deeplearning4j/deeplearning4j · error · IllegalStateException
Unknown array type (or not an array):
Error message
Unknown array type (or not an array):
What it means
ArrayUtil's dimension/length probe walks nested arrays and, at the deepest level, matches the element against supported primitive array types to return its length. When the leaf is not a recognized array type, it throws IllegalStateException including the offending object's class, since it cannot determine a dimension length. Like error 4263, it is an internal type-dispatch guard exposed by malformed input.
Solutions
- Verify the input is a properly nested array with supported primitive-array leaves before calling.
- Convert leaves to a supported type (e.g. float[]) prior to dimension probing.
- Use ArrayUtil.length(Object) or NDArray shape APIs (INDArray.shape()) for non-primitive data.
- Catch IllegalStateException and log the offending class (included in the message) to fix the input.
Example fix
// before
Object[] nested = { java.util.Arrays.asList(1f, 2f) };
int len = ArrayUtil.probeDimension(nested, 0); // throws Unknown array type (or not an array)
// after
Object[] nested = { new float[]{1f, 2f} };
int len = ArrayUtil.probeDimension(nested, 0); // 2 Defensive patterns
Strategy: type-guard
Validate before calling
// Java
static boolean isNestedPrimitiveArrayTree(Object o) {
if (o instanceof Object[]) {
for (Object e : (Object[]) o) if (!isNestedPrimitiveArrayTree(e)) return false;
return true;
}
return o != null && o.getClass().isArray() && !(o instanceof Object[]);
} Type guard
static boolean isPrimitiveArray(Object o) { return o != null && o.getClass().isArray() && !o.getClass().getComponentType().equals(Object.class); } Try / catch
try {
int len = ArrayUtil.probeDimension(nested, dim);
} catch (IllegalStateException e) {
// message contains offending class; fix input or use custom walker
} Prevention
- Verify input is truly nested arrays, not Collections or scalars
- Convert Lists to arrays before dimension probing
- Use ArrayUtil.length(Object) for generic objects
- Log the class name from the message to trace malformed data sources
When it happens
Trigger: Calling a dimension/length helper such as ArrayUtil.probeDimension(...) on an Object[] tree whose leaf element is a String, boxed number, POJO, or a primitive-array type not covered by the instanceof chain.
Common situations: Inspecting shapes of heterogeneous or deserialized object graphs; accidentally passing a single scalar or Collection instead of nested arrays; using byte[]/Object leaf data parsed from external sources.
Understand the failure class
Background: Type mismatch errors: IllegalArgumentException, TypeError and type guards across 150 open-source libraries — this error's family across 150 libraries.
Related errors
- Base array is not float[]
- Can't get value for non-regression Metric
- Can't get value for non-regression Metric
- Can't get value for non-ROC Metric
- Unknown array type
AI-assisted analysis of deeplearning4j/deeplearning4j@4c22ac5fe4 (2026-09-07).
Data as JSON: /api/errors/1f8cbf24d7bb1c98.
Report an issue: GitHub.
Appendix: source
Thrown at nd4j/nd4j-common/src/main/java/org/nd4j/common/util/ArrayUtil.java:4633
return ((Object[]) current).length;
} else if (current instanceof double[]) {
return ((double[]) current).length;
} else if (current instanceof float[]) {
return ((float[]) current).length;
} else if (current instanceof long[]) {
return ((long[]) current).length;
} else if (current instanceof int[]) {
return ((int[]) current).length;
} else if (current instanceof byte[]) {
return ((byte[]) current).length;
} else if (current instanceof char[]) {
return ((char[]) current).length;
} else if (current instanceof boolean[]) {
return ((boolean[]) current).length;
} else if (current instanceof short[]) {
return ((short[]) current).length;
} else
throw new IllegalStateException("Unknown array type (or not an array): " + current.getClass()); //Should never happen
}
/**
* Compute the inverse permutation indices for a permutation operation<br>
* Example: if input is [2, 0, 1] then output is [1, 2, 0]<br>
* The idea is that x.permute(input).permute(invertPermutation(input)) == x
*
* @param input 1D indices for permutation
* @return 1D inverted permutation
*/
public static int[] invertPermutation(int... input){
int[] target = new int[input.length];
for(int i = 0 ; i < input.length ; i++){
target[input[i]] = i;
}
return target;View on GitHub (pinned to 4c22ac5fe4)