fix: use nvidia-cutlass-dsl[cu13] extra for CUDA 13 libs

From 4.4 onwards, nvidia-cutlass-dsl requires the [cu13] extra to
install nvidia-cutlass-dsl-libs-cu13. Without it, CI fails.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: limin2021

requirements.txt

@@ -5,7 +5,7 @@ einops
 ninja
 numpy
 nvidia-cudnn-frontend>=1.13.0
-nvidia-cutlass-dsl>=4.4.2
+nvidia-cutlass-dsl[cu13]>=4.4.2
 nvidia-ml-py
 packaging>=24.2
 requests

scripts/setup_test_env.sh

@@ -29,7 +29,7 @@ if [ -n "${CUTLASS_DSL_VERSION:-}" ]; then
   echo "========================================"
   echo "Overriding nvidia-cutlass-dsl with version: ${CUTLASS_DSL_VERSION}"
   echo "========================================"
-  pip install --force-reinstall "nvidia-cutlass-dsl==${CUTLASS_DSL_VERSION}"
+  pip install --force-reinstall "nvidia-cutlass-dsl[cu13]==${CUTLASS_DSL_VERSION}"
   echo "nvidia-cutlass-dsl override complete."
   echo ""
 fi