top-data can't be synced, although it's in the `tr...
# timegpt-1point5o
Olivier Sprangers
07/28/2025, 9:45 PMtop-data can't be synced, although it's in the
train_constants.pyh
Han Wang
07/28/2025, 9:48 PMinteresting
Han Wang
07/28/2025, 9:48 PMis it eval or train?
Han Wang
07/28/2025, 9:52 PMtry again?
o
Olivier Sprangers
07/28/2025, 9:54 PMIt's train
Olivier Sprangers
07/28/2025, 9:54 PMAnd yes trying again
Olivier Sprangers
07/28/2025, 10:01 PMNope, same error
h
Han Wang
07/28/2025, 10:01 PMdo you have the path to the data?
Han Wang
07/28/2025, 10:01 PMor the config? let me try
o
Olivier Sprangers
07/28/2025, 10:05 PMSee train confs in this PR: https://github.com/Nixtla/calmecac/pull/59
minutely fails on top-data
daily_short also fails, I think it's due to the (large) synthetic being synced that times out
weekly failed too, but I got rid of almost all tests sets, now it runs. Apparently too large test sets also make it fail
Olivier Sprangers
07/28/2025, 10:06 PMI'm guessing the last two are time-out issues. The first I don't understand, as top-data should be an added volume
Olivier Sprangers
07/28/2025, 10:06 PM(also the last two can be overcome by just removing these larger datasets)
h
Han Wang
07/28/2025, 10:14 PMok we need to remove the test section completely
Han Wang
07/28/2025, 10:23 PMoh i see why
Han Wang
07/28/2025, 10:23 PMit is because we didn't create that bucket on azure
Han Wang
07/28/2025, 10:23 PMlet me fix
Han Wang
07/28/2025, 10:32 PM@Olivier Sprangers it is working now
o
Olivier Sprangers
07/29/2025, 6:56 AMThanks @Han Wang!
Olivier Sprangers
07/29/2025, 7:05 AMWorks now
8 Views