Fix build system and failures for linux

[Iainmon]

To build the image, run the following command in the root directory of the project:

$ docker build -t chai-build-env . --platform="linux/amd64" --progress=plain

To make a volume for the container, use the following command:

$ docker volume create chai-vol

This volume part might not be necessary.

To run the container, use the following command:

$ docker run -it --platform="linux/amd64" --name chai-vm --hostname chai-vm --privileged --memory 2g --cpus 2 -v chai-vol chai-build-env

Then you may run the following:

root@chai-vm:/app# cd demos/torchtest
root@chai-vm:/app/demos/torchtest# mkdir build
root@chai-vm:/app/demos/torchtest# cd build
root@chai-vm:/app/demos/torchtest/build# cmake ..
root@chai-vm:/app/demos/torchtest/build# make
root@chai-vm:/app/demos/torchtest/build# ./MyProject
Input: [1, 3, 10, 10]
Output: [1, 3, 10, 10]

Locally, you can do

$ export Torch_DIR=<ChAI-root>/libtorch
$ cd demos/torchtest
$ mkdir build
$ cd build
$ cmake -DCMAKE_PREFIX_PATH=<ChAI-root>/libtorch/share/cmake ..
$ make
$ ./MyProject
Input: [1, 3, 10, 10]
Output: [1, 3, 10, 10]

Current Status

Currently, you need to have a working installation of CMake, PyThon, and Chapel to use ChAI.

General build instructions

After cloning,

$ git clone https://github.com/chapel-lang/ChAI.git
$ cd ChAI

use CMake to build the project.

$ mkdir build
$ cd build
$ cmake -DCMAKE_BUILD_TYPE=Release ..

This will create a build directory and run CMake to generate the necessary build files.

To test if the build environment is working, you can run the following command:

$ make MyExample && make TorchTestBridge

$ ./MyExample
Hello, world!
a: ndarray([[ 0,  1,  2],
         [ 3,  4,  5],
         [ 6,  7,  8]],
       shape = (3, 3),
       rank = 2)

b: ndarray([[ 0,  1,  2],
         [ 3,  4,  5],
         [ 6,  7,  8]],
       shape = (3, 3),
       rank = 2)

c: ndarray([[15, 18, 21],
         [42, 54, 66],
         [69, 90, 111]],
       shape = (3, 3),
       rank = 2)

$ ./TorchTestBridge
Input: [1, 3, 10, 10]
Output: [1, 3, 10, 10]

The first command will compile the MyExample program, which is a simple "Hello, world!" program that uses ChAI. The second command will compile the TorchTestBridge program, which tests the PyTorch interop functionality of ChAI.

See docker-dev-run.md for instructions on how to run the examples in a Docker container. This is useful if either of these build steps fail on your system.

You may also try

$ cd examples/vgg
$ python3 dump_weights.py
$ cd ../..
$ make vgg
$ ./vgg images/frog.jpg
VGG Example Directory: .../ChAI/examples/vgg
Loading labels from .../ChAI/examples/vgg/imagenet/LOC_synset_mapping.txt
Loaded 1000 labels.
Constructing VGG16 model.
Constructed VGG16 model.
Loading VGG16 model weights.
Loaded VGG16 model.
Read file: .../ChAI/build/images/frog.jpg with ext: jpg
imageData shape: (3, 720, 1280)
normalizedImageData shape: (3, 720, 1280)
inputImage shape: (size = 3, sizes = 3 720 1280)
For 'images/frog.jpg' the top 5 predictions are: 
  0: label=tree frog, tree-frog; confidence=27%
  1: label=tailed frog, bell toad, ribbed toad, tailed toad, Ascaphus trui; confidence=8.2%
  2: label=leafhopper; confidence=4.5%
  3: label=leaf beetle, chrysomelid; confidence=1.9%
  4: label=weevil; confidence=1.5%