pip dependencies screw up pins

[k-dominik]

currently creating an environment will result in...

...
numpy                     1.19.5                   pypi_0    pypi
...

whereas the current pin states:

tiktorch/environment.yml

Line 23 in 7dbf87f

- numpy=1.18.5

there are also a couple more packages that pip will uninstall. I assume there is no way one could get tensorflow from conda-forge? Maybe once conda-forge/tensorflow-feedstock#110 is merged, it is an option...

[k-dominik]

since it's still bugging me some more detail:

Installing collected packages: rsa, pyasn1-modules, oauthlib, cachetools, requests-oauthlib, google-auth, werkzeug, tensorboard-plugin-wit, numpy, markdown, grpcio, google-auth-oauthlib, absl-py, wrapt, termcolor, tensorflow-estimator, tensorboard, opt-einsum, keras-preprocessing, h5py, google-pasta, gast, flatbuffers, astunparse, tensorflow, pytest-grpc, onnxruntime
  Attempting uninstall: numpy
    Found existing installation: numpy 1.18.5
    Uninstalling numpy-1.18.5:
      Successfully uninstalled numpy-1.18.5
  Attempting uninstall: grpcio
    Found existing installation: grpcio 1.33.2
    Uninstalling grpcio-1.33.2:
      Successfully uninstalled grpcio-1.33.2
  Attempting uninstall: h5py
    Found existing installation: h5py 3.1.0
    Uninstalling h5py-3.1.0:
      Successfully uninstalled h5py-3.1.0
Successfully installed absl-py-0.12.0 astunparse-1.6.3 cachetools-4.2.1 flatbuffers-1.12 gast-0.3.3 google-auth-1.28.0 google-auth-oauthlib-0.4.3 google-pasta-0.2.0 grpcio-1.32.0 h5py-2.10.0 keras-preprocessing-1.1.2 markdown-3.3.4 numpy-1.19.5 oauthlib-3.1.0 onnxruntime-1.7.0 opt-einsum-3.3.0 pyasn1-modules-0.2.8 pytest-grpc-0.7.0 requests-oauthlib-1.3.0 rsa-4.7.2 tensorboard-2.4.1 tensorboard-plugin-wit-1.8.0 tensorflow-2.4.1 tensorflow-estimator-2.4.0 termcolor-1.1.0 werkzeug-1.0.1 wrapt-1.12.1

[njzjz]

Hi, conda-forge/tensorflow-feedstock#110 has been merged.

[k-dominik]

Hey @njzjz, thanks a lot for the heads-up!!

note: TF2 still missing on win64

[k-dominik]

just a little update: while tensorflow is now happily on conda-forge, we're now missing onnxruntime. There is one discarded PR to bring it to conda forge. It's been a year now, so I think the reported problems there might be gone.