Date: April 01 Company: PCCOER https://www.pccoer.com Contact: KASHMIRE SAMRUDDHI<samruddhi.kashmire_comp21@pccoer.in> I have been using TimeGPT for time series forecasting tasks, and I encountered some challenges when attempting to use timestamps as part of the input data. It seems that TimeGPT does not fully support timestamps, as I encountered errors when trying to make forecasts using timestamp data. Upon reviewing the documentation, I couldn't find any mention of support for timestamps. Furthermore, it appears that TimeGPT may require input data with unique dates. I encountered issues related to duplicate dates when trying to make forecasts with my dataset. Could you please provide clarification on the following: 1. Does TimeGPT support timestamps as input data for time series forecasting tasks? If not, are there any plans to add support for timestamps in future releases? 2. Is TimeGPT designed to work only with input data containing unique dates? If so, what is the rationale behind this requirement? I appreciate any insights or guidance you can provide on these matters. If there are any alternative approaches or workarounds for using timestamps with TimeGPT, I would be grateful to learn about them. Thank you for your attention to this matter. I look forward to hearing from you soon.

@Cristian Challu, pinning you and tagging @Tracy Teal

This is a weird question... Hi XXX, Thanks for using TimeGPT. The

time_col

column should contain the timestamp of each observation. Can you provide more details on the error you are encountering? The most common error is a mismatch between values and frequency of the data. The column has some requirements as mentioned in our documentation (add link to docs.nixtla.io). One requirement is that the

time_col

should have unique values for each unique time series (

id_col

). There can be duplicate dates if they are for different series.