search

google-research / timesfm

TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.
https://research.google/blog/a-decoder-only-foundation-model-for-time-series-forecasting/
Apache License 2.0
3.02k stars 228 forks source link

Random Result #57

Open hvlgo opened 1 month ago

hvlgo commented 1 month ago

I would like to ask why the results are different each time even though I have fixed the random seed.

rajatsen91 commented 1 month ago

This should not happen because in eval model dropout is turned off: https://github.com/google-research/timesfm/blob/master/src/timesfm.py#L307

Are you seeing randomness on GPU or CPU and if so what is the magnitude of randomness relative to the prediced value?