Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

cleaner: bump 2.8.0; rebase patches; new numpy pins #205

Merged
merged 37 commits into from
May 18, 2022

Conversation

ngam
Copy link
Contributor

@ngam ngam commented Feb 6, 2022

cc @hmaarrfk

Closes #202
Closes #203
Closes #204
Closes #212

Checklist

  • Used a personal fork of the feedstock to propose changes
  • Bumped the build number (if the version is unchanged)
  • 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.

@conda-forge-linter
Copy link

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 ngam mentioned this pull request Feb 6, 2022
12 tasks
README.md Outdated Show resolved Hide resolved
@ngam
Copy link
Contributor Author

ngam commented Feb 6, 2022

Okay, everything broke for some reason! I will guess it is the 7f295b5 updating grpc_cpp to '1.43' since this is the only change from the other PR. grpc_cpp 1.43 wrecks mayhem basically: downgrading bazel, messing with cuda stuff, etc., so reverting to grpc_cpp 1.42 for now.

@ngam
Copy link
Contributor Author

ngam commented Feb 6, 2022

Is this looking good for you, @hmaarrfk? Should we basically lock this in for production?

We can assess the tensorflow-io thing later. Easiest path is pip install wheel but this one has a nasty matrix, cp37--310 for different arches, and no osx-arm64. (per #203 (comment) this is not feasible)

@hmaarrfk
Copy link
Contributor

hmaarrfk commented Feb 6, 2022

We really should fix

conda-forge/bazel-feedstock#90

@conda-forge-linter
Copy link

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

I wanted to let you know that I linted all conda-recipes in your PR (recipe) and found some lint.

Here's what I've got...

For recipe:

  • Old-style Python selectors (py27, py35, etc) are only available for Python 2.7, 3.4, 3.5, and 3.6. Please use explicit comparisons with the integer py, e.g. # [py==37] or # [py>=37]. See lines [85, 86, 160, 161]

For recipe:

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

@conda-forge-linter
Copy link

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 ngam changed the title cleaner --- bump 2.8.0; remove patches cleaner --- bump 2.8.0; remove patches; new numpy pins Feb 7, 2022
@ngam
Copy link
Contributor Author

ngam commented Apr 10, 2022

@conda-forge-admin, please rerender

@ngam
Copy link
Contributor Author

ngam commented Apr 10, 2022

@xhochy @hmaarrfk any idea why this has regressed significantly? Could you take a quick look and see if the errors look familiar?

@hmaarrfk
Copy link
Contributor

You might want to update the pinnings and take a few patches from:
#218

They seem to be included in the 2.9.0 milestone.

@hmaarrfk
Copy link
Contributor

I got stuck at the fact that grpc updated abseil/absl and that is causing compilation errors with CUDA builds.

We might want to do all the migrations except for GRPC

@ngam
Copy link
Contributor Author

ngam commented May 12, 2022

@conda-forge-admin, please rerender

@ngam
Copy link
Contributor Author

ngam commented May 12, 2022

@hmaarrfk @xhochy looks like you two did some magic in that last PR! Thanks a lot. This seems fine for now, so we could probably push forward. With the new singularity framework, we will hopefully have an easier time building going forward :)

Let me know if you know of any issues that I should be aware of, looks smooth to me so far, granted only ~10 mins...

@xhochy
Copy link
Member

xhochy commented May 12, 2022

I will give this a spin locally and can build out all packages (osx+linux) if it passes.

@ngam
Copy link
Contributor Author

ngam commented May 12, 2022

These are gonna be slightly longer builds than 2.7.x, looking at 23000 vs 17000 (compilation stuff). I should have one build ready soon-ish if it doesn't fail... Will post top-line results here

@ngam
Copy link
Contributor Author

ngam commented May 12, 2022

Okay, I think we are ready for this! @hmaarrfk @xhochy 🥳

Would love a quick review and then we can get this out of the way!

cuda 11.2 py38:

INFO: Analyzed 3 targets (441 packages loaded, 31257 targets configured).
INFO: Found 3 targets...
[0 / 40] [Prepa] Expanding template tensorflow/lite/tools/visualize
[599 / 5,648] Compiling tensorflow/python/util/determinism.cc; 3s local ... (128 actions, 65 running)
[849 / 5,880] Compiling llvm/utils/TableGen/CodeGenDAGPatterns.cpp; 6s local ... (128 actions, 65 running)
[5,329 / 10,076] Compiling llvm/utils/TableGen/GlobalISelEmitter.cpp; 9s local ... (128 actions, 65 running)
[5,957 / 14,512] Compiling third_party/icu/data/conversion_data.c; 7s local ... (128 actions, 65 running)
[6,558 / 15,731] Compiling third_party/icu/data/conversion_data.c; 14s local ... (128 actions, 65 running)
[7,649 / 18,934] Compiling tensorflow/lite/kernels/conv.cc; 8s local ... (128 actions, 65 running)
[8,082 / 20,176] Compiling tensorflow/lite/kernels/conv.cc; 18s local ... (128 actions, 65 running)
[8,438 / 21,142] Compiling src/cpu/cpu_convolution_list.cpp; 18s local ... (128 actions, 65 running)
[9,568 / 23,290] Compiling src/cpu/cpu_rnn_list.cpp; 29s local ... (128 actions, 65 running)
[10,163 / 23,290] Compiling src/cpu/rnn/ref_rnn.cpp; 37s local ... (128 actions, 65 running)
[10,444 / 23,290] Compiling llvm/lib/Transforms/Utils/SimplifyCFG.cpp; 9s local ... (128 actions, 64 running)
[10,746 / 23,290] Compiling llvm/lib/Passes/PassBuilder.cpp; 14s local ... (128 actions, 64 running)
[11,024 / 23,290] Compiling llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp; 22s local ... (128 actions, 64 running)
[11,699 / 23,290] Compiling llvm/lib/Target/X86/X86ISelLowering.cpp; 32s local ... (128 actions, 64 running)
[12,157 / 23,290] Compiling llvm/lib/Transforms/IPO/AttributorAttributes.cpp; 14s local ... (127 actions, 64 running)
[12,606 / 23,290] Compiling llvm/lib/Target/X86/X86ISelLowering.cpp; 16s local ... (128 actions, 65 running)
[14,614 / 23,290] Compiling mlir/lib/Dialect/LLVMIR/IR/LLVMDialect.cpp; 14s local ... (128 actions, 64 running)
[15,673 / 23,290] Compiling mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp; 27s local ... (128 actions, 65 running)
[16,636 / 23,290] Compiling tensorflow/compiler/mlir/hlo/lib/Dialect/mhlo/IR/hlo_ops.cc; 46s local ... (128 actions, 65 running)
[17,590 / 23,290] Compiling tensorflow/compiler/mlir/tensorflow/ir/tf_ops_a_m.cc; 63s local ... (128 actions, 65 running)
[18,162 / 23,290] Compiling tensorflow/core/kernels/rnn/lstm_ops_gpu.cu.cc; 111s local ... (128 actions, 65 running)
[18,987 / 23,290] Compiling tensorflow/core/kernels/sparse_fill_empty_rows_op_gpu.cu.cc; 187s local ... (128 actions, 65 running)
[19,067 / 23,290] Compiling tensorflow/core/kernels/training_ops_gpu.cu.cc; 270s local ... (128 actions, 65 running)
[19,336 / 23,290] Compiling tensorflow/core/kernels/cwise_op_gpu_mul.cu.cc; 209s local ... (128 actions, 65 running)
[19,482 / 23,290] Compiling tensorflow/core/kernels/cwise_op_gpu_random_grad.cu.cc; 328s local ... (128 actions, 65 running)
[20,720 / 23,291] Compiling tensorflow/core/kernels/cwise_op_gpu_random_grad.cu.cc; 467s local ... (128 actions, 65 running)
[21,193 / 23,291] Compiling tensorflow/core/kernels/cwise_op_gpu_igammas_float.cu.cc; 624s local ... (128 actions, 64 running)
[23,265 / 23,335] Compiling tensorflow/core/kernels/segment_reduction_ops_gpu_1.cu.cc; 671s local ... (7 actions running)
[23,273 / 23,335] Compiling tensorflow/core/kernels/pad_op_gpu.cu.cc; 567s local ... (2 actions running)
INFO: Elapsed time: 1838.318s, Critical Path: 947.64s
INFO: 23335 processes: 5923 internal, 17412 local.
INFO: Build completed successfully, 23335 total actions
INFO: Build completed successfully, 23335 total actions
...
####################################################################################
Resource usage summary:

Total time: 0:51:50.6
CPU usage: sys=0:00:30.3, user=0:01:11.8
Maximum memory usage observed: 1.2G
Total disk usage observed (not including envs): 2.2G
Validating outputs
+ validate_recipe_outputs tensorflow-feedstock
validation results:
{
  "linux-64/libtensorflow-2.8.0-cuda112h98e1010_0.tar.bz2": true,
  "linux-64/libtensorflow_cc-2.8.0-cuda112h98e1010_0.tar.bz2": true,
  "linux-64/tensorflow-base-2.8.0-cuda112py38h6a3b174_0.tar.bz2": true,
  "linux-64/tensorflow-estimator-2.8.0-cuda112py38hf5dcc89_0.tar.bz2": true,
  "linux-64/tensorflow-2.8.0-cuda112py38hded6998_0.tar.bz2": true,
  "linux-64/tensorflow-gpu-2.8.0-cuda112py38h0bbbad9_0.tar.bz2": true
}

@xhochy
Copy link
Member

xhochy commented May 13, 2022

The OSX builds don't pass. I know what to do and will tackle that later.

+ test -f /Users/uwe/mambaforge/conda-bld/tensorflow-split_1652379672357/_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_/lib/libtensorflow.dylib
+ ./test_libtensorflow.sh
+ arm64-apple-darwin20.0.0-clang -o test_c -L/Users/uwe/mambaforge/conda-bld/tensorflow-split_1652379672357/_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_/lib/ -ltensorflow -I/Users/uwe/mambaforge/conda-bld/tensorflow-split_1652379672357/_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_/include/ test_c.c
+ ./test_c
dyld[2693]: Library not loaded: @rpath/libtensorflow.2.dylib
  Referenced from: /Users/uwe/mambaforge/conda-bld/tensorflow-split_1652379672357/test_tmp/test_c
  Reason: tried: '/usr/local/lib/libtensorflow.2.dylib' (no such file), '/usr/lib/libtensorflow.2.dylib' (no such file)
./test_libtensorflow.sh: line 6:  2693 Abort trap: 6           ./test_c
Tests failed for libtensorflow-2.8.0-cpu_hb019c61_0.tar.bz2 - moving package to /Users/uwe/mambaforge/conda-bld/broken
WARNING:conda_build.build:Tests failed for libtensorflow-2.8.0-cpu_hb019c61_0.tar.bz2 - moving package to /Users/uwe/mambaforge/conda-bld/broken
TESTS FAILED: libtensorflow-2.8.0-cpu_hb019c61_0.tar.bz2

@ngam
Copy link
Contributor Author

ngam commented May 13, 2022

Looking good, I am happy to split the builds with you, so let me know. I am on standby for today and tomorrow (Saturday). After that, I will be mostly absent for ~10 days. I also have more stuff for you 😜 (cuda-full jaxlib + a preliminary PR for 2.8.1)

@xhochy
Copy link
Member

xhochy commented May 14, 2022

Enjoy being absent. There are still some small issues. I can iron them out and build things in the next days. Currently the osx-arm64 fail locally because MACOSX_SDK_VERSION is not set when running the local builds. Once all OSX builds have passed, I'll do the Linux builds.

@ngam
Copy link
Contributor Author

ngam commented May 15, 2022

Sounds good! Always here to help :) thanks a lot, you're a star 🌟

@xhochy xhochy mentioned this pull request May 16, 2022
5 tasks
@xhochy
Copy link
Member

xhochy commented May 17, 2022

@hmaarrfk Builds are ready in https://anaconda.org/uwe.korn/repo for osx-* and https://anaconda.org/uwe.korn-tf-gpu/repo for linux-* Can you have a look and approve? I would then copy over (or feel free to do so yourself)

Logs:

@xhochy
Copy link
Member

xhochy commented May 17, 2022

OSX is still uploading …

@xhochy
Copy link
Member

xhochy commented May 17, 2022

@hmaarrfk all uploaded, so ready for review/copy.

@hmaarrfk
Copy link
Contributor

Thanks!

for package in  tensorflow-estimator tensorflow-cpu tensorflow-base tensorflow  libtensorflow_cc libtensorflow; do
anaconda copy --from-label main --to-label main --to-owner conda-forge uwe.korn/${package}/2.8.0
done
for package in  tensorflow-estimator tensorflow-cpu tensorflow-gpu tensorflow-base tensorflow  libtensorflow_cc libtensorflow; do
anaconda copy --from-label main --to-label main --to-owner conda-forge uwe.korn-tf-gpu/${package}/2.8.0
done

@hmaarrfk hmaarrfk merged commit 51db7f8 into conda-forge:main May 18, 2022
@h-vetinari
Copy link
Member

Thanks everyone for sticking through this! 🥳

@ngam ngam deleted the v2.8.0 branch June 2, 2022 20:50
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants