Updating runners README files.

1. Adding README file for the docker runner.
2. Removing old content. Adding one-sentence description and a reference to the documentation pages folder.

runners/mlcommons_box_docker_run/README.md

@@ -0,0 +1,4 @@
+# MLCommons-Box Docker Runner
+MLCommons-Box Docker Runner runs boxes (packaged Machine Learning (ML) workloads) in the docker environment. Read 
+Docker Runner documentation [here](../../docs/runners/docker-runner.md).
+

runners/mlcommons_box_singularity_run/README.md

@@ -1,65 +1,4 @@
-# Singularity Runner
-Singularity runner uses Singularity runtime for running MBoxes. It supports two commands - `configure` and `build`:
-- `configure`: build Singularity container on a machine where it runs. 
-- `run`: run MLBox in a Singularity container build at configure phase.
-
-It follows the [MLBox Runners Specification v0.1](https://docs.google.com/document/d/1bL8bsAam71Ex8GI_6mQ59QlVf8RkVW-GvsbPweAd8C8)
-to implement the command line interface:
-```shell script
-python -m mlbox_singularity_run configure --mlbox=MLBOX_PATH --platform=PLATFORM_FILE_PATH --task=TASK_FILE_PATH
-```
-where:
-- `MLBOX_PATH` is the path to the MLBox root directory.
-- `PLATFORM_FILE_PATH` is the path to the Singularity platform definition file. This file is usually located in the
-  `platform` sub-directory of the MLBox.
-- `TASK_FILE_PATH` is the path to the task run file. This file is usually located in the `run` sub-directory of the
-  MLBox.
-
-### Singularity Platform Definition File
-The schema definition file is the part of the Singularity runner and is located in the source directory of the project:
-[mlbox-singularity.yaml](mlbox_singularity_run/mlbox-singularity.yaml). Onle one parameter is defined in the schema - 
-`image`. It is path to singularity image. It is relative to MLBOX_ROOT/workspace:
-- By default, containers are stored in `$MLBOX_ROOT/workspace` if image is a file name.
-- If it is a relative path, it is relative to `$MLBOX_ROOT/workspace`.
-- Absolute paths (starting with `/`) are used as is.
-
-One drawback of using workspace directory is that when using SSH runner, this directory is synchronized between remote
-and local hosts what results in transferring the Singularity image back to user's local machine which may not be
-desirable behavior. 
-
-If path to image does not exist, singularity runner will attempt to create one. 
-
-### Configure
-Configure command line interface requires two mandatory arguments - `mlbox` and `platform`:
-```shell script
-python -m mlbox_singularity_run configure --mlbox=MLBOX_PATH --platform=PLATFORM_FILE_PATH
-```
-
-### End to end example using MNIST MLBOX and GitHub source tree
-- Setup and activate python virtual environment, install runner requirements
-  ```shell script
-  virtualenv -p python3.8 ./env
-  source ./env/bin/activate
-  pip install typer mlspeclib
-  export PYTHONPATH=$(pwd)/mlcommons_box:$(pwd)/runners/mlbox_singularity_run
-  ```
-
-- Specify path to the Singularity image. In MNIST MLBox (`platform/singularity.yaml`), the path is
-  `/opt/singularity/mlperf_mlbox_mnist-0.01.simg`. Either change it, or make sure the `/opt/singularity` the directory
-  exists and writable, or users have permission to create one.
-
-- Setup environment. This is optional step. 
-  ```shell script
-  export https_proxy=${http_proxy}
-  ```
-
-- Configure MNIST MLBox.
-  ```shell script
-  python -m mlbox_singularity_run configure --mlbox=examples/mnist --platform=examples/mnist/platform/singularity.yaml
-  ```
-
-- Run two tasks - `download` (download data) and `train`
-  ```shell script
-  python -m mlbox_singularity_run run --mlbox=examples/mnist --platform=examples/mnist/platform/singularity.yaml --task=examples/mnist/run/download.yaml
-  python -m mlbox_singularity_run run --mlbox=examples/mnist --platform=examples/mnist/platform/singularity.yaml --task=examples/mnist/run/train.yaml
-  ```
+# MLCommons-Box Singularity Runner
+MLCommons-Box Singularity Runner runs boxes (packaged Machine Learning (ML) workloads) in the singularity environment.
+Read Singularity Runner documentation [here](../../docs/runners/singularity-runner.md).
+

runners/mlcommons_box_ssh_run/README.md

@@ -1,56 +1,3 @@
-# MLBox SSH runner
-
-This is the port of the SHH runner from this [repository](https://github.com/sergey-serebryakov/mlbox/tree/feature/mlbox_runners_v2)
-
-__Porting Status__  
-- Support for new MLBox directory structure.
-- Still runs only docker MLBoxes. Configure phase works, the run phase does not because of different directory tree.
-
-## Pre-requisites
-1. Remote server available via SSH/rsync.
-2. Password-less login with public/private keys. If it's not the case, the ssh runner will be asking for a password 
-   several times (quite annoying).
-3. Remote server must provide python interpreter with ssh runner requirements (`mlspeclib`, `typer`). It can be either
-   a system python or user python (virtualenv, conda etc.). Full path to the python executable must be known.
-4. Python version should be at least 3.5 (or maybe 3.6 - type annotations are used).
-
-## Current limitations (may not be the complete list)
-1. Only docker-based MLBoxes are supported. So, remote host must provide docker/nvidia-docker.
-2. Task files have the following path `run/{task_name}.yaml` (relative to MLBox root path).
-3. MLBox contains the `build` directory with Dockerfile. Images are built, not pulled.
-4. The whole `workspace` directory is synced with the local host after each task execution.
-5. The SHH runner was tested with MNIST MLBox only.
-
-> Pretty much all of these can easily be solved. In fact, the original repo do not have the first three limitations.
-
-## How to use SSH runner using MNIST example MLBox
-1. Copy the `ssh.yaml` platform file in the `platform` directory of a MLBox. 
-   `platform.yaml`.
-2. Edit `platform.yaml`. Change the following fields: 
-   - `host`: IP address of the remote host
-   - `user`: User name to use
-   - `env->interpreter->python`: Python interpreter to use on a remote host to run SSH runner. If it's a custom
-     installation, specify the absolute path.
-   - `env->variables`: Dictionary of environmental variables to use. Will be used for docker build/run. In some cases,
-     _http_proxy_ and _https_proxy_ variables need to be set.
-3. Configure remote host (copy mlbox runners, mlbox, build docker image) - run in the mlbox root directory:
-   ```shell script
-   export PYTHONPATH=$(pwd)/mlcommons_box:$(pwd)/runners/mlbox_singularity_run:$(pwd)/runners/mlbox_ssh_run
-
-   python -m mlbox_ssh_run configure --mlbox=examples/mnist --platform=examples/mnist/platform/ssh.yaml
-   ```  
-4. Run two tasks - download data and model training. After each task execution, local mlbox workspace directory
-   will contain tasks' output artifacts (data sets, log files, models etc.):
-   ```shell script
-    python -m mlbox_ssh_run run --mlbox=examples/mnist --platform=examples/mnist/platform/ssh.yaml --task=examples/mnist/run/download.yaml
-
-    python -m mlbox_ssh_run run --mlbox=examples/mnist --platform=examples/mnist/platform/ssh.yaml --task=examples/mnist/run/train.yaml
-   ```
-
-
-
-
-
-
-
-
+# MLCommons-Box SSH Runner
+MLCommons-Box SSH Runner runs boxes (packaged Machine Learning (ML) workloads) in the remote environment. Read
+SSH Runner documentation [here](../../docs/runners/ssh-runner.md).