Nixtla's Open Source Time Series Ecosystem.

Nixtla Community

Received an email this morning with a '500' error. Details in thread.

yes, it's a modal error, but i'm having a hard time debugging it hahaha

yes, <@U06UP1WPN7N>, by default we retry the errors on the sdk up to 5 times

image.png

there is a problem with the format of they ds column

<@U06MW6C1QM7> could we answer the following?

hey! thank you so much for using timegpt. we saw that there is a problem with the datetime column you're providing. it should be formatted in `%Y-%m-d %H-%M-%S` (since we don't support utc-based datestamps yet), could you try formatting in that way? here's a snippet:

```df["ds"] = df["ds"].str.replace("-06:00", "")
X_df["ds"] = X_df["ds"].str.replace("-06:00", "")```
and making the request again. thank you!

i mean, the format is not entirely wrong, since it is `datetime64[ns, UTC-06:00]` , but we don't support utc based datestamps yet, looping in <@U06E82QHSQG> to think in supporting this once he is back

I created an <https://github.com/Nixtla/nixtla/issues/419|issue> on Gh. Lmk if it needs to be changed.

awesome! thank you <@U066SSLDSPL>. also could you open one in <https://github.com/Nixtla/large-time-model-api|nixtla/large-time-model-api> linking it to that issue to keep track in the "back" of the api? thank you!

Thanks so much! <@U05UTMJQ85R> I'll pass your message (and not the screenshot) on to the user.

provisionally a more descriptive error when finding an unsupported datetime format would be helpful too