Tensorflow 2.14.0 published pypi metadata is missing platform specific dependencies installed via pip

[dre-hh]

Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

binary

TensorFlow version

2.14.0

Custom code

No

OS platform and distribution

macOS-13.3-arm64-arm-64bit

Mobile device

No response

Python version

3.11

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

No response

GPU model and memory

No response

Current behavior?

Tensorflow 2.14.0 still has wrong metadata published to pypi compared to what is actually installed
It does not list following platform specific dependencies.:

Requires-Dist: tensorflow-macos (==2.14.0) ; platform_system == "Darwin" and platform_machine == "arm64"
Requires-Dist: tensorflow-cpu-aws (==2.14.0) ; platform_system == "Linux" and (platform_machine == "arm64" or platform_machine == "aarch64")
Requires-Dist: tensorflow-intel (==2.14.0) ; platform_system == "Windows"

This leads to poetry not being able to install tensorflow properly.
See #61477

Standalone code to reproduce the issue

pip install tensorflow==2.14.0
cd python3.11/site-packages
cat tensorflow-2.14.0.dist-info/METADA

Requires-Dist: tensorflow-macos (==2.14.0) ; platform_system == "Darwin" and platform_machine == "arm64"
Requires-Dist: tensorflow-cpu-aws (==2.14.0) ; platform_system == "Linux" and (platform_machine == "arm64" or platform_machine == "aarch64")
Requires-Dist: tensorflow-intel (==2.14.0) ; platform_system == "Windows"

vs

curl -s https://pypi.org/pypi/tensorflow/2.14.0/json | jq '.info.requires_dist'

[
  "opt-einsum (>=2.3.2)",
  "absl-py (>=1.0.0)",
  "astunparse (>=1.6.0)",
  "flatbuffers (>=23.5.26)",
  "gast (!=0.5.0,!=0.5.1,!=0.5.2,>=0.2.1)",
  "google-pasta (>=0.1.1)",
  "h5py (>=2.9.0)",
  "libclang (>=13.0.0)",
  "ml-dtypes (==0.2.0)",
  "numpy (>=1.23.5)",
  "packaging",
  "protobuf (!=4.21.0,!=4.21.1,!=4.21.2,!=4.21.3,!=4.21.4,!=4.21.5,<5.0.0dev,>=3.20.3)",
  "setuptools",
  "six (>=1.12.0)",
  "termcolor (>=1.1.0)",
  "typing-extensions (>=3.6.6)",
  "wrapt (<1.15,>=1.11.0)",
  "tensorflow-io-gcs-filesystem (>=0.23.1)",
  "grpcio (<2.0,>=1.24.3)",
  "tensorboard (<2.15,>=2.14)",
  "tensorflow-estimator (<2.15,>=2.14.0)",
  "keras (<2.15,>=2.14.0)",
  "nvidia-cuda-runtime-cu11 (==11.8.89) ; extra == 'and-cuda'",
  "nvidia-cublas-cu11 (==11.11.3.6) ; extra == 'and-cuda'",
  "nvidia-cufft-cu11 (==10.9.0.58) ; extra == 'and-cuda'",
  "nvidia-cudnn-cu11 (==8.7.0.84) ; extra == 'and-cuda'",
  "nvidia-curand-cu11 (==10.3.0.86) ; extra == 'and-cuda'",
  "nvidia-cusolver-cu11 (==11.4.1.48) ; extra == 'and-cuda'",
  "nvidia-cusparse-cu11 (==11.7.5.86) ; extra == 'and-cuda'",
  "nvidia-nccl-cu11 (==2.16.5) ; extra == 'and-cuda'",
  "nvidia-cuda-cupti-cu11 (==11.8.87) ; extra == 'and-cuda'",
  "nvidia-cuda-nvcc-cu11 (==11.8.89) ; extra == 'and-cuda'",
  "tensorrt (==8.5.3.1) ; extra == 'and-cuda'"
]

### Relevant log output

_No response_

[sachinprasadhs]

cc: @kulinseth , @learning-to-play

[thorbenk]

When I run

poetry add "tensorflow==2.15.0.post1[and-cuda]"

I still get

Updating dependencies
Resolving dependencies... (0.6s)

Because tensorflow (2.15.0.post1) depends on tensorflow-intel (2.15.0.post1) which doesn't match any versions, tensorflow is forbidden.
So, because tf depends on tensorflow (2.15.0.post1), version solving failed.

I tried on tf-nightly as well, but there I already run into a different (more fundamental?) issue: #62928

[fuchspa]

Hi there, on my Windows 10 machine I get the same error when trying to use the CPU version.

poetry add tensorflow==2.15.0.post1

yields

Because tensorflow (2.15.0.post1) depends on tensorflow-intel (2.15.0.post1) which doesn't match any versions, tensorflow is forbidden.
So, because tf depends on tensorflow (2.15.0.post1), version solving failed.