Ivan Tikhonov

> @praasz, please comment on value outside Kaldi. Kaldi support was stopped a year ago. I see setting shapes for variables reasonable in the context that we started to produce...

@ilya-lavrenov @slyalin please take a look

> ave you compared IRs produced by Python and C++ paths for all models from the list? We agreed to run this model list: "hf-internal-testing/tiny-random-BloomForCausalLM", "hf-internal-testing/tiny-random-FalconForCausalLM", "hf-internal-testing/tiny-random-Starcoder2ForCausalLM", "hf-internal-testing/tiny-random-GPTJForCausalLM", "hf-internal-testing/tiny-random-StableLmForCausalLM", "hf-internal-testing/tiny-random-LlamaForCausalLM",...

As I can see, Jenkins and ARM jobs fail in other PRs and in the post-commit . Not related to these changes.

It looks like the layer tests failed @evkotov could you take a look?

Should we run any additional validations, e.g. performance? Could invoking these transformations break other patterns?

> Approve Tile transformation, but not the change in SharedOp. IR names would change because output choice is non deterministic. Can we keep the vector to store outputs and check...

Concat to Tile causes an issue in e2e tests with unknown root cause. We decided to exclude Tile from transformation and leave Concat to Broadcast case only. https://github.com/openvinotoolkit/openvino/pull/24661

please provide more details in the description section Details: Optimize loop after TF FE

it looks like, it will conflict with Slice op supporting in RoPE fusion patterns https://github.com/openvinotoolkit/openvino/pull/24379