank you for fixing this.
On Jan 11 2023 8:16 AM, Vatsal Raina wrote:
Hi @ebo I understand. The original link was used to provide the data during the running of the challenge. After finishing the challenge, all the canonical data is made available together at: https://storage.yandexcloud.net/yandex-research/shifts/weather/canonical-partitioned-dataset.tar This includes the files train, dev-in and dev-out. Hence, you can just select these 3 files to verify the examples in the tutorial. The tutorial itself has been updated to indicate this. Hope that helps!
As a note, I took a quick moment to check it out. The names of the directory and dataset has changed. I was able to replace "data/" with "canonical-paritioned-dataset/shiftscanonical" in the tutorial for the traing.csv, dev_in.csv, and dev_out.csv. Other than that I was able to get it to work up to the retention curves, where it died with an issue with sklearn (probably a problem on my end with a clean install in a new computer). I'll post a separate issue if there is anything needed to fix that in shifts. Other than that I look forward to testing this out further.
also, you might consider adding "--no-clobber" to the wget command so that you enforce only downloading once in case it gets rerun via a high level jupyter run/restart. If you want me to submit pull requests instead, I can do that as well.
I receive 404 response.