Use numba 0.61 for python 3.10+ to support numpy>=2

requirements/common.txt

@@ -1,7 +1,7 @@
 cachetools
 psutil
 sentencepiece  # Required for LLaMA tokenizer.
-numpy < 2.0.0
+numpy
 requests >= 2.26.0
 tqdm
 blake3

requirements/cuda.txt

@@ -1,7 +1,8 @@
 # Common dependencies
 -r common.txt
 
-numba == 0.60.0 # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
+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; python_version > '3.9'
 
 # Dependencies for NVIDIA GPUs
 ray[cgraph]>=2.43.0, !=2.44.* # Ray Compiled Graph, required for pipeline parallelism in V1.

requirements/rocm.txt

@@ -1,7 +1,8 @@
 # Common dependencies
 -r common.txt
 
-numba == 0.60.0 # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
+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; python_version > '3.9'
 
 # Dependencies for AMD GPUs
 awscli

requirements/test.in

@@ -38,7 +38,9 @@ buildkite-test-collector==0.1.9
 genai_perf==0.0.8
 tritonclient==2.51.0
 
-numpy < 2.0.0
+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; python_version > '3.9'
+numpy
 runai-model-streamer==0.11.0
 runai-model-streamer-s3==0.11.0
 fastsafetensors>=0.1.10

requirements/test.txt

@@ -219,7 +219,7 @@ libnacl==2.1.0
     # via tensorizer
 librosa==0.10.2.post1
     # via -r requirements/test.in
-llvmlite==0.43.0
+llvmlite==0.44.0
     # via numba
 lm-eval==0.4.4
     # via -r requirements/test.in
@@ -262,8 +262,10 @@ networkx==3.2.1
     # via torch
 nltk==3.9.1
     # via rouge-score
-numba==0.60.0
-    # via librosa
+numba==0.61.0
+    # via
+    #   -r requirements/test.in
+    #   librosa
 numexpr==2.10.1
     # via lm-eval
 numpy==1.26.4