Yes, we support specifying packages to be installed at runtime, but that
takes time. Also, tensorflow (attempts to) use semantic versioning, so a
user should be requiring tensorflow >=1.x,<2.0.

On Fri, Aug 31, 2018 at 12:36 PM Maximilian Michels <
[email protected]> wrote:

> *@mxm* approved this pull request.
> ------------------------------
>
> In sdks/python/container/Dockerfile
> <https://github.com/apache/beam/pull/6314#discussion_r214312578>:
>
> >      pip install "pandas == 0.18.1" && \
>      pip install "scipy == 1.0.0" && \
>      pip install "protobuf == 3.3.0" && \
> -    pip install "tensorflow == 1.4.0" && \
> +    pip install "tensorflow == 1.9" && \
>
> Do we support upgrading the provided packages? What if user code requires
> tensorflow 1.4?
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> <https://github.com/apache/beam/pull/6314#pullrequestreview-151342493>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/AAdqgRg2Dov1c8CuBbHLaz1KdcMlTHP-ks5uWRGlgaJpZM4WU-wP>
> .
>


[ Full content available at: https://github.com/apache/beam/pull/6314 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to