Caused by: ai.djl.engine.EngineException: invalid index of a 0-dim tensor. Use `tensor.item()` in Python or `tensor.item[T]()` in C++ to convert a 0-dim tensor to a number

[havetobe]

public void train(NDArray logMat, NDArray qMat, int batchSize, float learningRate, int epochs) {
    Tools.CustomLogger customLogger = new Tools.CustomLogger("logs", "both");
    Loss binaryCrossEntropy = Loss.sigmoidBinaryCrossEntropyLoss();

    for (int epoch = 0; epoch < epochs; epoch++) {
        float totalLoss = 0;
        int batchCount = 0;

        for (int i = 0; i < nUser; i += batchSize) {
            int batchEnd = Math.min(i + batchSize, nUser);
            NDArray batchTheta = thetaMat.get(String.format("%d:%d", i, batchEnd));
            NDArray batchLog = logMat.get(String.format("%d:%d", i, batchEnd));

            NDArray pred = forward(batchTheta, qMat);
            NDArray loss = binaryCrossEntropy.evaluate(
                    new NDList(batchLog),
                    **new NDList(pred)).get(0);**

            backwardAndUpdate(loss, batchTheta, qMat, learningRate);

            totalLoss += loss.mean().getFloat();
            batchCount++;
        }

        float avgLoss = totalLoss / batchCount;
        customLogger.write(String.format("Epoch %d, Loss: %.4f", epoch + 1, avgLoss), "both");
    }
}

“ NDArray loss = binaryCrossEntropy.evaluate( new NDList(batchLog), new NDList(pred)).get(0); “This line of code reports an error ”ai.djl.engine.EngineException: invalid index of a 0-dim tensor. use tensor.item() in Python or tensor.item[T]() in C++ to convert a 0-dim tensor to a number”, the solution on the web is ”Modify loss.data[0] to loss.item() ”, but there is no item() in DJL, and repeated use of chatgpt, claude, and even deepseek could not modify this error.

Full Error Reporting： org.opentest4j.AssertionFailedError: Unexpected exception thrown: ai.djl.engine.EngineException: invalid index of a 0-dim tensor. Use tensor.item() in Python or tensor.item[T]() in C++ to convert a 0-dim tensor to a number at com.idcd.model.IDCDModelTest.testTraining(IDCDModelTest.java:48) at java.util.ArrayList.forEach(ArrayList.java:1259) at java.util.ArrayList.forEach(ArrayList.java:1259) Caused by: ai.djl.engine.EngineException: invalid index of a 0-dim tensor. Use tensor.item() in Python or tensor.item[T]() in C++ to convert a 0-dim tensor to a number at ai.djl.pytorch.jni.PyTorchLibrary.torchIndexAdvGet(Native Method) at ai.djl.pytorch.jni.JniUtils.indexAdv(JniUtils.java:423) at ai.djl.pytorch.engine.PtNDArrayIndexer.get(PtNDArrayIndexer.java:74) at ai.djl.ndarray.NDArray.get(NDArray.java:594) at ai.djl.ndarray.NDArray.get(NDArray.java:583) at ai.djl.ndarray.NDArray.get(NDArray.java:628) at com.idcd.model.IDCDModel.train(IDCDModel.java:66) at com.idcd.model.IDCDModelTest.lambda$0(IDCDModelTest.java:49) ... 72 more