Skip to content

Latest commit

 

History

History
106 lines (95 loc) · 3.26 KB

run-transform-image.md

File metadata and controls

106 lines (95 loc) · 3.26 KB

Running a Transform Image

Here we address a simple use case of applying a single transform to a set of parquet files. We'll use one of the pre-built docker images from the repository to process the data. Additionally, what follows uses the python runtime image, but the examples below should also work for the ray or spark runtime images (-ray or -spark image name suffix instead of -python).

Getting an image

You may build the transform locally in the repository, for example,

cd transforms/universal/noop/python
make image
docker images | grep noop

produces

quay.io/dataprep1/data-prep-kit/noop-python                     latest   aac55fa3d82d  5 seconds ago  355 MB

Or, you can use the pre-built images (latest, or 0.2.1 or later tags) on quay.io found at https://quay.io/user/dataprep1.

Local Data - Python Runtime

To use an image to process local data we will mount the host input and output directories into the image. Any mount point can be used, but we will use /input and /output.

To process data in the /home/me/input directory and write it to the /home/me/output directory, we mount these directories into the image at the above mount points. So for example, using the locally built noop transform:

docker run  --rm 
    -v /home/me/input:/input \
    -v /home/me/output:/output \
    noop-python:latest 	\
	python noop_transform_python.py \
	--data_local_config "{ \
	    'input_folder'  : '/input', \
	    'output_folder' : '/output' \
	    }"

To run the quay.io located transform instead, substitute quay.io/dataprep1/data-prep-kit/noop-python:latest for noop-python:latest, as follows:

docker run  --rm 
    -v /home/me/input:/input \
    -v /home/me/output:/output \
    quay.io/dataprep1/data-prep-kit/noop-python:latest 	\
	python noop_transform_python.py \
	--data_local_config "{ \
	    'input_folder'  : '/input', \
	    'output_folder' : '/output' \
	    }"

Local Data - Ray Runtime

To use the ray runtime, we must

  1. Switch to using the ray-based image noop-ray:latest
  2. Use the ray runtime python main() defined in noop_transform_ray.py

For example, using the quay.io image

docker run  --rm 
    -v /home/me/input:/input \
    -v /home/me/output:/output \
    quay.io/dataprep1/data-prep-kit/noop-ray:latest 	\
	python noop_transform_ray.py \
	--data_local_config "{ \
	    'input_folder'  : '/input', \
	    'output_folder' : '/output' \
	    }"

This is functionally equivalent to the python-runtime, but additional configuration can be provided (see the ray launcher args) for details.

S3-located Data - Python Runtime

When processing data located in S3 buckets, one can use the same image and specify different --data_s3_* configuration as follows:

docker run  --rm 
    noop-python:latest 	\
	python noop_transform_python.py \
	--data_s3_cred "{ \
	    'access_key'  : '...', \
	    'secret_key' : '...', \
	    'url' : '...', \
	    }"  \
	--data_s3_config "{ \
	    'input_folder'  : '...', \
	    'output_folder' : '...', \
	    }"