Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

Hello, I am following this example for <https://docs.flyte.org/en/latest/user_guide/extending/custom_types.html#advanced-custom-types|defining a custom type> as a template for my own custom type transformer, but am having trouble getting caching to work with it.
• I think the issue might be be the random non-deterministic `uri` being returned from `to_literal` method in the example (I am using `uri` to store the path where the object is being stored using my custom serialization logic that is not json)
I _think_ the fix might be to return a deterministic `Literal` from my custom typetransformer’s `to_literal` method (by making the `uri` deterministic)
Curious if there are are any examples/pointers for this?
Or if someone has figured out a simpler way to do this :slightly_smiling_face:

that is true in that that’s likely what’s happening yeah.

and the issue is that you want tasks _downstream_ of this task, to _not_ run correct?  (because the output hasn’t really changed, just the uri)

&gt; and the issue is that you want tasks _downstream_ of this task, to _not_ run correct?
Yes, that is correct

I was previously using `ctx.file_access.get_random_remote_path` but that obviously causes misses

you should probably control this with hash method.

<https://docs.flyte.org/en/latest/user_guide/development_lifecycle/caching.html#how-does-caching-of-offloaded-objects-work>

this is a more accurate link <https://docs.flyte.org/en/latest/user_guide/development_lifecycle/caching.html#caching-of-non-flyte-offloaded-objects>