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

Problem installing requirement.yml #74

Open mariaclaraav opened 3 weeks ago

mariaclaraav commented 3 weeks ago

When I try to install the requirement.yml I get this error:

INFO: pip is looking at multiple versions of jax[cuda12] to determine which version is compatible with other requiremen.

Pip subprocess error: ERROR: Ignored the following versions that require a different python version: 1.6.2 Requires-Python >=3.7,<3.10; 1.6.30 ERROR: Could not find a version that satisfies the requirement jax-cuda12-plugin==0.4.26; extra == "cuda12" (from jax[c) ERROR: No matching distribution found for jax-cuda12-plugin==0.4.26; extra == "cuda12"

My CUDA 12.1 has been setup, so I can't understand the error. I've tried to change for jax[cuda12] but then I get another error:

The conflict is caused by: praxis 1.4.0 depends on lingvo==0.12.7 praxis 1.3.1 depends on lingvo praxis 1.3.0 depends on lingvo praxis 1.2.0 depends on lingvo praxis 1.1.0 depends on lingvo==0.12.7 praxis 1.0.0 depends on jaxlib==0.4.7 praxis 0.4.0 depends on lingvo praxis 0.3.0 depends on lingvo praxis 0.2.1 depends on lingvo praxis 0.2.0 depends on lingvo praxis 0.1 depends on lingvo

ilich917 commented 3 weeks ago

I'm having the same problem when do: pip install -e temp/timesfm, once copied repo

ChengEast commented 2 weeks ago

me toooooooo

ChengEast commented 1 week ago

change the python version to 3.10

magexo commented 16 hours ago

Same here.

Quick start should always work for any open source project, otherwise its difficult to participate...

Please fix so we can contribute as well.

siriuz42 commented 15 hours ago
  1. The lingvo install does not work for ARM architectures, therefore the current installation does not work for Apple Silicon. We will update once we get a solution.

  2. If it's a matter of the versioning of packages let's try pinning down in the yml that python==3.10 paxml==1.4.0 praxis==1.4.0 numpy==1.26.4 pandas==2.1.4 einshape==1.0.0