Elapsed Matmul_Query: 2059883 us
Elapsed Matmul_Key: 2039530 us
Elapsed Matmul_Value: 207731 us
Elapsed Transpose: 36324 us
Elapsed Matmul_QKt: 9010176 us
Elapsed Softmax: 1687256 us
Elapsed Matmul_Output: 475451 us
Validation results:
Softmax (rows sum & D) : OK
Output all finite: YES
* [MAQAO] Info: Dumping samples (host skylake, process 419776)
* [MAQAO] Info: Dumping source info for callchain nodes (host skylake, process 419776)
* [MAQAO] Info: Building/writing metadata (host skylake)
* [MAQAO] Info: Finished collect step (host skylake, process 419776)