Hi all I have a project where a simulation SIL is run and ex

Question

Hi all I have a project where a simulation SIL is run and exported signals from the simulation are processed in python The simulation happens inside a container The simulation container is exchanged more frequently than the python scripts are adapted I read in the Flyte documentation that the task is hard wired to a container image upon registration Is it possible to define a container image as input for a task So that someone can specify the path to the container e g within the Flyte GUI Thank you in advance for your answers

Ketan (kumare3) · Answer

It is indeed possible but you will have to use dynamic workflows

Ketan (kumare3) · Answer

Firstly welcome to the community and thank you for trying out flytes

Tim Hebbeler · Answer

Many thanks for the extremely fast reply and the friendly acceptance into the community I might have one more question The simulation consists of a simulator and the simulation code compiled as a library Would it be good to implement the simulation via dynamic workflows or would it make more sense to only provide the simulator as an image and the simulation libraries as a zip file via an input for the task

Ketan (kumare3) · Answer

You are lost welcome If you have Python bindings my recommendation would be to use flytekit to implement it as Flyte native workflows Flyte manages to zip and install things to run on new versions you get a good clean versioning story But I do not know your architecture to comment Why is simulator code a container image How often what changes the base simulator or the usage of the simulator Understanding this will Help us better help you

Ketan (kumare3) · Answer

But I also understand that the his could be proprietary

Tim Hebbeler · Answer

Currently everything is bundled inside one image simulator and simulation model sorry i think simulation code was not clear The simulation model like an <https fmi standard org |FMU> and simulator are written in C C++ We want to provide a workflow in which users can specify the FMU version as input via a GUI The signals of the FMU will later be analyzed in python We think Flyte s idea is great that each workflow is versioned and thus each user can work independently

Tim Hebbeler · Answer

< Ketan kumare3 > can you give me a hint how to set the container image dynamically I tried to instantiate the task decorator in the function but i don t think the task resolver DefaultTaskResolver works like that Error inside property self task resolver location python auto container py 151 ```Error looking for LHS in main ```

Ketan (kumare3) · Answer

You will have to use dynamic workflows tasks

Tim Hebbeler · Answer

I did I found a solution ```my task container image = image ``` inside a dynamic task But doesn t look very elegant Do you have better idea

Ketan (kumare3) · Answer

We can add that as a method

Ketan (kumare3) · Answer

The problem is container image is a fixed property to maintain reproducibility

Ketan (kumare3) · Answer

But you can also use raw container task

Ketan (kumare3) · Answer

Ohh there is one

Ketan (kumare3) · Answer

Using your own custom resolver

Ketan (kumare3) · Answer

But that is involved

Ketan (kumare3) · Answer

It seems you have a raw container

Ketan (kumare3) · Answer

Can I ping you later I am infact out today it s a holiday

Ketan (kumare3) · Answer

With family now

Tim Hebbeler · Answer

Sure have a nice holiday I will take a look a raw containers Thanks a lot for the hint