openvino

e2766874 - [GPU] fix torchvisionRaft fails on last tensor check issue (#31948)

[GPU] fix torchvisionRaft fails on last tensor check issue (#31948) ### Details: - Switch the activation function used by the Relu layer from CPU to GPU and fuse it with the upper Convolution layer to resolve the discrepancy between computed results and reference values during dynamic testing. ### Description of the issue(symptom, root-cause, how it was resolved) - ReLU choose CPU implementation (activation_cpu_impl) cannot handle padding correctly during dynamic test. - Test with choosing OCL implementation (activation_ref) by skipping mark_node, and the test result is correct. - Allow to fuse activation (ReLU, GeLU, and etc.) to Convolution, Gemm, and Fully_connected, so both of them will use OCL implementation. #### The code and line that caused this issue (if it is not changed directly) https://github.com/openvinotoolkit/openvino/blob/0a49c3257ff74d1f8e1ee60b72739808600fa4e9/src/plugins/intel_gpu/src/graph/graph_optimizer/prepare_primitive_fusing.cpp#L735 #### Reproduction step and snapshot (if applicable. Do not attach for customer model) - python -m pytest test_dynamism.py \ -n 4 --tb=native \ --env_conf=.automation/env_config.yml \ --test_conf=.automation/test_configs_custom/reshape/desktop_test_config.yml \ -m "not launch_only_if_manually_specified" \ --pregen_irs=models/irs_mapping.csv \ --tf_models_version=1.15.2 \ --modules pipelines/production/pytorch/heavy \ -k "PyTorch_TorchvisionRaftLarge_api_2_True" \ --dynamism_type=range_values \ --log-cli-level DEBUG #### Problematic graph - Original IR <img width="2166" height="1597" alt="image" src="https://github.com/user-attachments/assets/d0015490-3175-4758-ad7a-db1b4eb778fe" /> - Current IR <img width="3154" height="1413" alt="image" src="https://github.com/user-attachments/assets/d0c25bca-7b91-47dc-961f-93524dff19d3" /> #### Checklist - [v] Is it a proper fix? (not a workaround) - [v] Did you include test case for this fix, if necessary? - [v] Did you review existing test that can be extended to cover this scenario? Which test did you review? ### Tickets: - *CVS-169966*