A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 17:15 UTC
Use numba 0.61 for python 3.10+ to support numpy>=2 (#15692)
Signed-off-by: cyy <cyyever@outlook.com>
cyyever committed
Mar 29, 2025 at 12:11 UTC
8427f70493ed67bf26cb9e7fa98ac202b991c37d
5 files changed
+13
-7
requirements/common.txt
+1
-1
@@ -1,7 +1,7 @@
1
cachetools
2
psutil
3
sentencepiece # Required for LLaMA tokenizer.
4
-numpy < 2.0.0
4
+numpy
5
requests >= 2.26.0
6
tqdm
7
blake3
requirements/cuda.txt
+2
-1
@@ -1,7 +1,8 @@
1
# Common dependencies
2
-r common.txt
3
4
-numba == 0.60.0 # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
4
+numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
5
+numba == 0.61; python_version > '3.9'
6
7
# Dependencies for NVIDIA GPUs
8
ray[cgraph]>=2.43.0, !=2.44.* # Ray Compiled Graph, required for pipeline parallelism in V1.
requirements/rocm.txt
+2
-1
@@ -1,7 +1,8 @@
1
# Common dependencies
2
-r common.txt
3
4
-numba == 0.60.0 # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
4
+numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
5
+numba == 0.61; python_version > '3.9'
6
7
# Dependencies for AMD GPUs
8
awscli
requirements/test.in
+3
-1
@@ -38,7 +38,9 @@ buildkite-test-collector==0.1.9
38
genai_perf==0.0.8
39
tritonclient==2.51.0
40
41
-numpy < 2.0.0
41
+numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
42
+numba == 0.61; python_version > '3.9'
43
+numpy
44
runai-model-streamer==0.11.0
45
runai-model-streamer-s3==0.11.0
46
fastsafetensors>=0.1.10
requirements/test.txt
+5
-3
@@ -219,7 +219,7 @@ libnacl==2.1.0
219
# via tensorizer
220
librosa==0.10.2.post1
221
# via -r requirements/test.in
222
-llvmlite==0.43.0
222
+llvmlite==0.44.0
223
# via numba
224
lm-eval==0.4.4
225
# via -r requirements/test.in
@@ -262,8 +262,10 @@ networkx==3.2.1
262
# via torch
263
nltk==3.9.1
264
# via rouge-score
265
-numba==0.60.0
266
- # via librosa
265
+numba==0.61.0
266
+ # via
267
+ # -r requirements/test.in
268
+ # librosa
269
numexpr==2.10.1
270
# via lm-eval
271
numpy==1.26.4