Open mathpluscode opened 10 months ago
Hi @mathpluscode ,
The requirements updated for master branch which is having ml-dtypes~= 0.3.1
now.
This is an issue with Tf2.15v with tensorflow package also apart from tensorflow-macos. In setup.py ml-dtypes~= 0.2.0
but for all 3 requirements_lock_3_x.txt
it's still showing ml-dtypes~= 0.3.1
. Need to check on this. Thanks!
CC: @nitins17 , @learning-to-play , For your review please.
Any updates on a resolution to this issue? cc @nitins17 @learning-to-play
We are having the same problem on Linux. This issue is especially problematic as it affects Jax and related packages which demand a newer ml-dypes version.
If I'm understanding correctly, this appears to already be fixed on master
:
'ml_dtypes ~= 0.2.0',
'ml_dtypes ~= 0.3.1',
So it seems like the operative questions are:
Thanks for letting us know! #63020 updates the ml-dtypes version to 0.3.1. It should be included in the upcoming patch release for TF 2.15.
When will 2.15.1 be released?
I do not know the exact date but believe it should be out by end of March. Adding @learning-to-play to confirm.
@jakeBass TF 2.15 patch release is planned to be released before the end of March.
@nitins17 Thank you for cherry picking the fix to the r2.15 branch.
Issue type
Bug
Have you reproduced the bug with TensorFlow Nightly?
No
Source
binary
TensorFlow version
2.15.0
Custom code
No
OS platform and distribution
MacOS 14.1.2
Mobile device
Mac
Python version
3.9
Bazel version
No response
GCC/compiler version
No response
CUDA/cuDNN version
No response
GPU model and memory
No response
Current behavior?
Currently, we could not install the following packages together
From the requirement, it seems that it should require
ml-dtypes==0.3.1
.But if running
We will get
"ml-dtypes (~=0.2.0)",
.Note, it does work without
orbax-checkpoint
which is now recommended by Flax for checkpointing.Standalone code to reproduce the issue
Relevant log output