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
3.77k stars 320 forks source link

TensorFlow update #144

Open Juanosorio91 opened 1 month ago

Juanosorio91 commented 1 month ago

In my company we are having an issue running TimesFm due to a TensorFlow vulnerability that affects the versions of tensorflow, keras and flatbuffers. We are blocked because me need to run on TensorFlow 2.12.1 or higher. Do you know if in a next release is planned to update TensorFlow? I leave the link to the detailed issue. https://github.com/advisories/GHSA-gjh7-xx4r-x345 Thanks!

rajatsen91 commented 1 month ago

Just pushed the pytorch version. Would that work better with tf dependency? IT can also run on python >= 3.11

Juanosorio91 commented 1 month ago

Hi I checked the last Pull Request, and it stills runs in TensorFlow 2.9.3

rajatsen91 commented 1 month ago

We have not pushed to pypi yet. Did you follow this path ?