We want the building procedure generates Docker images so that we can run PaddlePaddle applications on Kubernetes clusters.
We want to make the building procedures:
We want to build .deb packages so that enterprise users can run PaddlePaddle applications without Docker.
1. Static, can reproduce easily.
1. Generate python `whl` packages that can be widely use cross many distributions.
1. Build different binaries per release to satisfy different environments:
- Binaries for different CUDA and CUDNN versions, like CUDA 7.5, 8.0, 9.0
- Binaries containing only capi
- Binaries for python with wide unicode support or not.
1. Build docker images with PaddlePaddle pre-installed, so that we can run
PaddlePaddle applications directly in docker or on Kubernetes clusters.
We want to minimize the size of generated Docker images and .deb packages so to reduce the download time.
To achieve this, we created a repo: https://github.com/PaddlePaddle/buildtools
which gives several docker images that are `manylinux1` sufficient. Then we
can build PaddlePaddle using these images to generate corresponding `whl`
binaries.
We want to encapsulate building tools and dependencies in a *development* Docker image so to ease the tools installation for developers.
## Run The Build
Developers use various editors (emacs, vim, Eclipse, Jupyter Notebook), so the development Docker image contains only building tools, not editing tools, and developers are supposed to git clone source code into their development computers and map the code into the development container.
### Build Evironments
We want the procedure and tools also work with testing, continuous integration, and releasing.
So we need two Docker images for each version of PaddlePaddle:
1.`paddle:<version>-dev`
This a development image contains only the development tools and standardizes the building procedure. Users include:
- developers -- no longer need to install development tools on the host, and can build their current work on the host (development computer).
Choose one docker image that suit your environment and run the following
- release engineers -- use this to build the official release from certain branch/tag on Github.com.
command to start a build:
- document writers / Website developers -- Our documents are in the source repo in the form of .md/.rst files and comments in source code. We need tools to extract the information, typeset, and generate Web pages.
Of course, developers can install building tools on their development computers. But different versions of PaddlePaddle might require different set or version of building tools. Also, it makes collaborative debugging easier if all developers use a unified development environment.
The development image should include the following tools:
cd Paddle
docker run --rm-v$PWD:/paddle -e"WITH_GPU=OFF"-e"WITH_AVX=ON"-e"WITH_TESTING=OFF"-e"RUN_TEST=OFF"-e"PYTHON_ABI=cp27-cp27mu" paddlepaddle/paddle_manylinux_devel /paddle/paddle/scripts/docker/build.sh
- gcc/clang
```
- nvcc
- Python
- sphinx
- woboq
- sshd
Many developers work on a remote computer with GPU; they could ssh into the computer and `docker exec` into the development container. However, running `sshd` in the container allows developers to ssh into the container directly.
After the build finishes, you can get output `whl` package under
`build/python/dist`.
1.`paddle:<version>`
This command mounts the source directory on the host into `/paddle` in the container, then run the build script `/paddle/paddle/scripts/docker/build.sh`
in the container. When it writes to `/paddle/build` in the container, it writes to `$PWD/build` on the host indeed.
This is the production image, generated using the development image. This image might have multiple variants:
### Build Options
- GPU/AVX `paddle:<version>-gpu`
Users can specify the following Docker build arguments with either "ON" or "OFF" value:
- GPU/no-AVX `paddle:<version>-gpu-noavx`
- no-GPU/AVX `paddle:<version>`
- no-GPU/no-AVX `paddle:<version>-noavx`
We allow users to choose between GPU and no-GPU because the GPU version image is much larger than then the no-GPU version.
| Option | Default | Description |
| ------ | -------- | ----------- |
| `WITH_GPU` | OFF | Generates NVIDIA CUDA GPU code and relies on CUDA libraries. |
| `WITH_AVX` | OFF | Set to "ON" to enable AVX support. |
| `WITH_TESTING` | ON | Build unit tests binaries. |
| `WITH_MKLDNN` | ON | Build with [Intel® MKL DNN](https://github.com/01org/mkl-dnn) support. |
| `WITH_MKLML` | ON | Build with [Intel® MKL](https://software.intel.com/en-us/mkl) support. |
| `WITH_GOLANG` | ON | Build fault-tolerant parameter server written in go. |
| `WITH_SWIG_PY` | ON | Build with SWIG python API support. |
| `WITH_C_API` | OFF | Build capi libraries for inference. |
| `WITH_PYTHON` | ON | Build with python support. Turn this off if build is only for capi. |
| `WITH_STYLE_CHECK` | ON | Check the code style when building. |
| `PYTHON_ABI` | "" | Build for different python ABI support, can be cp27-cp27m or cp27-cp27mu |
| `RUN_TEST` | OFF | Run unit test immediently after the build. |
| `WITH_DOC` | OFF | Build docs after build binaries. |
| `WOBOQ` | OFF | Generate WOBOQ code viewer under `build/woboq_out` |
We allow users the choice between AVX and no-AVX, because some cloud providers don't provide AVX-enabled VMs.
## Docker Images
## Development Environment
You can get the latest PaddlePaddle docker images by
`docker pull paddlepaddle/paddle:<version>` or build one by yourself.
Here we describe how to use above two images. We start from considering our daily development environment.
### Official Docker Releases
Developers work on a computer, which is usually a laptop or desktop:
Build PaddlePaddle docker images are quite simple since PaddlePaddle can
be installed by just running `pip install`. A sample `Dockerfile` is:
A principle here is that source code lies on the development computer (host) so that editors like Eclipse can parse the source code to support auto-completion.
```dockerfile
FROM nvidia/cuda:7.5-cudnn5-runtime-centos6
RUN yum install-y centos-release-SCL
RUN yum install-y python27
# This whl package is generated by previous build steps.
RUN pip install /paddlepaddle-0.10.0-cp27-cp27mu-linux_x86_64.whl &&rm-f /*.whl
```
Then build the image by running `docker build -t [REPO]/paddle:[TAG] .` under
the directory containing your own `Dockerfile`.
## Usages
- NOTE: note that you can choose different base images for your environment, you can find all the versions [here](https://hub.docker.com/r/nvidia/cuda/).
### Build the Development Docker Image
### Use Docker Images
The following commands check out the source code to the host and build the development image `paddle:dev`:
Suppose that you have written an application program `train.py` using
PaddlePaddle, we can test and run it using docker:
docker run --rm-it-v$PWD:/work paddlepaddle/paddle /work/a.py
cd paddle
docker build -t paddle:dev .
```
```
The `docker build` command assumes that `Dockerfile` is in the root source tree. Note that in this design, this `Dockerfile` is this only one in our repo.
But this works only if all dependencies of `train.py` are in the production image. If this is not the case, we need to build a new Docker image from the production image and with more dependencies installs.
Users can specify a Ubuntu mirror server for faster downloading:
Given the development image `paddle:dev`, the following command builds PaddlePaddle from the source tree on the development computer (host):
Our [book repo](https://github.com/paddlepaddle/book) also provide a docker
image to start a jupiter notebook inside docker so that you can run this book
using docker:
```bash
```bash
docker run --rm-v$PWD:/paddle -e"WITH_GPU=OFF"-e"WITH_AVX=ON"-e"WITH_TESTING=OFF"-e"RUN_TEST=OFF" paddle:dev
docker run -d-p 8888:8888 paddlepaddle/book
```
```
This command mounts the source directory on the host into `/paddle` in the container, so the default entry point of `paddle:dev`, `build.sh`, could build the source code with possible local changes. When it writes to `/paddle/build` in the container, it writes to `$PWD/build` on the host indeed.
Please refer to https://github.com/paddlepaddle/book if you want to build this
docker image by your self.
`build.sh` builds the following:
- PaddlePaddle binaries,
-`$PWD/build/paddle-<version>.deb` for production installation, and
-`$PWD/build/Dockerfile`, which builds the production Docker image.
Users can specify the following Docker build arguments with either "ON" or "OFF" value:
### Run Distributed Applications
-`WITH_GPU`: ***Required***. Generates NVIDIA CUDA GPU code and relies on CUDA libraries.
-`WITH_AVX`: ***Required***. Set to "OFF" prevents from generating AVX instructions. If you don't know what is AVX, you might want to set "ON".
-`WITH_TEST`: ***Optional, default OFF***. Build unit tests binaries. Once you've built the unit tests, you can run these test manually by the following command:
```bash
docker run --rm-v$PWD:/paddle -e"WITH_GPU=OFF"-e"WITH_AVX=ON" paddle:dev sh -c"cd /paddle/build; make coverall"
```
-`RUN_TEST`: ***Optional, default OFF***. Run unit tests after building. You can't run unit tests without building it.
### Build the Production Docker Image
In our [API design doc](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/api.md#distributed-training), we proposed an API that starts a distributed training job on a cluster. This API need to build a PaddlePaddle application into a Docker image as above and calls kubectl to run it on the cluster. This API might need to generate a Dockerfile look like above and call `docker build`.
The following command builds the production image:
Of course, we can manually build an application image and launch the job using the kubectl tool:
This production image is minimal -- it includes binary `paddle`, the shared library `libpaddle.so`, and Python runtime.
## Docker Images for Developers
### Run PaddlePaddle Applications
We have a special docker image for developers:
`paddlepaddle/paddle:<version>-dev`. This image is also generated from
https://github.com/PaddlePaddle/buildtools
Again the development happens on the host. Suppose that we have a simple application program in `a.py`, we can test and run it using the production image:
This a development image contains only the
development tools and standardizes the building procedure. Users include:
```bash
- developers -- no longer need to install development tools on the host, and can build their current work on the host (development computer).
docker run --rm-it-v$PWD:/work paddle /work/a.py
- release engineers -- use this to build the official release from certain branch/tag on Github.com.
```
- document writers / Website developers -- Our documents are in the source repo in the form of .md/.rst files and comments in source code. We need tools to extract the information, typeset, and generate Web pages.
But this works only if all dependencies of `a.py` are in the production image. If this is not the case, we need to build a new Docker image from the production image and with more dependencies installs.
Of course, developers can install building tools on their development computers. But different versions of PaddlePaddle might require different set or version of building tools. Also, it makes collaborative debugging easier if all developers use a unified development environment.
### Build and Run PaddlePaddle Applications
The development image contains the following tools:
We need a Dockerfile in https://github.com/paddlepaddle/book that builds Docker image `paddlepaddle/book:<version>`, basing on the PaddlePaddle production image:
- gcc/clang
- nvcc
- Python
- sphinx
- woboq
- sshd
```
Many developers work on a remote computer with GPU; they could ssh into the computer and `docker exec` into the development container. However, running `sshd` in the container allows developers to ssh into the container directly.
FROM paddlepaddle/paddle:<version>
RUN pip install -U matplotlib jupyter ...
COPY . /book
EXPOSE 8080
CMD ["jupyter"]
```
The book image is an example of PaddlePaddle application image. We can build it
```bash
### Development Workflow
git clone https://github.com/paddlepaddle/book
cd book
docker build -t book .
```
### Build and Run Distributed Applications
Here we describe how the workflow goes on. We start from considering our daily development environment.
In our [API design doc](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/api.md#distributed-training), we proposed an API that starts a distributed training job on a cluster. This API need to build a PaddlePaddle application into a Docker image as above and calls kubectl to run it on the cluster. This API might need to generate a Dockerfile look like above and call `docker build`.
Developers work on a computer, which is usually a laptop or desktop:
Of course, we can manually build an application image and launch the job using the kubectl tool:
A principle here is that source code lies on the development computer (host) so that editors like Eclipse can parse the source code to support auto-completion.
```
### Reading source code with woboq codebrowser
### Reading source code with woboq codebrowser
For developers who are interested in the C++ source code, please use -e "WOBOQ=ON" to enable the building of C++ source code into HTML pages using [Woboq codebrowser](https://github.com/woboq/woboq_codebrowser).
For developers who are interested in the C++ source code, please use -e "WOBOQ=ON" to enable the building of C++ source code into HTML pages using [Woboq codebrowser](https://github.com/woboq/woboq_codebrowser).
- The following command builds PaddlePaddle, generates HTML pages from C++ source code, and writes HTML pages into `$HOME/woboq_out` on the host:
- The following command builds PaddlePaddle, generates HTML pages from C++ source code, and writes HTML pages into `$HOME/woboq_out` on the host:
```bash
```bash
docker run -v$PWD:/paddle -v$HOME/woboq_out:/woboq_out -e"WITH_GPU=OFF"-e"WITH_AVX=ON"-e"WITH_TEST=ON"-e"WOBOQ=ON" paddle:dev
docker run -v$PWD:/paddle -v$HOME/woboq_out:/woboq_out -e"WITH_GPU=OFF"-e"WITH_AVX=ON"-e"WITH_TEST=ON"-e"WOBOQ=ON" paddlepaddle/paddle:latest-dev
```
```
- You can open the generated HTML files in your Web browser. Or, if you want to run a Nginx container to serve them for a wider audience, you can run:
- You can open the generated HTML files in your Web browser. Or, if you want to run a Nginx container to serve them for a wider audience, you can run:
