Skip to content

Commit 7338f9e

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 7338f9e

File tree

13 files changed

+74
-85
lines changed

13 files changed

+74
-85
lines changed

‎.circleci/config.yml‎

Lines changed: 38 additions & 49 deletions
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
@@ -188,19 +180,16 @@ commands:
188180
default: "amd64"
189181
cuda-pkg-name:
190182
type: string
191-
default: "cuda-toolkit-11-8"
183+
default: "cuda-toolkit-12-1"
192184
cuda-pkg-version:
193185
type: string
194-
default: "11-8"
195-
cuda-version:
196-
type: string
197-
default: "11.8.0"
186+
default: "12-0"
198187
cuda-string-version:
199188
type: string
200-
default: "cuda11.8"
189+
default: "cuda12.1"
201190
cudnn-version:
202191
type: string
203-
default: "8.8.0.121"
192+
default: "8.9.1.23"
204193
trt-version-short:
205194
type: string
206195
default: "8.6.1"
@@ -252,7 +241,7 @@ commands:
252241
default: "8.6.1"
253242
cudnn-version-long:
254243
type: string
255-
default: "8.8.0.121"
244+
default: "8.9.1.23"
256245
steps:
257246
- run:
258247
name: Set up python environment
@@ -261,21 +250,21 @@ commands:
261250
pip3 install wheel setuptools
262251
pip3 install nvidia-pyindex
263252
pip3 install tabulate
264-
pip3 install tensorrt==<< parameters.trt-version-long >> nvidia-cudnn-cu11==<< parameters.cudnn-version-long >>
253+
pip3 install tensorrt==<< parameters.trt-version-long >> nvidia-cudnn-cu12==<< parameters.cudnn-version-long >>
265254
pip3 install pytest parameterized expecttest nox
266255
267256
install-torch-from-index:
268257
description: "Install Pytorch"
269258
parameters:
270259
torch-build:
271260
type: string
272-
default: "2.1.0.dev20230605+cu118"
261+
default: "2.1.0.dev20230613+cu121"
273262
torchvision-build:
274263
type: string
275-
default: "0.16.0.dev20230605+cu118"
264+
default: "0.16.0.dev20230613+cu121"
276265
torch-build-index:
277266
type: string
278-
default: "https://download.pytorch.org/whl/nightly/cu118"
267+
default: "https://download.pytorch.org/whl/nightly/cu121"
279268
steps:
280269
- run:
281270
name: Install Torch
@@ -293,7 +282,7 @@ commands:
293282
- run:
294283
name: Build torch-tensorrt python release (pre-cxx11-abi)
295284
command: |
296-
export CUDA_HOME=/usr/local/cuda-11.8/
285+
export CUDA_HOME=/usr/local/cuda-12.1/
297286
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
298287
cd py
299288
python3 -m pip install wheel setuptools
@@ -313,7 +302,7 @@ commands:
313302
- run:
314303
name: Build torch-tensorrt python legacy release (pre-cxx11-abi)
315304
command: |
316-
export CUDA_HOME=/usr/local/cuda-11.8/
305+
export CUDA_HOME=/usr/local/cuda-12.1/
317306
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
318307
cd py
319308
python3 -m pip install wheel setuptools
@@ -345,7 +334,7 @@ commands:
345334
- run:
346335
name: Build torch-tensorrt python release package
347336
command: |
348-
export CUDA_HOME=/usr/local/cuda-11.8/
337+
export CUDA_HOME=/usr/local/cuda-12.1/
349338
cd ~/project/py
350339
python3 setup.py bdist_wheel --use-cxx11-abi --release
351340
python3 setup.py install --use-cxx11-abi --release
@@ -357,7 +346,7 @@ commands:
357346
- run:
358347
name: Build torch-tensorrt python package
359348
command: |
360-
export CUDA_HOME=/usr/local/cuda-11.8/
349+
export CUDA_HOME=/usr/local/cuda-12.1/
361350
cd ~/project/py
362351
python3 setup.py bdist_wheel --use-cxx11-abi
363352
python3 setup.py install --use-cxx11-abi
@@ -375,7 +364,7 @@ commands:
375364
- run:
376365
name: Build torch-tensorrt python release with only the fx backend
377366
command: |
378-
export CUDA_HOME=/usr/local/cuda-11.8/
367+
export CUDA_HOME=/usr/local/cuda-12.1/
379368
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
380369
cd py
381370
python3 -m pip install wheel setuptools
@@ -447,7 +436,7 @@ commands:
447436
name: Build torch-tensorrt library with CMake
448437
command: |
449438
mkdir build
450-
export PATH=$PATH:/usr/local/cuda-11.8/bin
439+
export PATH=$PATH:/usr/local/cuda-12.1/bin
451440
~/cmake/bin/cmake -S. -Bbuild \
452441
-DCMAKE_MODULE_PATH=cmake/Module \
453442
-DTorch_DIR=/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages/torch/share/cmake/Torch \
@@ -499,7 +488,7 @@ commands:
499488
name: Run core / C++ tests
500489
no_output_timeout: 15m
501490
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/:$LD_LIBRARY_PATH"
491+
LD_LIBRARY_PATH: "/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages/torch_tensorrt.libs:/home/circleci/project/bazel-project/external/libtorch_pre_cxx11_abi/lib/:/home/circleci/project/bazel-project/external/tensorrt/lib/:/usr/local/cuda-12.1/lib64/:$LD_LIBRARY_PATH"
503492
command: |
504493
set -e
505494
mv toolchains/ci_workspaces/WORKSPACE.<< parameters.platform >> WORKSPACE
@@ -529,7 +518,7 @@ commands:
529518
environment:
530519
USE_HOST_DEPS: "1"
531520
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/:$LD_LIBRARY_PATH"
521+
LD_LIBRARY_PATH: "/usr/lib/x86_64-linux-gnu/:/usr/local/cuda-12.1/lib64/:$LD_LIBRARY_PATH"
533522
command: |
534523
set -e
535524
mkdir -p /tmp/artifacts/test_results
@@ -819,7 +808,7 @@ jobs:
819808
type: boolean
820809
default: false
821810
machine:
822-
image: linux-cuda-11:2023.02.1
811+
image: linux-cuda-12:2023.05.1
823812
resource_class: gpu.nvidia.small
824813
steps:
825814
- checkout
@@ -881,7 +870,7 @@ jobs:
881870
cudnn-version:
882871
type: string
883872
machine:
884-
image: linux-cuda-11:2023.02.1
873+
image: linux-cuda-12:2023.05.1
885874
resource_class: gpu.nvidia.large
886875
parallelism: 4
887876
steps:
@@ -922,7 +911,7 @@ jobs:
922911
python-version:
923912
type: string
924913
machine:
925-
image: linux-cuda-11:2023.02.1
914+
image: linux-cuda-12:2023.05.1
926915
resource_class: gpu.nvidia.large
927916
steps:
928917
- checkout
@@ -956,7 +945,7 @@ jobs:
956945
type: string
957946
parallelism: 8
958947
machine:
959-
image: linux-cuda-11:2023.02.1
948+
image: linux-cuda-12:2023.05.1
960949
resource_class: gpu.nvidia.large
961950
steps:
962951
- checkout
@@ -992,7 +981,7 @@ jobs:
992981
type: string
993982
parallelism: 8
994983
machine:
995-
image: linux-cuda-11:2023.02.1
984+
image: linux-cuda-12:2023.05.1
996985
resource_class: gpu.nvidia.large
997986
steps:
998987
- checkout
@@ -1030,7 +1019,7 @@ jobs:
10301019
python-version:
10311020
type: string
10321021
machine:
1033-
image: linux-cuda-11:2023.02.1
1022+
image: linux-cuda-12:2023.05.1
10341023
resource_class: gpu.nvidia.large
10351024
steps:
10361025
- checkout
@@ -1068,7 +1057,7 @@ jobs:
10681057
type: string
10691058
parallelism: 4
10701059
machine:
1071-
image: linux-cuda-11:2023.02.1
1060+
image: linux-cuda-12:2023.05.1
10721061
resource_class: gpu.nvidia.small
10731062
steps:
10741063
- when:
@@ -1116,7 +1105,7 @@ jobs:
11161105
torch-build-index:
11171106
type: string
11181107
machine:
1119-
image: linux-cuda-11:2023.02.1
1108+
image: linux-cuda-12:2023.05.1
11201109
resource_class: gpu.nvidia.small
11211110
steps:
11221111
- when:
@@ -1252,7 +1241,7 @@ jobs:
12521241
python-version:
12531242
type: string
12541243
machine:
1255-
image: linux-cuda-11:2023.02.1
1244+
image: linux-cuda-12:2023.05.1
12561245
resource_class: gpu.nvidia.small
12571246
steps:
12581247
- checkout
@@ -1288,7 +1277,7 @@ jobs:
12881277
torch-base-image:
12891278
type: string
12901279
machine:
1291-
image: linux-cuda-11:2023.02.1
1280+
image: linux-cuda-12:2023.05.1
12921281
resource_class: gpu.nvidia.small
12931282
steps:
12941283
- checkout
@@ -1316,7 +1305,7 @@ jobs:
13161305
torch-base-image:
13171306
type: string
13181307
machine:
1319-
image: linux-cuda-11:2023.02.1
1308+
image: linux-cuda-12:2023.05.1
13201309
resource_class: gpu.nvidia.small
13211310
steps:
13221311
- when:
@@ -1352,13 +1341,13 @@ parameters:
13521341
# Nightly platform config
13531342
torch-build:
13541343
type: string
1355-
default: "2.1.0.dev20230605+cu118"
1344+
default: "2.1.0.dev20230613+cu121"
13561345
torchvision-build:
13571346
type: string
1358-
default: "0.16.0.dev20230605+cu118"
1347+
default: "0.16.0.dev20230613+cu121"
13591348
torch-build-index:
13601349
type: string
1361-
default: "https://download.pytorch.org/whl/nightly/cu118"
1350+
default: "https://download.pytorch.org/whl/nightly/cu121"
13621351
torch-build-legacy:
13631352
type: string
13641353
default: "1.13.1+cu117"
@@ -1373,7 +1362,7 @@ parameters:
13731362
default: true
13741363
cudnn-version:
13751364
type: string
1376-
default: "8.8.0.121"
1365+
default: "8.9.1.23"
13771366
trt-version-short:
13781367
type: string
13791368
default: "8.6.1"

‎README.md‎

Lines changed: 3 additions & 3 deletions
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

Lines changed: 10 additions & 10 deletions
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

Lines changed: 2 additions & 2 deletions
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)