<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
Hi,
<div><br>
</div>
<div>We demoed the use of the FFM API and code reflection at the Java One keynote, showing a way to translate Java code to ONNX models and execute in the ONNX runtime.</div>
<div><br>
</div>
<div>Here’s the link to the start of the keynote:</div>
<div><br>
</div>
<div> <a href="https://www.youtube.com/live/mk_2MIWxLI0">https://www.youtube.com/live/mk_2MIWxLI0</a></div>
<div><br>
</div>
<div>Here’s the point in the keynote that starts with FFM API and then proceeds with the ONNX demo:</div>
<div><br>
</div>
<div> <a href="https://youtu.be/mk_2MIWxLI0?t=4027">https://youtu.be/mk_2MIWxLI0?t=4027</a></div>
<div><br>
</div>
<div>The demo model code is here:</div>
<div><br>
</div>
<div> <a href="https://github.com/openjdk/babylon/blob/code-reflection/cr-examples/onnx/src/test/java/oracle/code/onnx/mnist/MNISTModel.java">https://github.com/openjdk/babylon/blob/code-reflection/cr-examples/onnx/src/test/java/oracle/code/onnx/mnist/MNISTModel.java</a></div>
<div><br>
</div>
<div>We are currently working to towards supporting the translation of if/for statements, and will hopefully add some more comprehensive models. We may even explore training if we can perhaps leverage ops in the domain ai.onnx.preview.training and/or other
stuff in the ONNX runtime. Implementing reverse mode auto-diff on ONNX models would be a difficult task, but also fascinating one. </div>
<div><br>
</div>
<div>Paul. </div>
</body>
</html>