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 16, 08:15 UTC
[Bugfix] Upgrade to pytorch 2.5.1 (#10001)
Signed-off-by: Bill Nell <bill@neuralmagic.com>
bnellnm committed
Nov 4, 2024 at 18:11 UTC
d93478b399535d4b31e49d584d323172e6060653
7 files changed
+11
-11
CMakeLists.txt
+2
-2
@@ -49,8 +49,8 @@ set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx11
49
# requirements.txt files and should be kept consistent. The ROCm torch
50
# versions are derived from Dockerfile.rocm
51
#
52
-set(TORCH_SUPPORTED_VERSION_CUDA "2.5.0")
53
-set(TORCH_SUPPORTED_VERSION_ROCM "2.5.0")
52
+set(TORCH_SUPPORTED_VERSION_CUDA "2.5.1")
53
+set(TORCH_SUPPORTED_VERSION_ROCM "2.5.1")
54
55
#
56
# Try to find python package with an executable that exactly matches
pyproject.toml
+1
-1
@@ -6,7 +6,7 @@ requires = [
6
"packaging",
7
"setuptools>=61",
8
"setuptools-scm>=8.0",
9
- "torch == 2.5.0",
9
+ "torch == 2.5.1",
10
"wheel",
11
"jinja2",
12
]
requirements-build.txt
+1
-1
@@ -4,6 +4,6 @@ ninja
4
packaging
5
setuptools>=61
6
setuptools-scm>=8
7
-torch==2.5.0
7
+torch==2.5.1
8
wheel
9
jinja2
requirements-cuda.txt
+3
-3
@@ -4,7 +4,7 @@
4
# Dependencies for NVIDIA GPUs
5
ray >= 2.9
6
nvidia-ml-py >= 12.560.30 # for pynvml package
7
-torch == 2.5.0
7
+torch == 2.5.1
8
# These must be updated alongside torch
9
-torchvision == 0.20 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
10
-xformers == 0.0.28.post2; platform_system == 'Linux' and platform_machine == 'x86_64' # Requires PyTorch 2.5.0
9
+torchvision == 0.20.1 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
10
+xformers == 0.0.28.post3; platform_system == 'Linux' and platform_machine == 'x86_64' # Requires PyTorch 2.5.1
requirements-openvino.txt
+1
-1
@@ -1,7 +1,7 @@
1
# Common dependencies
2
-r requirements-common.txt
3
4
-torch == 2.5.0 # should be aligned with "common" vLLM torch version
4
+torch == 2.5.1 # should be aligned with "common" vLLM torch version
5
openvino >= 2024.4.0 # since 2024.4.0 both CPU and GPU support Paged Attention
6
7
optimum @ git+https://github.com/huggingface/optimum.git@main # latest optimum is used to support latest transformers version
requirements-test.in
+1
-1
@@ -18,7 +18,7 @@ ray[adag]==2.35
18
sentence-transformers # required for embedding
19
soundfile # required for audio test
20
timm # required for internvl test
21
-torch==2.5.0
21
+torch==2.5.1
22
transformers_stream_generator # required for qwen-vl test
23
matplotlib # required for qwen-vl test
24
datamodel_code_generator # required for minicpm3 test
requirements-test.txt
+2
-2
@@ -492,7 +492,7 @@ timm==1.0.11
492
# via -r requirements-test.in
493
tokenizers==0.20.1
494
# via transformers
495
-torch==2.5.0
495
+torch==2.5.1
496
# via
497
# -r requirements-test.in
498
# accelerate
@@ -503,7 +503,7 @@ torch==2.5.0
503
# tensorizer
504
# timm
505
# torchvision
506
-torchvision==0.20.0
506
+torchvision==0.20.1
507
# via timm
508
tqdm==4.66.6
509
# via