Skip to content

Add test level 2 #376

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: dev
Choose a base branch
from
+62 −67
Open

Add test level 2 #376

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 20 additions & 0 deletions ci/_utils.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@ if [ -z "${TEST_SGPU}${TEST_MGPU}" ]; then
TEST_MGPU=1
fi

TEST_START_TS=`date +%s`

#To disable some logs trimming
export CI=1

Expand Down Expand Up @@ -233,3 +235,21 @@ configure_omp_threads() {
echo "Using OMP_NUM_THREADS=${OMP_NUM_THREADS}"
fi
}

time_elapsed() {
_arg=$1
date -d @$((`date +%s` - _arg)) +%${2:-T}
}

pytest_run() {
#args: tag1 tag2 level ...
check_level $3 || return
_test_variant_tag=`get_test_variant_tag $1 $2`
shift 3
_test_name_tag=`get_test_name_tag $1 $_test_variant_tag`
check_test_filter $_test_name_tag || return
_start_ts=`date +%s`
echo "Run [$_test_variant_tag] $@ at `time_elapsed $TEST_START_TS`"
pytest -v -rfEs `get_pytest_junitxml $_test_name_tag` $TEST_PYTEST_ARGS "$TEST_DIR/$@" || test_run_error "[$_test_variant_tag] $1"
echo "Done [$_test_variant_tag] $1 in `time_elapsed $_start_ts`"
}
52 changes: 19 additions & 33 deletions ci/jax.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,14 +26,7 @@ install_prerequisites() {
TEST_DIR=${TE_PATH}tests/jax

run() {
check_level $1 || return
shift
_test_variant_tag=`get_test_variant_tag $_fus_attn $_test_label`
_test_name_tag=`get_test_name_tag $1 $_test_variant_tag`
check_test_filter $_test_name_tag || return
echo "Run [$_test_variant_tag] $@"
pytest -v -rfEs `get_pytest_junitxml $_test_name_tag` "$TEST_DIR/$@" || test_run_error "[$_test_variant_tag] $1"
echo "Done [$_test_variant_tag] $1"
pytest_run $_fus_attn "" "$@"
}

run_default_fa() {
Expand All @@ -44,10 +37,7 @@ run_default_fa() {
}

run_lbl() {
_test_label="$1"
shift
run "$@"
_test_label=""
pytest_run $_fus_attn "$@"
}

run_default_fa_lbl() {
Expand All @@ -71,24 +61,16 @@ run_test_config() {

run_test_config_mgpu() {
echo ==== Run mGPU with Fused attention backend: $_fus_attn ====

_ver=$(pip show jaxlib | grep Version)
case "$_ver" in
*0.4.35*)
# Workaround for distributed tests hang with xla_flag
XLA_FLAGS="--xla_gpu_enable_nccl_comm_splitting=false" run 3 test_distributed_fused_attn.py -k 'not test_context_parallel_ring_attn'

# Test ring attention with xla_flag --xla_experimental_ignore_channel_id only
XLA_FLAGS="--xla_experimental_ignore_channel_id" run_lbl "parallel_ring" 3 test_distributed_fused_attn.py -k test_context_parallel_ring_attn
;;
*)
# Workaround for distributed tests hang with xla_flag
XLA_FLAGS="--xla_gpu_enable_nccl_comm_splitting=false" run 3 test_distributed_fused_attn.py
;;
esac

configure_omp_threads 8
if [ $_fus_attn = $_DEFAULT_FUSED_ATTN ]; then
_dfa_level=2
else
_dfa_level=3
fi
# Workaround for distributed tests hang with xla_flag
XLA_FLAGS="--xla_gpu_enable_nccl_comm_splitting=false" run $_dfa_level test_distributed_fused_attn.py
run_default_fa 3 test_distributed_layernorm.py
XLA_FLAGS="--xla_gpu_enable_nccl_comm_splitting=false" run_default_fa 3 test_distributed_layernorm_mlp.py
XLA_FLAGS="--xla_gpu_enable_nccl_comm_splitting=false" run_default_fa 2 test_distributed_layernorm_mlp.py
run_default_fa 3 test_distributed_softmax.py

run_default_fa 3 test_sanity_import.py
Expand All @@ -112,15 +94,19 @@ pip list | egrep "flax|fidle|jax|ml_dtypes|numpy|transformer_e|typing_ext"
for _fus_attn in auto ck aotriton; do
configure_fused_attn_env $_fus_attn || continue

#On basic (1) level tests are run with ck
#On full (3) level they are run with auto/aotriton
#On basic (1) level tests are run with auto
#On medium (2) level they are run with ck and aotriton
#On full (3) level they are run with auto and aotriton
#Do not use unfused becaue JAX tests either do not use FA or enforce it
if [ $TEST_LEVEL -ge 3 ]; then
_DEFAULT_FUSED_ATTN="auto"
test $_fus_attn = "ck" && continue
else
elif [ $TEST_LEVEL -ge 2 ]; then
_DEFAULT_FUSED_ATTN="ck"
test $_fus_attn != "ck" && continue
test $_fus_attn = "auto" && continue
else
_DEFAULT_FUSED_ATTN="auto"
test $_fus_attn != "auto" && continue
fi

if [ -n "$TEST_JOBS_MODE" ]; then
Expand Down
57 changes: 23 additions & 34 deletions ci/pytorch.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,20 +18,10 @@ install_prerequisites() {
script_error "Failed to install test prerequisites"
exit $rc
fi
NVTE_USE_ROCM=1 bash $TEST_DIR/custom_ort_ops/build.sh
}

run() {
check_level $1 || return
shift
_test_variant_tag=`get_test_variant_tag $_fus_attn $_test_label`
_test_name_tag=`get_test_name_tag $1 $_test_variant_tag`
check_test_filter $_test_name_tag || return
echo "Run [$_test_variant_tag] $@"
#: ${_WORKERS_COUNT:=1}
#_args=-n$_WORKERS_COUNT --max-worker-restart=$_WORKERS_COUNT
pytest -v -rfEs `get_pytest_junitxml $_test_name_tag` "$TEST_DIR/$@" || test_run_error "[$_test_variant_tag] $1"
echo "Done [$_test_variant_tag] $1"
pytest_run $_fus_attn "" "$@"
}

run_default_fa() {
Expand All @@ -44,10 +34,7 @@ run_default_fa() {

run_default_fa_lbl() {
if [ $_fus_attn = "$_DEFAULT_FUSED_ATTN" ]; then
_test_label="$1"
shift
run "$@"
_test_label=""
pytest_run $_fus_attn "$@"
fi
}

Expand All @@ -61,8 +48,8 @@ run_test_config(){
test $_fus_attn = auto -o $_fus_attn = ck -o $_fus_attn = aotriton && NVTE_FLASH_ATTN=0 run 1 test_cpu_offloading.py
run_default_fa 1 test_fused_rope.py
run_default_fa 1 test_fusible_ops.py
run_default_fa 3 test_gemm_autotune.py
run_default_fa 3 test_gemm_sm_count.py
run_default_fa 1 test_gemm_autotune.py
run_default_fa 1 test_gemm_sm_count.py
run 1 test_gqa.py
run 1 test_jit.py
run_default_fa 1 test_multi_tensor.py
Expand All @@ -78,24 +65,22 @@ run_test_config(){
run_default_fa 1 triton_kernels/test_norms.py
NVTE_TEST_TRITON_AUTOTUNE=1 run_default_fa_lbl "autotune" 3 triton_kernels/test_norms.py
run_default_fa 1 test_parallel_cross_entropy.py
NVTE_USE_DEQUANTIZE_TRITON=1 NVTE_USE_CAST_TRANSPOSE_TRITON=1 NVTE_USE_RMSNORM_TRITON=1 NVTE_USE_LAYERNORM_TRITON=1 run_default_fa_lbl "triton" 1 test_numerics.py
NVTE_USE_DEQUANTIZE_TRITON=1 NVTE_USE_CAST_TRANSPOSE_TRITON=1 NVTE_USE_RMSNORM_TRITON=1 NVTE_USE_LAYERNORM_TRITON=1 run_default_fa_lbl "triton" 3 test_numerics.py
NVTE_USE_RMSNORM_TRITON=1 run_default_fa_lbl "triton" 1 test_fusible_ops.py
NVTE_USE_CAST_TRANSPOSE_TRITON=1 run_default_fa_lbl "triton" 1 test_float8_current_scaling_exact.py
}

run_test_config_mgpu(){
#_WORKERS_COUNT=1
#test $TEST_WORKERS = 0 && _WORKERS_COUNT=0
if [ $_fus_attn = "$_DEFAULT_FUSED_ATTN" ]; then
echo ==== Run mGPU with Fused attention backend: $_fus_attn ====
run 3 test_fused_optimizer.py
run 3 test_sanity_import.py
run 3 distributed/test_fusible_ops.py
run 3 distributed/test_numerics.py
run 3 distributed/test_torch_fsdp2.py
run 3 distributed/test_torch_fsdp2_fp8.py
run 3 fused_attn/test_fused_attn_with_cp.py
fi
echo ==== Run mGPU with Fused attention backend: $_fus_attn ====
configure_omp_threads 8
run_default_fa 1 test_fused_optimizer.py
run_default_fa 3 test_sanity_import.py
run_default_fa 2 distributed/test_fusible_ops.py
run_default_fa 2 distributed/test_numerics.py
run_default_fa 1 distributed/test_torch_fsdp2.py
run_default_fa 2 distributed/test_torch_fsdp2_fp8.py
run_default_fa_lbl "flash" 3 fused_attn/test_fused_attn_with_cp.py -k "with_flash"
run_default_fa_lbl "fused" 2 fused_attn/test_fused_attn_with_cp.py -k "with_fused"
}

run_benchmark() {
Expand Down Expand Up @@ -136,13 +121,17 @@ for _fus_attn in auto flash ck aotriton unfused; do
#Flash - Fused attention is disabled
#CK/AOTriton - no Flash attention and only corresponding Fused attention backend is enabled
#Unfused - Flash and Fused attentions are disabled
#Level 1 - run in auto and unfused modes
#Level 3 - run in all but auto and unfused modes
#Level 1 - run in auto mode only
#Level 2 - run in ck and aotriton modes
#Level 3 - run in all but auto modes
if [ $TEST_LEVEL -ge 3 ]; then
test $_fus_attn = auto -o $_fus_attn = unfused && continue
test $_fus_attn = auto && continue
_DEFAULT_FUSED_ATTN="ck"
elif [ $TEST_LEVEL -eq 2 ]; then
test $_fus_attn != aotriton -a $_fus_attn != ck && continue
_DEFAULT_FUSED_ATTN="ck"
else
test $_fus_attn != auto -a $_fus_attn != unfused && continue
test $_fus_attn != auto && continue
_DEFAULT_FUSED_ATTN="auto"
fi

Expand Down