diff --git a/.github/workflows/unit-tests.yaml b/.github/workflows/unit-tests.yaml
index e739be01..fae3a106 100644
--- a/.github/workflows/unit-tests.yaml
+++ b/.github/workflows/unit-tests.yaml
@@ -16,11 +16,11 @@ jobs:
         # 1.14.0a0+410ce96
         - torch: "1.14"
           nvcr: 22.12-py3
-          dir: ../torch1
+          dir: torch1
         # 2.1.0a0+32f93b1
         - torch: "2.1"
           nvcr: 23.10-py3
-          dir: ../torch2
+          dir: torch2
     container:
       image: nvcr.io/nvidia/pytorch:${{ matrix.nvcr }}
       options: --privileged --ipc=host --gpus=all
diff --git a/dockerfile/torch2.1-cuda12.2.dockerfile b/dockerfile/torch2.1-cuda12.2.dockerfile
index 4dee17bb..3fd0a0e8 100644
--- a/dockerfile/torch2.1-cuda12.2.dockerfile
+++ b/dockerfile/torch2.1-cuda12.2.dockerfile
@@ -44,6 +44,8 @@ WORKDIR /opt/msamp
 ADD third_party third_party
 RUN cd third_party/msccl && \
     make -j ${NUM_MAKE_JOBS} src.build NVCC_GENCODE="\
+    -gencode=arch=compute_70,code=sm_70 \
+    -gencode=arch=compute_80,code=sm_80 \
     -gencode=arch=compute_90,code=sm_90" && \
     make install
 # cache TE build to save time in CI