BufferOverflowException in single task graph implementation

[otabuzzman]

Exception raised after adding 4th layer:

java.nio.BufferOverflowException
        at java.base/java.nio.Buffer.nextPutIndex(Buffer.java:744)
        at java.base/java.nio.HeapByteBuffer.putInt(HeapByteBuffer.java:509)
        at tornado.runtime@1.0.8/uk.ac.manchester.tornado.runtime.tasks.TornadoTaskGraph.addInner(TornadoTaskGraph.java:671)
        at tornado.runtime@1.0.8/uk.ac.manchester.tornado.runtime.tasks.TornadoTaskGraph.addInner(TornadoTaskGraph.java:2188)
        at tornado.runtime@1.0.8/uk.ac.manchester.tornado.runtime.tasks.TornadoTaskGraph.addTask(TornadoTaskGraph.java:2256)
        at tornado.api@1.0.8/uk.ac.manchester.tornado.api.TaskGraph.addTask(TaskGraph.java:82)
        at com.otabuzzman.llmj.GPT2.forward(GPT2.java:940)
        at com.otabuzzman.llmj.TestGpt2.run(TestGpt2.java:122)
        at com.otabuzzman.llmj.TestGpt2.main(TestGpt2.java:207)

[otabuzzman]

Workaround: set highLevelCode = new byte[20480] in ./TornadoVM/tornado-runtime/.../runtime/tasks/TornadoTaskGraph.java.