Adding tensorflow-io-gcs-filesystem to the build recipe by ganand1 · Pull Request #206 · conda-forge/tensorflow-feedstock

ganand1 · 2022-02-07T04:24:32Z

Checklist

[ x] Used a personal fork of the feedstock to propose changes
Bumped the build number (if the version is unchanged)
[ x] Reset the build number to 0 (if the version changed)
Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
Ensured the license file is being packaged.

…nda-forge-pinning 2022.02.06.08.21.47

conda-forge-linter · 2022-02-07T04:24:36Z

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe:

It looks like the 'tensorflow-base' output doesn't have any tests.

ngam · 2022-02-07T20:55:39Z

+  - name: tensorflow-io-gcs-filesystem
+    script: build_io_gcs_fs.sh   # [not win]
+    build:
+      string: cuda{{ cuda_compiler_version | replace('.', '') }}py{{ CONDA_PY }}h{{ PKG_HASH }}_{{ PKG_BUILDNUM }}  # [cuda_compiler_version != "None"]
+      string: cpu_py{{ CONDA_PY }}h{{ PKG_HASH }}_{{ PKG_BUILDNUM }}  # [cuda_compiler_version == "None"]
+    requirements:
+      build:
+        - {{ compiler('c') }}
+        - {{ compiler('cxx') }}
+        - {{ compiler('cuda') }}                 # [cuda_compiler_version != "None"]
+        - bazel
+        - bazel >=4.2.1  # [osx and arm64]
+      host:
+        - python
+        - pip
+        - setuptools
+        - wheel
+        - openssl
+        - {{ pin_subpackage('tensorflow-base', exact=True) }}
+      run:
+        - python
+        - {{ pin_subpackage('tensorflow-base', exact=True) }}
+    test:
+      requires:
+        - pip
+      commands:
+        - pip check
+


Suggested change

- name: tensorflow-io-gcs-filesystem

script: build_io_gcs_fs.sh # [not win]

build:

string: cuda{{ cuda_compiler_version | replace('.', '') }}py{{ CONDA_PY }}h{{ PKG_HASH }}_{{ PKG_BUILDNUM }} # [cuda_compiler_version != "None"]

string: cpu_py{{ CONDA_PY }}h{{ PKG_HASH }}_{{ PKG_BUILDNUM }} # [cuda_compiler_version == "None"]

requirements:

build:

- {{ compiler('c') }}

- {{ compiler('cxx') }}

- {{ compiler('cuda') }} # [cuda_compiler_version != "None"]

- bazel

- bazel >=4.2.1 # [osx and arm64]

host:

- python

- pip

- setuptools

- wheel

- openssl

- {{ pin_subpackage('tensorflow-base', exact=True) }}

run:

- python

- {{ pin_subpackage('tensorflow-base', exact=True) }}

test:

requires:

- pip

commands:

- pip check

- name: tensorflow-io

script: build_io_gcs_fs.sh # [not win]

build:

string: cuda{{ cuda_compiler_version | replace('.', '') }}py{{ CONDA_PY }}h{{ PKG_HASH }}_{{ PKG_BUILDNUM }} # [cuda_compiler_version != "None"]

string: cpu_py{{ CONDA_PY }}h{{ PKG_HASH }}_{{ PKG_BUILDNUM }} # [cuda_compiler_version == "None"]

requirements:

build:

- {{ compiler('c') }}

- {{ compiler('cxx') }}

- {{ compiler('cuda') }} # [cuda_compiler_version != "None"]

- bazel

- bazel >=4.2.1 # [osx and arm64]

host:

- python

- pip

- setuptools

- wheel

- openssl

- {{ pin_subpackage('tensorflow', exact=True) }}

run:

- python

- {{ pin_subpackage('tensorflow', exact=True) }}

test:

requires:

- pip

commands:

- pip check

Hi @ganand1, looks you're very close. Could you try building tensorflow-io instead of just the gcs-filesystem? tensorflow-io includes tensorflow-io-gcs-filesystem. Also, I would pin against tensorflow, not the base here. This may help, I can give it a shot locally if you want me to (it will compile relatively faster). Let me know if you have any thoughts or observations so far :) thanks for picking this up! We could maybe implement it in a 2.7.1 release or a rebuild of 2.8.0 once everyone is satisfied

ngam · 2022-02-07T20:56:58Z

    folder: tensorflow-estimator
+  - url: https://github.com/tensorflow/io/archive/refs/tags/v{{ io_gcs_version }}.tar.gz
+    sha256: 7a4b57f6f438402bab4919c360e931c32f4d8d8afa23116d32a48461dddd91dc
+    folder: tensorflow-io-gcs-filesystem


Suggested change

folder: tensorflow-io-gcs-filesystem

folder: tensorflow-io

ngam · 2022-02-07T20:57:27Z

+
+set -exuo pipefail
+
+pushd tensorflow-io-gcs-filesystem


Suggested change

pushd tensorflow-io-gcs-filesystem

pushd tensorflow-io

ngam · 2022-02-07T20:57:48Z

+
+./configure.sh
+
+bazel build //tensorflow_io_gcs_filesystem/...


Suggested change

bazel build //tensorflow_io_gcs_filesystem/...

bazel build //tensorflow_io/...

ngam · 2022-02-07T21:02:27Z

        - python
        - {{ pin_subpackage('tensorflow-base', exact=True) }}
        - {{ pin_subpackage('tensorflow-estimator', exact=True) }}
+        - {{ pin_subpackage('tensorflow-io-gcs-filesystem', exact=True) }}


Suggested change

- {{ pin_subpackage('tensorflow-io-gcs-filesystem', exact=True) }}

- {{ pin_subpackage('tensorflow-io', exact=True) }}

We can probably do io only here or we will want to output it tensorflow-io and tensorflow-io-gcs-filesystem separately

ngam · 2022-02-07T21:03:59Z

@ganand1 please change the title to indicate you're adding io-gcs-filesystem :)

ganand1 · 2022-02-07T22:25:34Z

@ngam - i think we could generate only tensorflow-io-gcs-filesystem. i followed this readme
https://github.com/tensorflow/io/blob/master/docs/development.md'

specifically python setup.py bdist_wheel --data bazel-bin --project tensorflow-io-gcs-filesystem
-- builds the exact wheel we need. i see some versioning differences. have fixed for the same. Lets see if the fix helps with versioning tensorflow-io-gcs-filesystem to 0.24.

ngam · 2022-02-07T22:33:48Z

Okay, but please at least have the io package depend on tensorflow-base AND estimator (or simply tensorflow). Otherwise, it will fail again. Have you seen the errors from previous runs? They were, "couldn't find tensorflow-estimator" and couldn't find "tensorflow-io-gcs-filesystem needed for tensorflow-io"

ganand1 · 2022-02-07T22:51:27Z

Okay, but please at least have the io package depend on tensorflow-base AND estimator (or simply tensorflow). Otherwise, it will fail again. Have you seen the errors from previous runs? They were, "couldn't find tensorflow-estimator" and couldn't find "tensorflow-io-gcs-filesystem needed for tensorflow-io"

ok changing to tensorflow.

ngam · 2022-02-13T05:09:07Z

So... @xhochy knows this best than anyone around here I think. He earlier said this:

Yes, that is a very old vendored setuptools version that is Python 3.10 incompatible.

I am looking at your bash script, and I am thinking, maybe you should follow what xhochy did for estimator? Could you try that please?

tensorflow-feedstock/recipe/build_estimator.sh

Lines 1 to 13 in bbd7cdf

    
           #!/bin/bash 
        
           set -exuo pipefail 
        
           pushd tensorflow-estimator 
        
           WHEEL_DIR=${PWD}/wheel_dir 
        
           mkdir -p ${WHEEL_DIR} 
        
           bazel build tensorflow_estimator/tools/pip_package:build_pip_package 
        
           bazel-bin/tensorflow_estimator/tools/pip_package/build_pip_package ${WHEEL_DIR} 
        
           ${PYTHON} -m pip install --no-deps ${WHEEL_DIR}/*.whl 
        
           bazel clean 
        
           popd

In particular, I am not sure if you're altering or overriding the custom_toolchain we have here in your script (e.g. --expunge and ./configure.sh).

tensorflow-feedstock/recipe/build_io_gcs_fs.sh

Lines 1 to 16 in bbd7cdf

    
           #!/bin/bash 
        
           set -exuo pipefail 
        
           pushd tensorflow-io 
        
           bazel clean --expunge 
        
           bazel shutdown 
        
           ./configure.sh 
        
           bazel build //tensorflow_io_gcs_filesystem/... 
        
           python setup.py bdist_wheel --data bazel-bin --project tensorflow-io-gcs-filesystem 
        
           ${PYTHON} -m pip install --no-deps dist/*.whl 
        
           bazel clean 
        
           popd

Or maybe the problem is something simpler that I cannot see... I am not so sure. It's surprising that it's working for python <3.10

ganand1 · 2022-02-13T05:22:38Z

So... @xhochy knows this best than anyone around here I think. He earlier said this:

Yes, that is a very old vendored setuptools version that is Python 3.10 incompatible.

I am looking at your bash script, and I am thinking, maybe you should follow what xhochy did for estimator? Could you try that please?

tensorflow-feedstock/recipe/build_estimator.sh

Lines 1 to 13 in bbd7cdf

#!/bin/bash

set -exuo pipefail

pushd tensorflow-estimator

WHEEL_DIR=${PWD}/wheel_dir

mkdir -p ${WHEEL_DIR}

bazel build tensorflow_estimator/tools/pip_package:build_pip_package

bazel-bin/tensorflow_estimator/tools/pip_package/build_pip_package ${WHEEL_DIR}

${PYTHON} -m pip install --no-deps ${WHEEL_DIR}/*.whl

bazel clean

popd

In particular, I am not sure if you're altering or overriding the custom_toolchain we have here in your script (e.g. --expunge and ./configure.sh).

tensorflow-feedstock/recipe/build_io_gcs_fs.sh

Lines 1 to 16 in bbd7cdf

#!/bin/bash

set -exuo pipefail

pushd tensorflow-io

bazel clean --expunge

bazel shutdown

./configure.sh

bazel build //tensorflow_io_gcs_filesystem/...

python setup.py bdist_wheel --data bazel-bin --project tensorflow-io-gcs-filesystem

${PYTHON} -m pip install --no-deps dist/*.whl

bazel clean

popd

Or maybe the problem is something simpler that I cannot see... I am not so sure. It's surprising that it's working for python <3.10

you mean following tensorflow-estimator and build using build_pip_package like commands?

ngam · 2022-02-13T05:40:37Z

you mean following tensorflow-estimator and build using build_pip_package like commands?

yes, or more generally, trying to see how the tensorflow-estimator building was translated from upstream to here and then you could mimic it. Maybe that will help, but maybe not. That's the only thing that comes to mind...

ganand1 · 2022-02-13T05:49:21Z

tensorflow-io doesnt have that option..
but errors comes from here
https://github.com/tensorflow/io/blob/b9c150b39c62c8a8b93b346f527d9b2e3591cf5f/WORKSPACE#L99
this is too old...
i am thinking does this support python 3.10?

ngam · 2022-02-13T06:04:32Z

tensorflow-io doesnt have that option..
but errors comes from here
https://github.com/tensorflow/io/blob/b9c150b39c62c8a8b93b346f527d9b2e3591cf5f/WORKSPACE#L99
this is too old...
i am thinking does this support python 3.10?

I like your patch :)

ngam · 2022-02-13T06:06:13Z

fingers crossed 🤞

ganand1 · 2022-02-13T06:13:07Z

tensorflow-io doesnt have that option.. but errors comes from here https://github.com/tensorflow/io/blob/b9c150b39c62c8a8b93b346f527d9b2e3591cf5f/WORKSPACE#L99 this is too old... i am thinking does this support python 3.10?

i think it should be fine to remove the lint dependencies as well if this doesnt help. As i read from the build documents, lint has been added for checking lint errors. we could remove it i guess. if this doesnt work will give that a try

xhochy · 2022-02-13T08:06:51Z

tensorflow-io doesnt have that option..
but errors comes from here
https://github.com/tensorflow/io/blob/b9c150b39c62c8a8b93b346f527d9b2e3591cf5f/WORKSPACE#L99
this is too old...
i am thinking does this support python 3.10?

I like your patch :)

Yes, sensible approach 👍

ganand1 · 2022-02-14T06:15:47Z

tensorflow-io doesnt have that option..
but errors comes from here
https://github.com/tensorflow/io/blob/b9c150b39c62c8a8b93b346f527d9b2e3591cf5f/WORKSPACE#L99
this is too old...
i am thinking does this support python 3.10?

I like your patch :)

Yes, sensible approach 👍

have 3.10 running as well. please take a look

ganand1 · 2022-02-14T06:22:37Z

@conda-forge-admin, please restart ci
trying again to see if python3.9 also could pass by running on a different machine

ganand1 · 2022-02-14T06:23:52Z

passing log for python3.10 from the previous run
https://dev.azure.com/conda-forge/84710dde-1620-425b-80d0-4cf5baca359d/_apis/build/builds/458732/logs/111

ngam · 2022-02-14T06:47:38Z

passing log for python3.10 from the previous run
https://dev.azure.com/conda-forge/84710dde-1620-425b-80d0-4cf5baca359d/_apis/build/builds/458732/logs/111

Very cool!! Was it just that patch or was there something else needed?

ngam · 2022-02-14T06:49:34Z

BTW, I am happy to build the osx ones here too. Just give me a signal once you think we are good to go, and then I can find a time to build them soon.

ganand1 · 2022-02-14T07:19:59Z

BTW, I am happy to build the osx ones here too. Just give me a signal once you think we are good to go, and then I can find a time to build them soon.

thank you @ngam . please go ahead .. it was the patch - old version incompatibility. I just retriggered to get python3.9 passing which was failing for. unknown reason

please try OSX at your convenience. thanks for the extended help.

ngam · 2022-05-13T20:46:01Z

@ganand1 still interested in this?

Co-authored-by: ganand1

ganand1 · 2022-05-14T02:17:51Z

@ganand1 still interested in this?

it will be good to merge it so we could have tensorflow-io-gcs-filesystem released from conda-forge.

ngam · 2022-05-16T22:32:33Z

@ganand1 still interested in this?

it will be good to merge it so we could have tensorflow-io-gcs-filesystem released from conda-forge.

It requires some decent amount of work though. I built essentially lightly modified version of this, but it fails in practice. Why? Because the package of interest is actually tensorflow-io, which depends on tensorflow-io-gcs-filesystem. But when I tried to build them both together, they refused to cooperate well enough. So some work is needed. So I suggest we move this to a staged-recipes feedstock. Let me look into this in approx two weeks, and I will tag you in the PR over at conda-forge/staged-recipes. I personally want to see this added to conda-forge asap as well :)

ngam added 6 commits February 6, 2022 14:15

bump 2.8.0; remove patches

8395a27

MNT: Re-rendered with conda-build 3.21.8, conda-smithy 3.16.2, and co…

7f295b5

…nda-forge-pinning 2022.02.06.08.21.47

force grpc_cpp to 1.42

52a503b

MNT: Re-rendered with conda-build 3.21.8, conda-smithy 3.16.2, and co…

2ac43a0

…nda-forge-pinning 2022.02.06.08.21.47

Update conda_build_config.yaml

664d955

pin numpy to 1.20/1.21 and rerender

feada08

ganand1 requested review from FarhanTejani, ghego, h-vetinari, hajapy, hmaarrfk, jschueller, njzjz, waitingkuo, wolfv and xhochy as code owners February 7, 2022 04:24

ganand1 force-pushed the update_2.8.0 branch from 62ff6d2 to 9b97f67 Compare February 7, 2022 15:46

fixing patches, etc.

b688924

ngam reviewed Feb 7, 2022

View reviewed changes

ganand1 force-pushed the update_2.8.0 branch 2 times, most recently from 3f85ad0 to dfe3553 Compare February 7, 2022 22:23

ganand1 force-pushed the update_2.8.0 branch from bbd7cdf to 0cf1f4d Compare February 13, 2022 05:59

ganand1 force-pushed the update_2.8.0 branch 2 times, most recently from dbdb72a to ac68305 Compare February 13, 2022 22:50

Adding tensorflow-io-gcs-filesystem.

f15a521

ganand1 force-pushed the update_2.8.0 branch from ac68305 to f15a521 Compare February 13, 2022 23:53

conda-forge-linter closed this Feb 14, 2022

conda-forge-linter reopened this Feb 14, 2022

ngam mentioned this pull request May 13, 2022

tracking "easy" updates and improvements (for patch release 2.8.1?) #234

Closed

3 tasks

ngam added a commit to ngam/tensorflow-feedstock that referenced this pull request May 13, 2022

import PR conda-forge#206

e322ce1

Co-authored-by: ganand1

ngam mentioned this pull request May 14, 2022

v2.8.1 + add cpu_count, xla #235

Closed

5 tasks

ngam mentioned this pull request Aug 8, 2022

checklist for 2.10.x (+ request for comment, feature requests, etc.) #263

Closed

5 tasks

ngam mentioned this pull request Oct 9, 2022

Compatibility with Tensorflow Text #276

Open

ehfd mentioned this pull request Jan 29, 2025

Package request: tensorflow-text conda-forge/staged-recipes#19578

Open

2 tasks


		./configure.sh

		bazel build //tensorflow_io_gcs_filesystem/...

	bazel build //tensorflow_io_gcs_filesystem/...
	bazel build //tensorflow_io/...

	- {{ pin_subpackage('tensorflow-io-gcs-filesystem', exact=True) }}
	- {{ pin_subpackage('tensorflow-io', exact=True) }}

Uh oh!

Conversation

ganand1 commented Feb 7, 2022

Uh oh!

conda-forge-linter commented Feb 7, 2022

Uh oh!

ngam Feb 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ngam Feb 7, 2022

Choose a reason for hiding this comment

Uh oh!

ngam Feb 7, 2022

Choose a reason for hiding this comment

Uh oh!

ngam Feb 7, 2022

Choose a reason for hiding this comment

Uh oh!

ngam Feb 7, 2022

Choose a reason for hiding this comment

Uh oh!

ngam commented Feb 7, 2022

Uh oh!

ganand1 commented Feb 7, 2022

Uh oh!

ngam commented Feb 7, 2022

Uh oh!

ganand1 commented Feb 7, 2022

Uh oh!

ngam commented Feb 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ganand1 commented Feb 13, 2022

Uh oh!

ngam commented Feb 13, 2022

Uh oh!

ganand1 commented Feb 13, 2022

Uh oh!

ngam commented Feb 13, 2022

Uh oh!

ngam commented Feb 13, 2022

Uh oh!

ganand1 commented Feb 13, 2022

Uh oh!

xhochy commented Feb 13, 2022

Uh oh!

ganand1 commented Feb 14, 2022

Uh oh!

ganand1 commented Feb 14, 2022

Uh oh!

ganand1 commented Feb 14, 2022

Uh oh!

ngam commented Feb 14, 2022

Uh oh!

ngam commented Feb 14, 2022

Uh oh!

ganand1 commented Feb 14, 2022

Uh oh!

ngam commented May 13, 2022

Uh oh!

ganand1 commented May 14, 2022

Uh oh!

ngam commented May 16, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

ngam Feb 7, 2022 •

edited

Loading

ngam commented Feb 13, 2022 •

edited

Loading