Damian Kurek

### Details: - Fix interpolate 11 op case when sizes or scales are not constant. The additional argument shadowed TensorAccessor underlying tensor on port 2 and therefore shape inference was...

[GPU] Swiglu OPT kernel implementation

5

### Details: - Implement Swiglu OPT kernel - Use fast erf function instead of slow OpenCL erf ### Tickets: -

[GPU] Variadic split merge implementation

1

### Details: - Implement variadic split merge - 10% performance improvement in Phi-4 model due to reduced strided slices - Very slight performance improvement in dynamic Stable Diffusion due to...