[V1] TPU - Add tensor parallel support via Ray (#13618)

Signed-off-by: Alexander Matveev <amatveev@redhat.com>

tests/entrypoints/llm/test_accuracy.py

@@ -42,6 +42,10 @@ def run_test(more_args=None):
             ), f"Expected: {EXPECTED_VALUE} |  Measured: {measured_value}"
 
 
+# TODO: [AlexM] Fix it with new CI/CD tests
+TPU_TP_TEST_STR = ""  #"tensor_parallel_size=4"
+
+
 @pytest.mark.skipif(not current_platform.is_cuda()
                     and not current_platform.is_tpu(),
                     reason="V1 is currently only supported on CUDA and TPU")
@@ -56,6 +60,10 @@ def test_lm_eval_accuracy_v1_engine(monkeypatch):
             # Limit compilation time for TPU V1
             more_args = "max_num_seqs=64"
 
+            # Add TP test (if provided)
+            if TPU_TP_TEST_STR:
+                more_args += ",{}".format(TPU_TP_TEST_STR)
+
         run_test(more_args)