Skip to content

feat(build_deps): Remove unused cudnn for lite deployment #495

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

feat(build_deps): Remove unused cudnn for lite deployment #495

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a21175a
feat(build_deps): Remove unused cudnn for lite deployment
Lifann Jun 29, 2025
9afee0e
Remove cudnn from CI procedure
Lifann Jun 29, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 0 additions & 9 deletions .github/workflows/make_wheel_Linux_x86.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,40 +11,32 @@ export TF_NAME='tensorflow'
if [[ "$TF_VERSION" =~ ^2\.(16)\.[0-9]+$ ]] ; then
export BUILD_IMAGE="tfra/nosla-cuda12.3-cudnn8.9-ubuntu20.04-manylinux2014-python$PY_VERSION"
export TF_CUDA_VERSION="12.3"
export TF_CUDNN_VERSION="8.9"
export TF_USE_LEGACY_KERAS=1
elif [[ "$TF_VERSION" =~ ^2\.(15)\.[0-9]+$ ]] ; then
export BUILD_IMAGE="tfra/nosla-cuda12.2-cudnn8.9-ubuntu20.04-manylinux2014-python$PY_VERSION"
export TF_CUDA_VERSION="12.2"
export TF_CUDNN_VERSION="8.9"
elif [[ "$TF_VERSION" =~ ^2\.(14)\.[0-9]+$ ]] ; then
export BUILD_IMAGE="tfra/nosla-cuda11.8-cudnn8.7-ubuntu20.04-manylinux2014-python$PY_VERSION"
export TF_CUDA_VERSION="11.8"
export TF_CUDNN_VERSION="8.7"
elif [[ "$TF_VERSION" =~ ^2\.(12|13)\.[0-9]+$ ]] ; then
export BUILD_IMAGE="tfra/nosla-cuda11.8-cudnn8.6-ubuntu20.04-manylinux2014-python$PY_VERSION"
export TF_CUDA_VERSION="11.8"
export TF_CUDNN_VERSION="8.6"
elif [[ "$TF_VERSION" =~ ^2\.([6-9]|10|11)\.[0-9]+$ ]] ; then
export BUILD_IMAGE="tfra/nosla-cuda11.2-cudnn8-ubuntu20.04-manylinux2014-python$PY_VERSION"
export TF_CUDA_VERSION="11.2"
export TF_CUDNN_VERSION="8.1"
elif [ $TF_VERSION == "2.4.1" ] ; then
export BUILD_IMAGE='tfra/nosla-cuda11.0-cudnn8-ubuntu18.04-manylinux2010-multipython'
export TF_CUDA_VERSION="11.0"
export TF_CUDNN_VERSION="8.0"
elif [ $TF_VERSION == "1.15.2" ] ; then
export BUILD_IMAGE='tfra/nosla-cuda10.0-cudnn7-ubuntu16.04-manylinux2010-multipython'
export TF_CUDA_VERSION="10.0"
export TF_CUDNN_VERSION="7.6"
else
echo "TF_VERSION is invalid: $TF_VERSION!"
exit 1
fi

echo "BUILD_IMAGE is $BUILD_IMAGE"
echo "TF_CUDA_VERSION is $TF_CUDA_VERSION"
echo "TF_CUDNN_VERSION is $TF_CUDNN_VERSION"

if [ -z $HOROVOD_VERSION ] ; then
export HOROVOD_VERSION='0.28.1'
Expand All @@ -66,7 +58,6 @@ DOCKER_BUILDKIT=1 docker build --no-cache \
--build-arg TF_NAME \
--build-arg TF_NEED_CUDA \
--build-arg TF_CUDA_VERSION \
--build-arg TF_CUDNN_VERSION \
--build-arg HOROVOD_VERSION \
--build-arg BUILD_IMAGE \
--build-arg NIGHTLY_FLAG \
Expand Down
64 changes: 30 additions & 34 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,26 +104,24 @@ is compiled differently. A typical example of this would be `conda`-installed Te
#### Compatibility Matrix
*GPU is supported by version `0.2.0` and later.*

| TFRA | TensorFlow | Compiler | CUDA | CUDNN | Compute Capability | CPU |
|:------|:-----------|:-----------|:-----|:------|:-----------------------------|:---------|
| 0.8.0 | 2.16.2 | GCC 8.2.1 | 12.3 | 8.9 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 | x86 |
| 0.8.0 | 2.16.2 | Xcode 13.1 | - | - | - | Apple M1 |
| 0.7.0 | 2.15.1 | GCC 8.2.1 | 12.2 | 8.9 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 | x86 |
| 0.7.0 | 2.15.1 | Xcode 13.1 | - | - | - | Apple M1 |
| 0.6.0 | 2.8.3 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.6.0 | 2.6.0 | Xcode 13.1 | - | - | - | Apple M1 |
| 0.5.1 | 2.8.3 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.5.1 | 2.6.0 | Xcode 13.1 | - | - | - | Apple M1 |
| 0.5.0 | 2.8.3 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.5.0 | 2.6.0 | Xcode 13.1 | - | - | - | Apple M1 |
| 0.4.0 | 2.5.1 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.4.0 | 2.5.0 | Xcode 13.1 | - | - | - | Apple M1 |
| 0.3.1 | 2.5.1 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.2.0 | 2.4.1 | GCC 7.3.1 | 11.0 | 8.0 | 6.0, 6.1, 7.0, 7.5, 8.0 | x86 |
| 0.2.0 | 1.15.2 | GCC 7.3.1 | 10.0 | 7.6 | 6.0, 6.1, 7.0, 7.5 | x86 |
| 0.1.0 | 2.4.1 | GCC 7.3.1 | - | - | - | x86 |

Check [nvidia-support-matrix](https://docs.nvidia.com/deeplearning/cudnn/support-matrix/index.html) for more details.
| TFRA | TensorFlow | Compiler | CUDA | Compute Capability | CPU |
|:------|:-----------|:-----------|:-----|:-----------------------------|:---------|
| 0.8.0 | 2.16.2 | GCC 8.2.1 | 12.3 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 | x86 |
| 0.8.0 | 2.16.2 | Xcode 13.1 | - | - | Apple M1 |
| 0.7.0 | 2.15.1 | GCC 8.2.1 | 12.2 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 | x86 |
| 0.7.0 | 2.15.1 | Xcode 13.1 | - | - | Apple M1 |
| 0.6.0 | 2.8.3 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.6.0 | 2.6.0 | Xcode 13.1 | - | - | Apple M1 |
| 0.5.1 | 2.8.3 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.5.1 | 2.6.0 | Xcode 13.1 | - | - | Apple M1 |
| 0.5.0 | 2.8.3 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.5.0 | 2.6.0 | Xcode 13.1 | - | - | Apple M1 |
| 0.4.0 | 2.5.1 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.4.0 | 2.5.0 | Xcode 13.1 | - | - | Apple M1 |
| 0.3.1 | 2.5.1 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 | x86 |
| 0.2.0 | 2.4.1 | GCC 7.3.1 | 11.0 | 6.0, 6.1, 7.0, 7.5, 8.0 | x86 |
| 0.2.0 | 1.15.2 | GCC 7.3.1 | 10.0 | 6.0, 6.1, 7.0, 7.5 | x86 |
| 0.1.0 | 2.4.1 | GCC 7.3.1 | - | - | x86 |

**NOTICE**

Expand Down Expand Up @@ -178,9 +176,7 @@ export TF_VERSION="2.15.1" # "2.11.0" is well tested.
export PY_VERSION="3.9"
export TF_NEED_CUDA=1
export TF_CUDA_VERSION=12.2 # nvcc --version to check version
export TF_CUDNN_VERSION=8.9 # print("cuDNN version:", tf.sysconfig.get_build_info()["cudnn_version"])
export CUDA_TOOLKIT_PATH="/usr/local/cuda"
export CUDNN_INSTALL_PATH="/usr/lib/x86_64-linux-gnu"

python configure.py
```
Expand Down Expand Up @@ -284,18 +280,18 @@ sess_config.gpu_options.allow_growth = True
### With TensorFlow Serving

#### Compatibility Matrix
| TFRA | TensorFlow | Serving branch | Compiler | CUDA | CUDNN | Compute Capability |
|:------|:-----------|:---------------|:----------|:-----|:------|:-----------------------------|
| 0.8.0 | 2.16.2 | r2.16 | GCC 8.2.1 | 12.3 | 8.9 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 |
| 0.7.0 | 2.15.1 | r2.15 | GCC 8.2.1 | 12.2 | 8.9 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 |
| 0.6.0 | 2.8.3 | r2.8 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.5.1 | 2.8.3 | r2.8 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.5.0 | 2.8.3 | r2.8 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.4.0 | 2.5.1 | r2.5 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.3.1 | 2.5.1 | r2.5 | GCC 7.3.1 | 11.2 | 8.1 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.2.0 | 2.4.1 | r2.4 | GCC 7.3.1 | 11.0 | 8.0 | 6.0, 6.1, 7.0, 7.5, 8.0 |
| 0.2.0 | 1.15.2 | r1.15 | GCC 7.3.1 | 10.0 | 7.6 | 6.0, 6.1, 7.0, 7.5 |
| 0.1.0 | 2.4.1 | r2.4 | GCC 7.3.1 | - | - | - |
| TFRA | TensorFlow | Serving branch | Compiler | CUDA | Compute Capability |
|:------|:-----------|:---------------|:----------|:-----|:-----------------------------|
| 0.8.0 | 2.16.2 | r2.16 | GCC 8.2.1 | 12.3 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 |
| 0.7.0 | 2.15.1 | r2.15 | GCC 8.2.1 | 12.2 | 7.0, 7.5, 8.0, 8.6, 8.9, 9.0 |
| 0.6.0 | 2.8.3 | r2.8 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.5.1 | 2.8.3 | r2.8 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.5.0 | 2.8.3 | r2.8 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.4.0 | 2.5.1 | r2.5 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.3.1 | 2.5.1 | r2.5 | GCC 7.3.1 | 11.2 | 6.0, 6.1, 7.0, 7.5, 8.0, 8.6 |
| 0.2.0 | 2.4.1 | r2.4 | GCC 7.3.1 | 11.0 | 6.0, 6.1, 7.0, 7.5, 8.0 |
| 0.2.0 | 1.15.2 | r1.15 | GCC 7.3.1 | 10.0 | 6.0, 6.1, 7.0, 7.5 |
| 0.1.0 | 2.4.1 | r2.4 | GCC 7.3.1 | - | - |

Serving TFRA-enable models by custom ops in TensorFlow Serving.

Expand Down
22 changes: 0 additions & 22 deletions build_deps/toolchains/gpu/cuda/BUILD.tpl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -117,27 +117,6 @@ cc_library(
visibility = ["//visibility:public"],
)

cc_library(
name = "cudnn",
srcs = ["cuda/lib/%{cudnn_lib}"],
data = ["cuda/lib/%{cudnn_lib}"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)

cc_library(
name = "cudnn_header",
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
)

cc_library(
name = "cufft",
srcs = ["cuda/lib/%{cufft_lib}"],
Expand Down Expand Up @@ -169,7 +148,6 @@ cc_library(
":cublas",
":cuda_headers",
":cudart",
":cudnn",
":cufft",
":curand",
],
Expand Down
17 changes: 0 additions & 17 deletions build_deps/toolchains/gpu/cuda/BUILD.windows.tpl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,22 +91,6 @@ cc_import(
visibility = ["//visibility:public"],
)

cc_import(
name = "cudnn",
interface_library = "cuda/lib/%{cudnn_lib}",
system_provided = 1,
visibility = ["//visibility:public"],
)

cc_library(
name = "cudnn_header",
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
)

cc_import(
name = "cufft",
interface_library = "cuda/lib/%{cufft_lib}",
Expand All @@ -128,7 +112,6 @@ cc_library(
":cublas",
":cuda_headers",
":cudart",
":cudnn",
":cufft",
":curand",
],
Expand Down
1 change: 0 additions & 1 deletion build_deps/toolchains/gpu/cuda/cuda_config.h.tpl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,6 @@ limitations under the License.
#define TF_CUDA_CAPABILITIES %{cuda_compute_capabilities}

#define TF_CUDA_VERSION "%{cuda_version}"
#define TF_CUDNN_VERSION "%{cudnn_version}"

#define TF_CUDA_TOOLKIT_PATH "%{cuda_toolkit_path}"

Expand Down
Loading
Loading