.. _error-code-evrf049:

.. meta::
   :description: AWS Neuron SDK Graph Compiler error code documentation for error EVRF049.

NCC_EVRF049
===========

**Error message**: ScaledMatmul custom call could not parse backend_config.

The ``__op$block_scaled_dot`` custom call performs matrix multiplication on
MXFP8-quantized tensors. The compiler accepts an empty ``backend_config``,
``"None"``, or valid JSON. It raises this error when JSON parsing fails or a
dimension list contains a value that cannot be converted to an integer.

The ``scaled_dot_backend_config`` object and all of its fields are optional.
Missing fields use empty values and do not trigger this error. Output dtype is
determined by the result tensor's element type, not by a JSON field.

The optional fields in ``scaled_dot_backend_config`` with their defaults are:

- ``lhs_batch_dimensions``: array of batch dimension indices for LHS (default ``[]``)
- ``rhs_batch_dimensions``: array of batch dimension indices for RHS (default ``[]``)
- ``lhs_contracting_dimensions``: array of contracting dimension indices for LHS (default ``[]``)
- ``rhs_contracting_dimensions``: array of contracting dimension indices for RHS (default ``[]``)
- ``element_dtype``: FP8 element dtype ``"float8_e5m2"`` or ``"float8_e4m3fn"`` if specified (default empty string)

To fix this error, provide valid JSON and use integer values in each dimension
array.