[Tensor parallel] Fix meta tensor split state propagation (#27574)

* ggml : fix meta tensor split state propagation

* Add test-llama-archs to CI
This commit is contained in:
Gaurav Garg
2026-08-23 18:49:12 +08:00
committed by GitHub
parent 8144f3192e
commit d3371929bb
4 changed files with 113 additions and 24 deletions
+29
View File
@@ -300,6 +300,31 @@ function gg_sum_ctest_release {
gg_printf '```\n'
}
# test_llama_archs_tensor_split
function gg_run_test_llama_archs_tensor_split {
cd ${SRC}
set -e
GGML_CUDA_DEVICES=1 ./build-ci-release/bin/test-llama-archs -s 1 2>&1
GGML_CUDA_DEVICES=2 ./build-ci-release/bin/test-llama-archs -s 1 2>&1
GGML_CUDA_DEVICES=3 ./build-ci-release/bin/test-llama-archs -s 1 2>&1
GGML_CUDA_DEVICES=4 ./build-ci-release/bin/test-llama-archs -s 1 2>&1
set +e
}
function gg_sum_test_llama_archs_tensor_split {
gg_printf '### %s\n\n' "${ci}"
gg_printf 'Runs test-llama-archs with 1 to 4 CUDA devices\n'
gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
gg_printf '```\n'
gg_printf '%s\n' "$(cat $OUT/${ci}.log)"
gg_printf '```\n'
}
# test_scripts
function gg_run_test_scripts {
@@ -751,6 +776,10 @@ ret=0
test $ret -eq 0 && gg_run ctest_debug
test $ret -eq 0 && gg_run ctest_release
if [ ! -z ${GG_BUILD_CUDA} ]; then
test $ret -eq 0 && gg_run test_llama_archs_tensor_split
fi
if [ ! -z ${GG_BUILD_HIGH_PERF} ]; then
test $ret -eq 0 && gg_run test_backend_ops_cpu
fi