Why versions after 2.11 are not available on Windows?
AsakusaRinne opened this issue · 2 comments
Hi there,
I noticed that 2.10 is the last version on windows. I know that tensorflow has dropped the cuda support for native-windows since 2.11. However CPU version on Windows still works. May I ask the reason why tensorflow_text not even support tensorflow CPU version on Windows since 2.11?
Apologies for the late reply, I've actually switched projects and did not notice this question was assigned to me. This is actually addressed in the README here:
https://github.com/tensorflow/text#a-note-about-different-operating-system-packages
tl;dr: the team doesn't have the bandwidth to maintain packages for many different platforms other than Linux. When TF switched to their collaborator model for Windows builds, we saw that there were very few downloads for Windows and the little interest didn't support the maintenance cost.
Thank you a lot for your response. :) May I ask another question here? In fact there's some reasons for me to use tensorflow.text in windows (WSL2 is not enough). As I can see, tensorflow completely dropped the GPU support for native windows, and tensorflow.text left the support for windows cpu and gpu to community. Though I could stay in v2.10 to use gpu. However as time pass, there will be many inconvenience and I can't enjoy the new features.
Do you think I should switch to Jax to make NLP apps on native windows? Actually I'm a little scared. Because tensorflow has been open-sourced for 8 years and now it has suddenly dropped the gpu support, leaving some of my efforts on it meaningless. Will NLP components of Jax still drop windows in the future? I don't know where I can go if they all give up supporting windows....