[Perf] Optimize MRoPR position preparing performance with numba

requirements/common.txt

@@ -47,3 +47,5 @@ opentelemetry-sdk>=1.26.0  # vllm.tracing
 opentelemetry-api>=1.26.0  # vllm.tracing
 opentelemetry-exporter-otlp>=1.26.0  # vllm.tracing
 opentelemetry-semantic-conventions-ai>=0.4.1  # vllm.tracing
+numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding & Qwen2/2.5-VL/Omni
+numba == 0.61.2; python_version > '3.9'

requirements/cuda.txt

@@ -1,9 +1,6 @@
 # Common dependencies
 -r common.txt
 
-numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9'
-
 # Dependencies for NVIDIA GPUs
 ray[cgraph]>=2.43.0, !=2.44.* # Ray Compiled Graph, required for pipeline parallelism in V1.
 torch==2.7.0

requirements/rocm.txt

@@ -1,9 +1,6 @@
 # Common dependencies
 -r common.txt
 
-numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9'
-
 # Dependencies for AMD GPUs
 boto3
 botocore