openvino

7b10f31d - [GPU] Fix out_md of onednn reduction primitive when input/output dimensions are different (#33637)

[GPU] Fix out_md of onednn reduction primitive when input/output dimensions are different (#33637) ### Description of the issue(symptom, root-cause, how it was resolved) - The onednn reduction primitive requires the source and destination tensors to have the same number of dimensions. But model in ticket has reducesum with input 5d and output 4d. Current onednn reduce pd creation doesn't consider it. - Create output_md with same dimension of input shape. #### The code and line that caused this issue (if it is not changed directly) - https://github.com/openvinotoolkit/openvino/blob/251976936f0607a6e655e9c33508781a45509eba/src/plugins/intel_gpu/src/graph/impls/onednn/reduce_onednn.cpp#L61 #### Reproduction step and snapshot (if applicable. Do not attach for customer model) - $ benchmark_app -d GPU.1 -m ~/task/cvs179227/WavLM-Base+_88080_new_sim.xml -hint latency #### Problematic graph - ReduceSum has 5d input and 4d output <img width="484" height="526" alt="image" src="https://github.com/user-attachments/assets/333f2f1b-6a09-4699-a7fb-b26eed3de0a7" /> #### Checklist - [x] Is it a proper fix? (not a workaround) - [x] Did you include test case for this fix, if necessary? - [x] Did you review existing test that can be extended to cover this scenario? Which test did you review? ### Tickets: - 179227