.params vs .pt

[DialingSpoon527]

How do I get a model to load from a .params, or save to a .pt?
If I run model.load(Paths.get("path/to/model.params")) I get

Caused by: ai.djl.engine.EngineException: PytorchStreamReader failed reading zip archive: failed finding central directory
	at knot//ai.djl.pytorch.jni.PyTorchLibrary.moduleLoad(Native Method)
	at knot//ai.djl.pytorch.jni.JniUtils.loadModule(JniUtils.java:1789)
	at knot//ai.djl.pytorch.engine.PtModel.load(PtModel.java:99)
	at knot//ai.djl.Model.load(Model.java:110)

I did see this. I only repeat it because it's unanswered, ancient, and not in this category.

[frankfliu]

1. .pt file in DJL referring to torchscript file. You can use python to trace the pytorch model and save it as .pt file
2. .params is DJL parameter file, it's saved by DJL training code. You have to use DJL Block to load the .params file. Means, you need call Model.setBlock() first.
3. The Model.load() method expect a folder or a torchscript file (for PyTorch). The file name extension is only being used to detect the model file when a folder is specified.