Upgrade to use CUDA 11.8

CONTRIBUTING.md

@@ -20,9 +20,9 @@ Your contributions will fall into two categories:
 
          For GPU
            ```bash
-           python ts_scripts/install_dependencies.py --environment=dev --cuda=cu102
+           python ts_scripts/install_dependencies.py --environment=dev --cuda=cu118
            ```
-            > Supported cuda versions as cu117, cu116, cu113, cu111, cu102, cu101, cu92
+            > Supported cuda versions as cu118, cu117, cu116, cu113, cu111, cu102, cu101, cu92
         - Install `pre-commit` to your Git flow:
             ```bash
             pre-commit install

benchmarks/auto_benchmark.py

@@ -149,7 +149,7 @@ def install_torchserve(skip_ts_install, hw, ts_version):
 
     # install_dependencies.py
     if hw == "gpu":
-        cmd = "python ts_scripts/install_dependencies.py --environment dev --cuda cu117"
+        cmd = "python ts_scripts/install_dependencies.py --environment dev --cuda cu118"
     elif hw == "neuronx":
         cmd = "python ts_scripts/install_dependencies.py --environment dev --neuronx"
     else:

docker/README.md

@@ -52,7 +52,7 @@ Creates a docker image with publicly available `torchserve` and `torch-model-arc
 ./build_image.sh
 ```
 
- - To create a GPU based image with cuda 10.2. Options are `cu92`, `cu101`, `cu102`, `cu111`, `cu113`, `cu116`, `cu117`
+ - To create a GPU based image with cuda 10.2. Options are `cu92`, `cu101`, `cu102`, `cu111`, `cu113`, `cu116`, `cu117`, `cu118`
 
   ```bash
   ./build_image.sh -g -cv cu102

docker/build_image.sh

@@ -43,7 +43,7 @@ do
         -g|--gpu)
           MACHINE=gpu
           DOCKER_TAG="pytorch/torchserve:latest-gpu"
-          BASE_IMAGE="nvidia/cuda:11.7.1-base-ubuntu20.04"
+          BASE_IMAGE="nvidia/cuda:11.8.0-base-ubuntu20.04"
           CUDA_VERSION="cu117"
           shift
           ;;

docker/build_upload_release.py

@@ -28,7 +28,7 @@
     # Upload pytorch/torchserve docker binaries
     try_and_handle(f"./build_image.sh -t {organization}/torchserve:latest", dry_run)
     try_and_handle(
-        f"./build_image.sh -g -cv cu117 -t {organization}/torchserve:latest-gpu",
+        f"./build_image.sh -g -cv cu118 -t {organization}/torchserve:latest-gpu",
         dry_run,
     )
     try_and_handle(

docker/docker_nightly.py

@@ -34,7 +34,7 @@
     # Build Nightly images and append the date in the name
     try_and_handle(f"./build_image.sh -t {organization}/{cpu_version}", dry_run)
     try_and_handle(
-        f"./build_image.sh -g -cv cu117 -t {organization}/{gpu_version}",
+        f"./build_image.sh -g -cv cu118 -t {organization}/{gpu_version}",
         dry_run,
     )
 

examples/pt2/README.md

@@ -9,7 +9,7 @@ We strongly recommend you leverage newer hardware so for GPUs that would be an A
 Install torchserve and ensure that you're using at least `torch>=2.0.0`
 
 ```sh
-python ts_scripts/install_dependencies.py --cuda=cu117
+python ts_scripts/install_dependencies.py --cuda=cu118
 pip install torchserve torch-model-archiver
 ```