[code-reflection] RFR: Model lifetimes of onnx session-related objects more explicitly

Adam Sotona asotona at openjdk.org
Mon Mar 3 10:07:06 UTC 2025


On Fri, 28 Feb 2025 12:42:24 GMT, Maurizio Cimadamore <mcimadamore at openjdk.org> wrote:

> The class representing an onnx session is auto closeable. But, in the current code, a session is closed immediately after its `run` method is called. This is problematic because a session returns some ORTValues (tensors) which also need to be freed, but that cannot be freed immediately after calling `run` (as they need to be used by clients).
> 
> To address this problem, I tweaked the session code to accept an external arena. All the allocation of session-related data structures now happens using that external arena. This means that the client can now be in charge of managing the lifetime of a session (see changes to MNIST demo).
> 
> To test, I tweaked the MNIST code to do 10K iterations on each button pressed. Predictably, a single button pressed resulted in over 3g of memory being leaked. With these changes the memory arrives at ~400K (there is still some minor leak, but not sure worth pushing more).
> 
> If the changes to the demo are not deemed good, I can withdraw this PR -- I mostly wanted to capture the result of my exploration somewhere.

I'm a bit confused. Are we discussing to hide Onnx API from the Onnx API and expose the whole stuff as an extension of FFM API / Arena ?

I was thinking a bit opposite - to wrap the low-level FFM components like memory segments and arenas handling inside the Onnx API. Practically it may look almost the same, just the naming should match what Onnx users are familiar with:


try (OnnxEnvironment onnxEnv= OnnxEnvironment.of(logLevel)) {
      var prediction = onnxEnv.execute(MethodHandles.lookup(),
                           () -> cnn(imageTensor));
      var result = prediction.data().toArray(ValueLayout.JAVA_FLOAT);
      ...
}


However that does not solve constrution of tensors. We have two options:
- Tesor.ofXyz(arena, ...
- onnxEnv.createTensorXyz(...

-------------

PR Comment: https://git.openjdk.org/babylon/pull/332#issuecomment-2693852933


More information about the babylon-dev mailing list