[BUG] Tensorflow install issue in some environments

[AjayThorve]

Tensorflow installs in an environment with rapids libraries fail in some environments, like the rapidsai/ci containers. This is likely due to a compatibility issue between tensorflow and protobuf, which causes cudf - tensorflow compatibility issue.

conda create -n test -c rapidsai-nightly -c conda-forge -c nvidia cudf=23.12 tensorflow

results in

Looking for: ['cudf=23.12', 'tensorflow', 'cuda-version=11.8']

conda-forge/linux-64                                        Using cache
conda-forge/noarch                                          Using cache
nvidia/linux-64                                             Using cache
nvidia/noarch                                               Using cache
rapidsai-nightly/linux-64                                     No change
rapidsai-nightly/noarch                                       No change
Could not solve for environment specs
The following packages are incompatible
├─ cuda-version 11.8**  is requested and can be installed;
├─ cudf 23.12**  is installable with the potential options
│  ├─ cudf [23.12.00a710|23.12.00a713] would require
│  │  ├─ protobuf >=4.24,<5.0a0 , which requires
│  │  │  └─ libabseil >=20230802.1,<20230803.0a0 , which requires
│  │  │     └─ abseil-cpp 20230802.1* , which can be installed;
│  │  └─ python_abi 3.9.* *_cp39, which can be installed;
│  ├─ cudf [23.12.00a710|23.12.00a713] would require
│  │  ├─ protobuf >=4.24,<5.0a0 , which can be installed (as previously explained);
│  │  └─ python_abi 3.10.* *_cp310, which can be installed;
│  └─ cudf [23.12.00a710|23.12.00a713] would require
│     └─ cuda-version >=12,<13.0a0 , which conflicts with any installable versions previously reported;
└─ tensorflow   is uninstallable because there are no viable options
   ├─ tensorflow [0.10.0|0.11.0|...|1.9.0] would require
   │  └─ python_abi * *_cp27mu, which conflicts with any installable versions previously reported;
   ├─ tensorflow [0.10.0|0.11.0|...|0.9.0] would require
   │  └─ python_abi * *_cp34m, which conflicts with any installable versions previously reported;
   ├─ tensorflow [0.10.0|0.11.0|...|1.9.0] would require
   │  └─ python_abi * *_cp35m, which conflicts with any installable versions previously reported;
   ├─ tensorflow [0.12.1|1.0.0|...|1.9.0] would require
   │  └─ python_abi * *_cp36m, which conflicts with any installable versions previously reported;
   ├─ tensorflow [1.13.1|1.13.2] would require
   │  ├─ tensorflow-base [1.13.1 py27h76b4ce7_8|1.13.1 py36h58012e3_6|...|1.13.2 py27h76b4ce7_0], which requires
   │  │  └─ tensorflow-estimator >=1.13.0,<1.14.0a0 , which does not exist (perhaps a missing channel);
   │  └─ tensorflow-estimator >=1.13.0,<1.14.0a0 , which does not exist (perhaps a missing channel);
   ├─ tensorflow 1.13.1 would require
   │  └─ tensorflow-estimator >=1.13.0,<1.14.0rc0 , which does not exist (perhaps a missing channel);
   ├─ tensorflow 1.14.0 would require
   │  └─ python_abi 3.7.* *_cp37m, which conflicts with any installable versions previously reported;
   ├─ tensorflow 1.14.0 would require
   │  └─ python_abi 3.6.* *_cp36m, which conflicts with any installable versions previously reported;
   ├─ tensorflow [2.10.0|2.11.0|2.11.1] would require
   │  └─ tensorflow-base [2.10.0 cpu_py310hc537a0e_0|2.10.0 cpu_py37h50bd216_0|...|2.11.1 cuda112py39h0320e7d_0], which requires
   │     └─ libabseil 20220623.0 cxx17*, which conflicts with any installable versions previously reported;
   ├─ tensorflow 2.4.0 would require
   │  └─ tensorflow-base [2.4.0 py36h515a7b5_0|2.4.0 py37h00a14e9_0|2.4.0 py38h01d9eeb_0], which requires
   │     └─ grpc-cpp >=1.36.4,<1.37.0a0 , which requires
   │        └─ abseil-cpp >=20200923.3,<20200923.4.0a0 , which conflicts with any installable versions previously reported;
   ├─ tensorflow 2.4.1 would require
   │  └─ tensorflow-base [2.4.1 py36h312d151_0|2.4.1 py37he2fe834_0|2.4.1 py38h83f5f1d_0|2.4.1 py39h23a8cbf_0], which requires
   │     └─ grpc-cpp >=1.37.0,<1.38.0a0  but there are no viable options
   │        ├─ grpc-cpp [1.36.4|1.37.0], which cannot be installed (as previously explained);
   │        ├─ grpc-cpp [1.37.0|1.37.1] would require
   │        │  └─ abseil-cpp >=20210324.0,<20210324.1.0a0 , which conflicts with any installable versions previously reported;
   │        ├─ grpc-cpp 1.37.1 would require
   │        │  └─ abseil-cpp >=20210324.1,<20210324.2.0a0 , which conflicts with any installable versions previously reported;
   │        └─ grpc-cpp [1.37.1|1.39.1|1.40.0|1.41.1|1.42.0] would require
   │           └─ abseil-cpp >=20210324.2,<20210324.3.0a0 , which conflicts with any installable versions previously reported;
   ├─ tensorflow [2.4.3|2.6.0] would require
   │  └─ tensorflow-base [2.4.3 py36h312d151_0|2.4.3 py37he2fe834_0|...|2.6.0 py39h23a8cbf_0], which requires
   │     └─ grpc-cpp >=1.37.1,<1.38.0a0 , which cannot be installed (as previously explained);
   ├─ tensorflow [2.6.0|2.6.2|...|2.8.1] would require
   │  └─ tensorflow-base [2.6.0 cpu_py37hc5ef7b8_2|2.6.0 cpu_py38h4611ba2_2|...|2.8.1 cuda112py39he716a45_0], which requires
   │     ├─ abseil-cpp >=20210324.2,<20210324.3.0a0 , which conflicts with any installable versions previously reported;
   │     └─ grpc-cpp [>=1.39.1,<1.40.0a0 |>=1.40.0,<1.41.0a0 |>=1.41.1,<1.42.0a0 |>=1.42.0,<1.43.0a0 ], which cannot be installed (as previously explained);
   ├─ tensorflow 2.9.1 would require
   │  └─ tensorflow-base [2.9.1 cpu_py310h8df3ab6_0|2.9.1 cpu_py37hb97876d_0|...|2.9.1 cuda112py39h81abfd3_0], which requires
   │     └─ abseil-cpp >=20211102.0,<20211102.1.0a0 , which conflicts with any installable versions previously reported;
   └─ tensorflow [2.12.1|2.13.1|2.14.0] would require
      └─ tensorflow-base [2.12.1 cpu_py310h57e0714_0|2.12.1 cpu_py310h91b7bb3_1|...|2.14.0 cuda118py39hbe86951_0], which requires
         └─ libabseil >=20230125.3,<20230126.0a0 , which conflicts with any installable versions previously reported.