[CI] Fix CI (python and cuda versions) (#1621)

pytorch · Oct 12, 2023 · 38dfc21 · 38dfc21
1 parent 70c650e
commit 38dfc21
Show file tree

Hide file tree

Showing 32 changed files with 254 additions and 96 deletions.
diff --git a/.github/unittest/linux_libs/scripts_brax/install.sh b/.github/unittest/linux_libs/scripts_brax/install.sh
@@ -32,7 +32,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall  --progress-bar off
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall  --progress-bar off
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall  --progress-bar off
 fi
 
 # install tensordict
@@ -42,7 +42,7 @@ pip install git+https://github.com/pytorch/tensordict.git --progress-bar off
 python -c "import functorch;import tensordict"
 
 printf "* Installing torchrl\n"
-pip3 install -e .
+python setup.py develop
 
 # smoke test
 python -c "import torchrl"
diff --git a/.github/unittest/linux_libs/scripts_d4rl/install.sh b/.github/unittest/linux_libs/scripts_d4rl/install.sh
@@ -35,7 +35,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict
@@ -45,7 +45,7 @@ pip install git+https://github.com/pytorch/tensordict.git
 python -c "import functorch;import tensordict"
 
 printf "* Installing torchrl\n"
-pip3 install -e .
+python setup.py develop
 
 # smoke test
 python -c "import torchrl"
diff --git a/.github/unittest/linux_libs/scripts_habitat/install.sh b/.github/unittest/linux_libs/scripts_habitat/install.sh
@@ -20,7 +20,7 @@ version="$(python -c "print('.'.join(\"${CUDA_VERSION}\".split('.')[:2]))")"
 git submodule sync && git submodule update --init --recursive
 
 printf "Installing PyTorch with %s\n" "${CU_VERSION}"
-pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 
 # install tensordict
 pip3 install git+https://github.com/pytorch/tensordict.git
@@ -29,7 +29,7 @@ pip3 install git+https://github.com/pytorch/tensordict.git
 python3 -c "import functorch;import tensordict"
 
 printf "* Installing torchrl\n"
-pip3 install -e .
+python setup.py develop
 
 # smoke test
 python3 -c "import torchrl"
diff --git a/.github/unittest/linux_libs/scripts_jumanji/install.sh b/.github/unittest/linux_libs/scripts_jumanji/install.sh
@@ -32,7 +32,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict
@@ -42,7 +42,7 @@ pip install git+https://github.com/pytorch/tensordict.git
 python -c "import functorch;import tensordict"
 
 printf "* Installing torchrl\n"
-pip3 install -e .
+python setup.py develop
 
 # smoke test
 python -c "import torchrl"
diff --git a/.github/unittest/linux_libs/scripts_pettingzoo/install.sh b/.github/unittest/linux_libs/scripts_pettingzoo/install.sh
@@ -32,7 +32,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict

diff --git a/.github/unittest/linux_libs/scripts_rlhf/install.sh b/.github/unittest/linux_libs/scripts_rlhf/install.sh
@@ -35,7 +35,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict
@@ -45,7 +45,7 @@ pip install git+https://github.com/pytorch/tensordict.git
 python -c "import tensordict"
 
 printf "* Installing torchrl\n"
-pip3 install -e .
+python setup.py develop
 
 # smoke test
 python -c "import torchrl"
diff --git a/.github/unittest/linux_libs/scripts_robohive/install_and_run_test.sh b/.github/unittest/linux_libs/scripts_robohive/install_and_run_test.sh
@@ -43,7 +43,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict

diff --git a/.github/unittest/linux_libs/scripts_sklearn/install.sh b/.github/unittest/linux_libs/scripts_sklearn/install.sh
@@ -35,7 +35,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict
@@ -45,7 +45,7 @@ pip install git+https://github.com/pytorch/tensordict.git
 python -c "import functorch;import tensordict"
 
 printf "* Installing torchrl\n"
-pip3 install -e .
+python setup.py develop
 
 # smoke test
 python -c "import torchrl"
diff --git a/.github/unittest/linux_libs/scripts_smacv2/install.sh b/.github/unittest/linux_libs/scripts_smacv2/install.sh
@@ -32,7 +32,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict

diff --git a/.github/unittest/linux_libs/scripts_vmas/install.sh b/.github/unittest/linux_libs/scripts_vmas/install.sh
@@ -32,7 +32,7 @@ if [ "${CU_VERSION:-}" == cpu ] ; then
 #    conda install -y pytorch cpuonly -c pytorch-nightly
     pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu --force-reinstall
 else
-    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu116 --force-reinstall
+    pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 --force-reinstall
 fi
 
 # install tensordict

diff --git a/.github/workflows/test-linux-brax.yml b/.github/workflows/test-linux-brax.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -27,8 +31,8 @@ jobs:
       script: |
         set -euo pipefail
 
-        export PYTHON_VERSION="3.8"
-        export CU_VERSION="11.7"
+        export PYTHON_VERSION="3.9"
+        export CU_VERSION="12.1"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"
         export TF_CPP_MIN_LOG_LEVEL=0

diff --git a/.github/workflows/test-linux-d4rl.yml b/.github/workflows/test-linux-d4rl.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -25,7 +29,7 @@ jobs:
       timeout: 120
       script: |
         set -euo pipefail
-        export PYTHON_VERSION="3.8"
+        export PYTHON_VERSION="3.9"
         export CU_VERSION="cu117"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"

diff --git a/.github/workflows/test-linux-envpool.yml b/.github/workflows/test-linux-envpool.yml
@@ -11,6 +11,10 @@ on:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -22,8 +26,8 @@ jobs:
       timeout: 120
       script: |
         set -euo pipefail
-        export PYTHON_VERSION="3.8"
-        export CU_VERSION="11.7"
+        export PYTHON_VERSION="3.9"
+        export CU_VERSION="12.1"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"
         export TF_CPP_MIN_LOG_LEVEL=0

diff --git a/.github/workflows/test-linux-examples.yml b/.github/workflows/test-linux-examples.yml
@@ -22,8 +22,8 @@ jobs:
   tests:
     strategy:
       matrix:
-        python_version: ["3.9"] # "3.8", "3.9", "3.10", "3.11"
-        cuda_arch_version: ["11.6"] # "11.6", "11.7"
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
       fail-fast: false
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
@@ -36,11 +36,8 @@ jobs:
       script: |
         # Set env vars from matrix
         export PYTHON_VERSION=${{ matrix.python_version }}
-        # Commenting these out for now because the GPU test are not working inside docker
         export CUDA_ARCH_VERSION=${{ matrix.cuda_arch_version }}
         export CU_VERSION="cu${CUDA_ARCH_VERSION:0:2}${CUDA_ARCH_VERSION:3:1}"
-        # Remove the following line when the GPU tests are working inside docker, and uncomment the above lines
-        #export CU_VERSION="cpu"
 
         echo "PYTHON_VERSION: $PYTHON_VERSION"
         echo "CU_VERSION: $CU_VERSION"

diff --git a/.github/workflows/test-linux-gym.yml b/.github/workflows/test-linux-gym.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -27,7 +31,7 @@ jobs:
       timeout: 120
       script: |
         set -euxo pipefail
-        export PYTHON_VERSION="3.8"
+        export PYTHON_VERSION="3.9"
         # export CU_VERSION="${{ inputs.gpu-arch-version }}"
         export CU_VERSION="11.4"
         export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/work/mujoco-py/mujoco_py/binaries/linux/mujoco210/bin"

diff --git a/.github/workflows/test-linux-jumanji.yml b/.github/workflows/test-linux-jumanji.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -27,7 +31,7 @@ jobs:
       script: |
         set -euo pipefail
         export PYTHON_VERSION="3.9"
-        export CU_VERSION="11.7"
+        export CU_VERSION="12.1"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"
         export TF_CPP_MIN_LOG_LEVEL=0

diff --git a/.github/workflows/test-linux-olddeps.yml b/.github/workflows/test-linux-olddeps.yml
@@ -11,6 +11,10 @@ on:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.8"]
+        cuda_arch_version: ["11.6"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl

diff --git a/.github/workflows/test-linux-pettingzoo.yml b/.github/workflows/test-linux-pettingzoo.yml
@@ -27,7 +27,7 @@ jobs:
       script: |
         set -euo pipefail
         export PYTHON_VERSION="3.9"
-        export CU_VERSION="11.7"
+        export CU_VERSION="12.1"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"
         export TF_CPP_MIN_LOG_LEVEL=0

diff --git a/.github/workflows/test-linux-rlhf.yml b/.github/workflows/test-linux-rlhf.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -27,7 +31,7 @@ jobs:
       timeout: 120
       script: |
         set -euo pipefail
-        export PYTHON_VERSION="3.8"
+        export PYTHON_VERSION="3.9"
         export CU_VERSION="cu117"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"

diff --git a/.github/workflows/test-linux-robohive.yml b/.github/workflows/test-linux-robohive.yml
@@ -11,6 +11,10 @@ on:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -19,7 +23,7 @@ jobs:
       timeout: 120
       script: |
         set -euo pipefail
-        export PYTHON_VERSION="3.8"
+        export PYTHON_VERSION="3.9"
         export CU_VERSION="cu117"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"

diff --git a/.github/workflows/test-linux-sklearn.yml b/.github/workflows/test-linux-sklearn.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -27,7 +31,7 @@ jobs:
       timeout: 120
       script: |
         set -euo pipefail
-        export PYTHON_VERSION="3.8"
+        export PYTHON_VERSION="3.9"
         export CU_VERSION="cu117"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"

diff --git a/.github/workflows/test-linux-smacv2.yml b/.github/workflows/test-linux-smacv2.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     if: ${{ github.event_name == 'push' || contains(github.event.pull_request.labels.*.name, 'Environments') }}
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
@@ -28,7 +32,7 @@ jobs:
       script: |
         set -euo pipefail
         export PYTHON_VERSION="3.9"
-        export CU_VERSION="11.7"
+        export CU_VERSION="12.1"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"
         export TF_CPP_MIN_LOG_LEVEL=0

diff --git a/.github/workflows/test-linux-vmas.yml b/.github/workflows/test-linux-vmas.yml
@@ -17,6 +17,10 @@ concurrency:
 
 jobs:
   unittests:
+    strategy:
+      matrix:
+        python_version: ["3.9"]
+        cuda_arch_version: ["12.1"]
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
       repository: pytorch/rl
@@ -27,7 +31,7 @@ jobs:
       script: |
         set -euo pipefail
         export PYTHON_VERSION="3.9"
-        export CU_VERSION="11.7"
+        export CU_VERSION="12.1"
         export TAR_OPTIONS="--no-same-owner"
         export UPLOAD_CHANNEL="nightly"
         export TF_CPP_MIN_LOG_LEVEL=0

diff --git a/.github/workflows/test-macos-cpu.yml b/.github/workflows/test-macos-cpu.yml
@@ -22,7 +22,7 @@ jobs:
   tests:
     strategy:
       matrix:
-        python_version: ["3.8", "3.9", "3.10", "3.11"]
+        python_version: ["3.8", "3.11"]
       fail-fast: false
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     with: