[Library Upgrade] uni2ts, momentfm, tsfm_public (Moirai, MOMENT, Granite TTM) #9340
Description
The following libraries are required to use popular Hugging Face time series forecasting models in LEAN algorithms:
- uni2ts — required to use Salesforce/moirai-1.0-R-small/base/large, the Moirai universal time series foundation model
- momentfm — required to use AutonLab/MOMENT-1-large, the MOMENT time series foundation model
- tsfm_public — required to use ibm-granite/granite-timeseries-ttm-r1, IBM's Granite Tiny Time Mixer
These packages cannot be installed at backtest runtime because the QC Cloud backtest environment does not have outbound network access. They need to be pre-installed in the Python environment alongside existing ML libraries like chronos, transformers, and torch.
Checklist
- I have completely filled out this template
- I have confirmed that this issue exists on the current
masterbranch - I have confirmed that this is not a duplicate issue by searching issues