Failed to get registration from op code BATCH_MATMUL

TensorFlow Special Interest Groups Micro
,
1

I am trying to build and attention based tflite model for a microcontroller and this requires batch matrix multiplication. After going through the tflite source code, I found out “micro_mutable_op_resolver.h” does not have the addBatchMatMul() function, however “schema_generated.h” has “BuiltinOperator_BATCH_MATMUL” and “flatbufferconversions.cc” has the “ParseBatchMatMul”. The problem is I cannot find the .cc file with where "Register_BATCH_MAT_MUL() function is located.
Any help in getting matrix multiplication to work within my tflite model would be greatly appreciated.

3

Hi @Kvenn,

Is it the same function `Register_BATCH_MATMUL_REF()’ what you are looking. If it is the same, please find the path for .cc file here.

Thank You.