Value Error: Units 'M', 'Y' and 'y'

[humbertodlltrr]

I have monthly data that I am generating forecasts for, I have only found issues with model.plot_validations() function so far. It seems this end in an eventual Timedelta(r, unit) which is no longer supported for anything higher than days D. Is there a workaround at the moment that I haven't yet percieved?

[winedarksea]

I just ran a test on monthly data and plot_validations worked fine with pandas==2.1.3. Could you provide more details about your error?

[humbertodlltrr]

Im using the following:
python 3.10.0
pandas==2.1.3
autots==0.6.3

The call in particular is in evaluator\auto_ts.py, line 3142 where it evaluates my models frequency, M, into "1M"

I'm thinking its up to a dependency that causes in this virtual environment to take pd.to_timedelta(used_freq), where used_freq is 1M, to the specified error. On my personal machine it seems to work fine but the same code is evaluated to Timedelta('0 days 00:01:00') which would still be an unwanted output.

[winedarksea]

I've got it now, thanks for the details: ValueError: Units 'M', 'Y' and 'y' do not represent unambiguous timedelta values and are not supported.

I'll try to put together a fix for the next release

[winedarksea]

I have pushed a fix to dev.
For now, if you need an immediate fix. Try setting start_date="2023-01-01", or some other date, in plot_validations, which will bypass the bug