The Golem Task model is particularly useful if you can split the job into chunks that can be run in parallel.
In this challenge, you will learn how to run tasks in parallel.
Create an app that will take a folder name as input and will upscale all jpg
images from that folder on remote nodes. You can use the realesr
tool in the provided image.
For this task create a set of 5 low-resolution anime images as input data. Your script should execute the tasks in parallel. Results should be stored in the same folder under upscaled_<originalName>.jpg
name.
Please observe how long it takes to complete the technical part of the tasks. We will ask you for such information in the feedback form.
-
Prepare test data.
-
Create a requestor script that will run the image conversion.
-
Add changes in the
run-tasks-in-parallel
directory. -
Fill out the feedback form and publish changes in your repository fork.
-
Golem JS QuickStart.
-
That image's
hash
is28716c377b8484abe9fb11dfb94b97831233f18a42658d677657f88b
. -
The image had WORKDIR and VOLUME set as
/golem/work
. -
The usage example of the upscaling tool:
realesrgan-ncnn-vulkan -i <input-image-name> -o <output-image_name> -n realesr-animevideov3-x2 -s 2 -f <jpg|png>
. -
Golem JS tasks running examples are here.
-
You can always ask on the Discord
#js-discussion
channel in the SDK section.