We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent d39fa23 commit 450fc8fCopy full SHA for 450fc8f
.github/container/build-te.sh
@@ -126,8 +126,10 @@ print_var SRC_PATH_XLA
126
echo "=================================================="
127
128
# Parse SM_LIST into the format accepted by TransformerEngine's build system
129
-# "1.2,3.4,5.6" -> "12-real;34-real;56", i.e. SASS plus PTX for the last one
130
-NVTE_CUDA_ARCHS="${SM_LIST//,/-real;}"
+# "1.2,3.4,5.6" -> "12;34;56". In principle we would like to compile SASS-only
+# plus PTX for the highest known architecture, but TransformerEngine's build
131
+# system does not currently handle that.
132
+NVTE_CUDA_ARCHS="${SM_LIST//,/;}"
133
set -x
134
export NVTE_CUDA_ARCHS="${NVTE_CUDA_ARCHS//./}"
135
# Parallelism within nvcc invocations.
0 commit comments