[ROCm][CI] Fix spec decode profile assertion and logprob test determinism (#35043)

Signed-off-by: Andreas Karatzas <akaratza@amd.com>

tests/v1/sample/test_logprobs.py

@@ -20,6 +20,7 @@ from tests.v1.sample.utils import (
 from vllm import SamplingParams
 from vllm.config.model import LogprobsMode
 from vllm.distributed import cleanup_dist_env_and_memory
+from vllm.platforms import current_platform
 
 from ...conftest import HfRunner, VllmRunner
 
@@ -31,6 +32,23 @@ SAMPLE = BatchLogprobsComposition.SAMPLE
 PROMPT = BatchLogprobsComposition.PROMPT
 SAMPLE_PROMPT = BatchLogprobsComposition.SAMPLE_PROMPT
 
+# On ROCm, floating-point reductions in attention and GEMM kernels are
+# non-associative and sensitive to batch geometry. The ref LLM (no spec
+# decode, default scheduling) and the spec-decode LLM (chunked prefill,
+# different effective batch sizes) follow different reduction orders,
+# producing numerically divergent logprobs that get mis-attributed to
+# spec-decode incorrectness.
+#
+# Force LLM instances into an identical, deterministic execution
+# mode so the test isolates spec-decode correctness only:
+ROCM_DETERMINISM_KWARGS: dict = (
+    dict(
+        max_num_seqs=1,
+    )
+    if current_platform.is_rocm()
+    else {}
+)
+
 
 @pytest.fixture(
     scope="module",
@@ -1035,6 +1053,7 @@ def test_spec_decode_logprobs(
         logprobs_mode=logprobs_mode,
         gpu_memory_utilization=0.4,
         enable_prefix_caching=False,
+        **ROCM_DETERMINISM_KWARGS,
     )
     ref_results = ref_llm.generate(
         [prompt, prompt], [sampling_params, penalty_sampling_params]
@@ -1064,6 +1083,7 @@ def test_spec_decode_logprobs(
         enable_chunked_prefill=True,
         max_num_batched_tokens=32,
         enable_prefix_caching=False,
+        **ROCM_DETERMINISM_KWARGS,
     )
     spec_results = spec_llm.generate(
         [prompt, prompt], [sampling_params, penalty_sampling_params]