Skip to content

Commit f0f1df3

Browse files
gs-olivegs-olive
committed
chore: Upgrade to CUDA 12.1 and CuDNN 8.9
- Upgrade CUDA version across CI, dependencies, and build systems - Upgrade versions for Torch, CuDNN, TensorRT, and other dependencies in accordance with new CUDA version 12.1 - Upgrade Torch nightly date to latest available
1 parent 1735088 commit f0f1df3

14 files changed

+73
-90
lines changed

‎.circleci/config.yml

+36-53
Original file line numberDiff line numberDiff line change
@@ -92,17 +92,9 @@ commands:
9292
key: cmake-<< parameters.cache-key >>-<< parameters.version >>
9393
paths:
9494
- << parameters.install-dir >>
95-
uninstall-cuda:
96-
description: "Uninstall CUDA-11.4"
97-
steps:
98-
- run:
99-
name: "Uninstall CUDA-11.4"
100-
command: |
101-
sudo apt-get --purge remove "cuda*" "nsight*"
102-
sudo apt-get --purge remove "*nvidia*"
10395

10496
install-cudnn:
105-
description: "Install CUDNN 8.8.0"
97+
description: "Install CUDNN 8.9.1"
10698
parameters:
10799
os:
108100
type: string
@@ -112,10 +104,10 @@ commands:
112104
default: "x86_64"
113105
cudnn-version:
114106
type: string
115-
default: "8.8.0.121"
107+
default: "8.9.1.23"
116108
cuda-version:
117109
type: string
118-
default: "cuda11.8"
110+
default: "cuda12.1"
119111
steps:
120112
- run:
121113
name: Install CUDNN
@@ -149,13 +141,13 @@ commands:
149141
default: "amd64"
150142
cuda-pkg-name:
151143
type: string
152-
default: "cuda-toolkit-11-8"
144+
default: "cuda-toolkit-12-1"
153145
cuda-pkg-version:
154146
type: string
155-
default: "11-8"
147+
default: "12-1"
156148
cuda-version:
157149
type: string
158-
default: "11.8.0"
150+
default: "12.1.1"
159151
steps:
160152
- run:
161153
name: Install CUDA
@@ -186,21 +178,12 @@ commands:
186178
architecture:
187179
type: string
188180
default: "amd64"
189-
cuda-pkg-name:
190-
type: string
191-
default: "cuda-toolkit-11-8"
192-
cuda-pkg-version:
193-
type: string
194-
default: "11-8"
195-
cuda-version:
196-
type: string
197-
default: "11.8.0"
198181
cuda-string-version:
199182
type: string
200-
default: "cuda11.8"
183+
default: "cuda12.0"
201184
cudnn-version:
202185
type: string
203-
default: "8.8.0.121"
186+
default: "8.9.1.23"
204187
trt-version-short:
205188
type: string
206189
default: "8.6.1"
@@ -252,7 +235,7 @@ commands:
252235
default: "8.6.1"
253236
cudnn-version-long:
254237
type: string
255-
default: "8.8.0.121"
238+
default: "8.9.1.23"
256239
steps:
257240
- run:
258241
name: Set up python environment
@@ -261,21 +244,21 @@ commands:
261244
pip3 install wheel setuptools
262245
pip3 install nvidia-pyindex
263246
pip3 install tabulate
264-
pip3 install tensorrt==<< parameters.trt-version-long >> nvidia-cudnn-cu11==<< parameters.cudnn-version-long >>
247+
pip3 install tensorrt==<< parameters.trt-version-long >> nvidia-cudnn-cu12==<< parameters.cudnn-version-long >>
265248
pip3 install pytest parameterized expecttest nox
266249
267250
install-torch-from-index:
268251
description: "Install Pytorch"
269252
parameters:
270253
torch-build:
271254
type: string
272-
default: "2.1.0.dev20230605+cu118"
255+
default: "2.1.0.dev20230613+cu121"
273256
torchvision-build:
274257
type: string
275-
default: "0.16.0.dev20230605+cu118"
258+
default: "0.16.0.dev20230613+cu121"
276259
torch-build-index:
277260
type: string
278-
default: "https://download.pytorch.org/whl/nightly/cu118"
261+
default: "https://download.pytorch.org/whl/nightly/cu121"
279262
steps:
280263
- run:
281264
name: Install Torch
@@ -293,7 +276,7 @@ commands:
293276
- run:
294277
name: Build torch-tensorrt python release (pre-cxx11-abi)
295278
command: |
296-
export CUDA_HOME=/usr/local/cuda-11.8/
279+
export CUDA_HOME=/usr/local/cuda-12.1/
297280
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
298281
cd py
299282
python3 -m pip install wheel setuptools
@@ -313,7 +296,7 @@ commands:
313296
- run:
314297
name: Build torch-tensorrt python legacy release (pre-cxx11-abi)
315298
command: |
316-
export CUDA_HOME=/usr/local/cuda-11.8/
299+
export CUDA_HOME=/usr/local/cuda-12.1/
317300
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
318301
cd py
319302
python3 -m pip install wheel setuptools
@@ -345,7 +328,7 @@ commands:
345328
- run:
346329
name: Build torch-tensorrt python release package
347330
command: |
348-
export CUDA_HOME=/usr/local/cuda-11.8/
331+
export CUDA_HOME=/usr/local/cuda-12.1/
349332
cd ~/project/py
350333
python3 setup.py bdist_wheel --use-cxx11-abi --release
351334
python3 setup.py install --use-cxx11-abi --release
@@ -357,7 +340,7 @@ commands:
357340
- run:
358341
name: Build torch-tensorrt python package
359342
command: |
360-
export CUDA_HOME=/usr/local/cuda-11.8/
343+
export CUDA_HOME=/usr/local/cuda-12.1/
361344
cd ~/project/py
362345
python3 setup.py bdist_wheel --use-cxx11-abi
363346
python3 setup.py install --use-cxx11-abi
@@ -375,7 +358,7 @@ commands:
375358
- run:
376359
name: Build torch-tensorrt python release with only the fx backend
377360
command: |
378-
export CUDA_HOME=/usr/local/cuda-11.8/
361+
export CUDA_HOME=/usr/local/cuda-12.1/
379362
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
380363
cd py
381364
python3 -m pip install wheel setuptools
@@ -447,7 +430,7 @@ commands:
447430
name: Build torch-tensorrt library with CMake
448431
command: |
449432
mkdir build
450-
export PATH=$PATH:/usr/local/cuda-11.8/bin
433+
export PATH=$PATH:/usr/local/cuda-12.1/bin
451434
~/cmake/bin/cmake -S. -Bbuild \
452435
-DCMAKE_MODULE_PATH=cmake/Module \
453436
-DTorch_DIR=/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages/torch/share/cmake/Torch \
@@ -499,7 +482,7 @@ commands:
499482
name: Run core / C++ tests
500483
no_output_timeout: 15m
501484
environment:
502-
LD_LIBRARY_PATH: "/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages/torch_tensorrt.libs:/home/circleci/project/bazel-project/external/libtorch_pre_cxx11_abi/lib/:/home/circleci/project/bazel-project/external/tensorrt/lib/:/usr/local/cuda-11.8/lib64/:$LD_LIBRARY_PATH"
485+
LD_LIBRARY_PATH: "/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages/torch_tensorrt.libs:/home/circleci/project/bazel-project/external/libtorch_pre_cxx11_abi/lib/:/home/circleci/project/bazel-project/external/tensorrt/lib/:/usr/local/cuda-12.1/lib64/:$LD_LIBRARY_PATH"
503486
command: |
504487
set -e
505488
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
@@ -529,7 +512,7 @@ commands:
529512
environment:
530513
USE_HOST_DEPS: "1"
531514
PYT_PATH: "/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages/"
532-
LD_LIBRARY_PATH: "/usr/lib/x86_64-linux-gnu/:/usr/local/cuda-11.8/lib64/:$LD_LIBRARY_PATH"
515+
LD_LIBRARY_PATH: "/usr/lib/x86_64-linux-gnu/:/usr/local/cuda-12.1/lib64/:$LD_LIBRARY_PATH"
533516
command: |
534517
set -e
535518
mkdir -p /tmp/artifacts/test_results
@@ -819,7 +802,7 @@ jobs:
819802
type: boolean
820803
default: false
821804
machine:
822-
image: linux-cuda-11:2023.02.1
805+
image: linux-cuda-12:2023.05.1
823806
resource_class: gpu.nvidia.small
824807
steps:
825808
- checkout
@@ -881,7 +864,7 @@ jobs:
881864
cudnn-version:
882865
type: string
883866
machine:
884-
image: linux-cuda-11:2023.02.1
867+
image: linux-cuda-12:2023.05.1
885868
resource_class: gpu.nvidia.large
886869
parallelism: 4
887870
steps:
@@ -922,7 +905,7 @@ jobs:
922905
python-version:
923906
type: string
924907
machine:
925-
image: linux-cuda-11:2023.02.1
908+
image: linux-cuda-12:2023.05.1
926909
resource_class: gpu.nvidia.large
927910
steps:
928911
- checkout
@@ -956,7 +939,7 @@ jobs:
956939
type: string
957940
parallelism: 8
958941
machine:
959-
image: linux-cuda-11:2023.02.1
942+
image: linux-cuda-12:2023.05.1
960943
resource_class: gpu.nvidia.large
961944
steps:
962945
- checkout
@@ -992,7 +975,7 @@ jobs:
992975
type: string
993976
parallelism: 8
994977
machine:
995-
image: linux-cuda-11:2023.02.1
978+
image: linux-cuda-12:2023.05.1
996979
resource_class: gpu.nvidia.large
997980
steps:
998981
- checkout
@@ -1030,7 +1013,7 @@ jobs:
10301013
python-version:
10311014
type: string
10321015
machine:
1033-
image: linux-cuda-11:2023.02.1
1016+
image: linux-cuda-12:2023.05.1
10341017
resource_class: gpu.nvidia.large
10351018
steps:
10361019
- checkout
@@ -1068,7 +1051,7 @@ jobs:
10681051
type: string
10691052
parallelism: 4
10701053
machine:
1071-
image: linux-cuda-11:2023.02.1
1054+
image: linux-cuda-12:2023.05.1
10721055
resource_class: gpu.nvidia.small
10731056
steps:
10741057
- when:
@@ -1116,7 +1099,7 @@ jobs:
11161099
torch-build-index:
11171100
type: string
11181101
machine:
1119-
image: linux-cuda-11:2023.02.1
1102+
image: linux-cuda-12:2023.05.1
11201103
resource_class: gpu.nvidia.small
11211104
steps:
11221105
- when:
@@ -1252,7 +1235,7 @@ jobs:
12521235
python-version:
12531236
type: string
12541237
machine:
1255-
image: linux-cuda-11:2023.02.1
1238+
image: linux-cuda-12:2023.05.1
12561239
resource_class: gpu.nvidia.small
12571240
steps:
12581241
- checkout
@@ -1288,7 +1271,7 @@ jobs:
12881271
torch-base-image:
12891272
type: string
12901273
machine:
1291-
image: linux-cuda-11:2023.02.1
1274+
image: linux-cuda-12:2023.05.1
12921275
resource_class: gpu.nvidia.small
12931276
steps:
12941277
- checkout
@@ -1316,7 +1299,7 @@ jobs:
13161299
torch-base-image:
13171300
type: string
13181301
machine:
1319-
image: linux-cuda-11:2023.02.1
1302+
image: linux-cuda-12:2023.05.1
13201303
resource_class: gpu.nvidia.small
13211304
steps:
13221305
- when:
@@ -1352,13 +1335,13 @@ parameters:
13521335
# Nightly platform config
13531336
torch-build:
13541337
type: string
1355-
default: "2.1.0.dev20230605+cu118"
1338+
default: "2.1.0.dev20230613+cu121"
13561339
torchvision-build:
13571340
type: string
1358-
default: "0.16.0.dev20230605+cu118"
1341+
default: "0.16.0.dev20230613+cu121"
13591342
torch-build-index:
13601343
type: string
1361-
default: "https://download.pytorch.org/whl/nightly/cu118"
1344+
default: "https://download.pytorch.org/whl/nightly/cu121"
13621345
torch-build-legacy:
13631346
type: string
13641347
default: "1.13.1+cu117"
@@ -1373,7 +1356,7 @@ parameters:
13731356
default: true
13741357
cudnn-version:
13751358
type: string
1376-
default: "8.8.0.121"
1359+
default: "8.9.1.23"
13771360
trt-version-short:
13781361
type: string
13791362
default: "8.6.1"

‎README.md

+3-3
Original file line numberDiff line numberDiff line change
@@ -116,9 +116,9 @@ torch.jit.save(trt_ts_module, "trt_torchscript_module.ts") # save the TRT embedd
116116
These are the following dependencies used to verify the testcases. Torch-TensorRT can work with other versions, but the tests are not guaranteed to pass.
117117

118118
- Bazel 5.2.0
119-
- Libtorch 2.1.0.dev20230605 (built with CUDA 11.8)
120-
- CUDA 11.8
121-
- cuDNN 8.8.0
119+
- Libtorch 2.1.0.dev20230613 (built with CUDA 12.1)
120+
- CUDA 12.1
121+
- cuDNN 8.9.1
122122
- TensorRT 8.6.1
123123

124124
## Prebuilt Binaries and Wheel files

‎WORKSPACE

+10-10
Original file line numberDiff line numberDiff line change
@@ -41,7 +41,7 @@ local_repository(
4141
new_local_repository(
4242
name = "cuda",
4343
build_file = "@//third_party/cuda:BUILD",
44-
path = "/usr/local/cuda-11.8/",
44+
path = "/usr/local/cuda-12.1/",
4545
)
4646

4747
#############################################################################################################
@@ -51,17 +51,17 @@ new_local_repository(
5151
http_archive(
5252
name = "libtorch",
5353
build_file = "@//third_party/libtorch:BUILD",
54-
sha256 = "999becce82b73e566d0ffe010cd21fea8cf3a33f90f09dcc6b01150b820ae063",
54+
sha256 = "1c3712b3b1de34e9989549f53675b557f6f0ca0b800ccbbc80c941af68abcc65",
5555
strip_prefix = "libtorch",
56-
urls = ["https://download.pytorch.org/libtorch/nightly/cu118/libtorch-cxx11-abi-shared-with-deps-2.1.0.dev20230605%2Bcu118.zip"],
56+
urls = ["https://download.pytorch.org/libtorch/nightly/cu121/libtorch-cxx11-abi-shared-with-deps-2.1.0.dev20230613%2Bcu121.zip"],
5757
)
5858

5959
http_archive(
6060
name = "libtorch_pre_cxx11_abi",
6161
build_file = "@//third_party/libtorch:BUILD",
62-
sha256 = "786cc728c63ea69c40bd8fb535cf8e5e1dfff1d43eaad3eb5256b9ed89c1b268",
62+
sha256 = "6bbb53f2f9533804175a0d2eeae5093ee1907158a9ec75c7fb9d10e2103d5df5",
6363
strip_prefix = "libtorch",
64-
urls = ["https://download.pytorch.org/libtorch/nightly/cu118/libtorch-shared-with-deps-2.1.0.dev20230605%2Bcu118.zip"],
64+
urls = ["https://download.pytorch.org/libtorch/nightly/cu121/libtorch-shared-with-deps-2.1.0.dev20230613%2Bcu121.zip"],
6565
)
6666

6767
# Download these tarballs manually from the NVIDIA website
@@ -71,20 +71,20 @@ http_archive(
7171
http_archive(
7272
name = "cudnn",
7373
build_file = "@//third_party/cudnn/archive:BUILD",
74-
sha256 = "36fff137153ef73e6ee10bfb07f4381240a86fb9fb78ce372414b528cbab2293",
75-
strip_prefix = "cudnn-linux-x86_64-8.8.0.121_cuda11-archive",
74+
sha256 = "35163c5c542be0c511738b27e25235193cbeedc5e0e006e44b1cdeaf1922e83e",
75+
strip_prefix = "cudnn-linux-x86_64-8.9.1.23_cuda12-archive",
7676
urls = [
77-
"https://developer.download.nvidia.com/compute/cudnn/secure/8.8.0/local_installers/11.8/cudnn-linux-x86_64-8.8.0.121_cuda11-archive.tar.xz",
77+
"https://developer.nvidia.com/downloads/compute/cudnn/secure/8.9.1/local_installers/12.x/cudnn-linux-x86_64-8.9.1.23_cuda12-archive.tar.xz",
7878
],
7979
)
8080

8181
http_archive(
8282
name = "tensorrt",
8383
build_file = "@//third_party/tensorrt/archive:BUILD",
84-
sha256 = "15bfe6053d45feec45ecc7123a9106076b0b43fa0435f242d89dca0778337759",
84+
sha256 = "0f8157a5fc5329943b338b893591373350afa90ca81239cdadd7580cd1eba254",
8585
strip_prefix = "TensorRT-8.6.1.6",
8686
urls = [
87-
"https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.1/tars/TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-11.8.tar.gz",
87+
"https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.1/tars/TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0.tar.gz",
8888
],
8989
)
9090

‎docker/Dockerfile

+2-2
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
# Base image starts with CUDA
2-
ARG BASE_IMG=nvidia/cuda:11.8.0-devel-ubuntu22.04
2+
ARG BASE_IMG=nvidia/cuda:12.1.1-devel-ubuntu22.04
33
FROM ${BASE_IMG} as base
4-
ENV BASE_IMG=nvidia/cuda:11.8.0-devel-ubuntu22.04
4+
ENV BASE_IMG=nvidia/cuda:12.1.1-devel-ubuntu22.04
55

66
ARG TENSORRT_VERSION
77
ENV TENSORRT_VERSION=${TENSORRT_VERSION}

0 commit comments

Comments
 (0)