| Previous | Next |
| WINML_ERR_INVALID_BINDING | WINML_ERR_SIZE_MISMATCH |
WINML_ERR_VALUE_NOTFOUND
Investigate unknown Windows ML input or output name at the first returning call, before a wrapper substitutes a generic message. The binding attempted to address a feature name that is absent from the loaded model description, often after model regeneration or an incorrect generated wrapper.
Read the status in context
Windows ML separates model loading, session creation, feature binding, and evaluation. The phase returning the HRESULT decides whether metadata, tensors, device selection, or execution is the useful boundary.
This is name resolution, not tensor allocation or device support. A correctly sized value still fails if its key is not exported by the model. Keep WINML_ERR_VALUE_NOTFOUND, its numeric value, and the first returning operation together.
Diagnostic facts
| Capture | Diagnostic value |
|---|---|
| Exact UTF-16 binding key, model hash/version, descriptor names, generated-code version, and case. | Identifies the concrete object and operation associated with unknown Windows ML input or output name. |
| Whether the missing name is an input, output, or intermediate name not exposed by the model contract. | Separates argument or lifecycle state from a lower-layer provider failure. |
| First WinML/DirectML diagnostic event and the operation phase: load, session, bind, or evaluate. | Creates a stable before-and-after comparison. |
| Known-good model and input produced by the same export toolchain. | Shows whether this condition is the first result or a translated summary. |
Preserve whether the missing name is an input, output, or intermediate name not exposed by the model contract before reinstalling, rebooting, clearing state, or substituting another device or provider.
Verification path
- Print the loaded descriptor names and bind using one copied verbatim. Record the first returned status and any state transition observed.
- Load the previous model version with the same application binary to prove schema drift. Treat a changed result as a separate failure rather than automatic resolution.
- Run one evaluation with deterministic input and capture the descriptor-to-buffer mapping. Repeat the original supported operation so the check remains tied to the same API boundary.
Closing the incident
Targeted correction. Regenerate wrappers or update the binding map from the actual deployed model and version model/schema changes together.
Acceptance criterion. The deployed model and application agree on feature names and a deliberate unknown key fails before evaluation.
Technical references
References for WINML_ERR_VALUE_NOTFOUND on the deployed platform version.
- Microsoft Open Specifications: HRESULT values — defines the status namespace used.
- Microsoft: HRESULT facility extraction — documents the API or lifecycle behind this condition.
- Microsoft: Windows ML desktop workflow — provides ABI, implementation, or protocol context.
- Microsoft: LearningModelBinding — supports the portability and verification limits.
Looking for a different code? Search another status or error code.