mxnet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hen <bay...@apache.org>
Subject Re: [GitHub] szha closed pull request #11154: Revert "[MXNET-503] Website landing page for MMS (#11037)"
Date Mon, 11 Jun 2018 02:21:55 GMT
It wasn't clear why this was commit was reverted. Things that stood out as
odd:

* I didn't see an email to dev@ on the topic of a revert.
* Rather than reverting, if there is a minor item requiring a fix it could
simply be fixed; if a major item then it should be raised on dev@.
* I didn't see a reason to revert in the revert PR (11154).
* The original PR has github:szha asking for github:piiswrong to review
with no context; I'm concerned that it was implied that the commit could
not go in without this review.
* I don't see anything in the original PR to earn a revert. At best
'github:john-andrilla' being asked if "a flexible, scalable,
multi-framework serving solution" was okay.
* I find it odd that github:lupesko is a reviewer.

Hen



On Tue, Jun 5, 2018 at 5:08 PM, GitBox <git@apache.org> wrote:

> szha closed pull request #11154: Revert "[MXNET-503] Website landing page
> for MMS (#11037)"
> URL: https://github.com/apache/incubator-mxnet/pull/11154
>
>
>
>
> This is a PR merged from a forked repository.
> As GitHub hides the original diff on merge, it is displayed below for
> the sake of provenance:
>
> As this is a foreign pull request (from a fork), the diff is supplied
> below (as it won't show otherwise due to GitHub magic):
>
> diff --git a/docs/mms/index.md b/docs/mms/index.md
> deleted file mode 100644
> index ff6edae414b..00000000000
> --- a/docs/mms/index.md
> +++ /dev/null
> @@ -1,114 +0,0 @@
> -# Model Server for Apache MXNet (incubating)
> -
> -[Model Server for Apache MXNet (incubating)](https://github.
> com/awslabs/mxnet-model-server), otherwise known as MXNet Model Server
> (MMS), is an open source project aimed at providing a simple yet scalable
> solution for model inference. It is a set of command line tools for
> packaging model archives and serving them. The tools are written in Python,
> and have been extended to support containers for easy deployment and
> scaling. MMS also supports basic logging and advanced metrics with Amazon
> CloudWatch integration.
> -
> -
> -## Multi-Framework Model Support with ONNX
> -
> -MMS supports both *symbolic* MXNet and *imperative* Gluon models. While
> the name implies that MMS is just for MXNet, it is in fact much more
> flexible, as it can support models in the [ONNX](https://onnx.ai) format.
> This means that models created and trained in PyTorch, Caffe2, or other
> ONNX-supporting frameworks can be served with MMS.
> -
> -To find out more about MXNet's support for ONNX models and using ONNX
> with MMS, refer to the following resources:
> -
> -* [MXNet-ONNX Docs](../api/python/contrib/onnx.md)
> -* [Export an ONNX Model to Serve with MMS](https://github.com/
> awslabs/mxnet-model-server/docs/export_from_onnx.md)
> -
> -## Getting Started
> -
> -To install MMS with ONNX support, make sure you have Python installed,
> then for Ubuntu run:
> -
> -```bash
> -sudo apt-get install protobuf-compiler libprotoc-dev
> -pip install mxnet-model-server
> -```
> -
> -Or for Mac run:
> -
> -```bash
> -conda install -c conda-forge protobuf
> -pip install mxnet-model-server
> -```
> -
> -
> -## Serving a Model
> -
> -To serve a model you must first create or download a model archive. Visit
> the [model zoo](https://github.com/awslabs/mxnet-model-server/
> docs/model_zoo.md) to browse the models. MMS options can be explored as
> follows:
> -
> -```bash
> -mxnet-model-server --help
> -```
> -
> -Here is an easy example for serving an object classification model. You
> can use any URI and the model will be downloaded first, then served from
> that location:
> -
> -```bash
> -mxnet-model-server \
> -  --models squeezenet=https://s3.amazonaws.com/model-server/
> models/squeezenet_v1.1/squeezenet_v1.1.model
> -```
> -
> -
> -### Test Inference on a Model
> -
> -Assuming you have run the previous `mxnet-model-server` command to start
> serving the object classification model, you can now upload an image to its
> `predict` REST API endpoint. The following will download a picture of a
> kitten, then upload it to the prediction endpoint.
> -
> -```bash
> -curl -O https://s3.amazonaws.com/model-server/inputs/kitten.jpg
> -curl -X POST http://127.0.0.1:8080/squeezenet/predict -F
> "data=@kitten.jpg"
> -```
> -
> -The predict endpoint will return a prediction response in JSON. It will
> look something like the following result:
> -
> -```
> -{
> -  "prediction": [
> -    [
> -      {
> -        "class": "n02124075 Egyptian cat",
> -        "probability": 0.9408261179924011
> -      },
> -...
> -```
> -
> -For more examples of serving models visit the following resources:
> -
> -* [Quickstart: Model Serving](https://github.com/
> awslabs/mxnet-model-server/README.md#serve-a-model)
> -* [Running the Model Server](https://github.com/
> awslabs/mxnet-model-server/docs/server.md)
> -
> -
> -## Create a Model Archive
> -
> -Creating a model archive involves rounding up the required model
> artifacts, then using the `mxnet-model-export` command line interface. The
> process for creating archives is likely to evolve. As the project adds
> features, we recommend that you review the following resources to get the
> latest instructions:
> -
> -* [Quickstart: Export a Model](https://github.com/
> awslabs/mxnet-model-server/README.md#export-a-model)
> -* [Model Artifacts](https://github.com/awslabs/mxnet-model-server/
> docs/export_model_file_tour.md)
> -* [Loading and Serving Gluon Models](https://github.com/
> awslabs/mxnet-model-server/tree/master/examples/gluon_alexnet)
> -* [Creating a MMS Model Archive from an ONNX Model](https://github.com/
> awslabs/mxnet-model-server/docs/export_from_onnx.md)
> -* [Create an ONNX model (that will run with MMS) from PyTorch](
> https://github.com/onnx/onnx-mxnet/blob/master/README.md#quick-start)
> -
> -
> -## Using Containers
> -
> -Using Docker or other container services with MMS is a great way to scale
> your inference applications. You can use Docker to pull the latest version:
> -
> -```
> -docker pull awsdeeplearningteam/mms_gpu
> -```
> -
> -It is recommended that you review the following resources for more
> information:
> -
> -* [MMS Docker Hub](https://hub.docker.com/u/awsdeeplearningteam/)
> -* [Using MMS with Docker Quickstart](https://github.
> com/awslabs/mxnet-model-server/docker/README.md)
> -* [MMS on Fargate](https://github.com/awslabs/mxnet-model-server/
> docs/mms_on_fargate.md)
> -* [Optimized Container Configurations for MMS](https://github.com/
> awslabs/mxnet-model-server/docs/optimized_config.md)
> -* [Orchestrating, monitoring, and scaling with MMS, Amazon Elastic
> Container Service, AWS Fargate, and Amazon CloudWatch)](https://aws.
> amazon.com/blogs/machine-learning/apache-mxnet-model-
> server-adds-optimized-container-images-for-model-serving-at-scale/)
> -
> -
> -## Community & Contributions
> -
> -The MMS project is open to contributions from the community. If you like
> the idea of a flexible, scalable, multi-framework serving solution for your
> models and would like to provide feedback, suggest features, or even jump
> in and contribute code or examples, please visit the [project page on
> GitHub](https://github.com/awslabs/mxnet-model-server). You can create an
> issue there, or join the discussion on the forum.
> -
> -* [MXNet Forum - MMS Discussions](https://discuss.
> mxnet.io/c/mxnet-model-server)
> -
> -
> -## Further Reading
> -
> -* [GitHub](https://github.com/awslabs/mxnet-model-server)
> -* [MMS Docs](https://github.com/awslabs/mxnet-model-server/docs)
>
>
>
>
> ----------------------------------------------------------------
> This is an automated message from the Apache Git Service.
> To respond to the message, please log on GitHub and use the
> URL above to go to the specific comment.
>
> For queries about this service, please contact Infrastructure at:
> users@infra.apache.org
>
>
> With regards,
> Apache Git Services
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message