How to adapt the software version 2.10.0 to the linux-aarhc64-aarhc64

[ZerryNi]

Comment:

No response

[h-vetinari]

2.10 on aarch will not happen. @dslarm has been leading a heroic effort to get linux-aarch64 builds supported in #438, but this will only land for 2.18 or later, certainly not 2.10.

[ZerryNi]

2.10 on aarch will not happen. @dslarm has been leading a heroic effort to get linux-aarch64 builds supported in #438, but this will only land for 2.18 or later, certainly not 2.10.

Hi,however,I need to resolve this conflict issue.Could you please advise on how to do so ?

[dslarm]

Ideally.. in order of preference

1. Get the target package to migrate to keras 3

2. Try compatibility method for old keras identified here (see backwards compatibility section with TF_USE_LEGACY_KERAS=1 etc.). You may need to create a new package tf_keras (I can't check right now)
   : https://keras.io/getting_started/#:~:text=TensorFlow%20compatibility,0%20%26%20keras~%3D2.14.0

3. Figure out how to get an older version of tensorflow into conda-forge.. It might go ok, if you specifically do not try to do cuda. Don't know what maintainers think of that idea.