Setup Github Actions workflow CI for horovod on CPU
See original GitHub issue🚀 Feature
Idea is to add run Horovod tests on CPU via Github Actions. We can either: 1) install horovod using pip or 2) otherwise compile horovod from source as it is done for Circle CI (without NCCL support). The goal is to be able to run CPU distributed tests with Horovod backend.
TODO:
- create new file
hvd-tests.yml
in.github/workflows
like https://github.com/pytorch/ignite/blob/master/.github/workflows/tpu-tests.yml - setup horovod (described above)
- run horovod tests like in circle ci yaml
Quick play around github actions can be done in a separate repository like https://github.com/vfdev-5/github-actions-playground
Issue Analytics
- State:
- Created 3 years ago
- Comments:7 (4 by maintainers)
Top Results From Across the Web
Add GitHub Workflow CI · horovod/horovod@0974e3b · GitHub
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. - Add GitHub Workflow CI · horovod/horovod@0974e3b.
Read more >horovod/ci.yaml at master - GitHub
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. - horovod/ci.yaml at master · horovod/horovod.
Read more >horovod/Dockerfile.test.cpu at master - GitHub
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. - horovod/Dockerfile.test.cpu at master · horovod/horovod.
Read more >horovod/install.rst at master - GitHub
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. - horovod/install.rst at master · horovod/horovod.
Read more >intel/intel-horovod: Distributed training framework for ... - GitHub
To run on CPUs: $ pip install horovod. To run on GPUs with NCCL: $ HOROVOD_GPU_OPERATIONS=NCCL pip install horovod. For more details on...
Read more >Top Related Medium Post
No results found
Top Related StackOverflow Question
No results found
Troubleshoot Live Code
Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start FreeTop Related Reddit Thread
No results found
Top Related Hackernoon Post
No results found
Top Related Tweet
No results found
Top Related Dev.to Post
No results found
Top Related Hashnode Post
No results found
Top GitHub Comments
hey @sdesrozis , I am working on this. will give a PR in few days. 🙂
@ramesht007 in circle ci we use docker env to easily have pytorch distribution without installing conda etc as it is done in github actions. For github actions, let’s follow the same setup as https://github.com/pytorch/ignite/blob/master/.github/workflows/unittests.yml#L29 , Setup Miniconda and Install dependencies steps. Let’s use python 3.7 and the latest pytorch version.