Skip to content

examples/llama: fix FSDP grad-accum-fusion regression + propagate train exit status #145

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

Merged
sudhu2k merged 2 commits into from
Jul 7, 2026
Merged

examples/llama: fix FSDP grad-accum-fusion regression + propagate train exit status #145

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2c193ad
examples/llama: disable gradient accumulation fusion on FSDP paths
wenchenvincent Jun 30, 2026
7d8e7e4
examples/llama: propagate training exit status from train scripts
wenchenvincent Jun 30, 2026
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
22 changes: 21 additions & 1 deletion examples/llama/train_llama2.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -123,6 +123,13 @@ fi

if [ "$FSDP" -eq 1 ] || [ "$MEGATRON_FSDP" -eq 1 ]; then
unset CUDA_DEVICE_MAX_CONNECTIONS
# Gradient accumulation fusion is incompatible with FSDP: torch-FSDP2 hard-asserts against it
# in arguments.py, and Megatron-FSDP crashes at runtime with an fsdp_grads "No buffer found for
# bucket_id" assertion. Force it off so the default-on toggle does not break the FSDP suites.
if [ "$GRADIENT_ACCUMULATION_FUSION" -eq 1 ]; then
echo "FSDP is incompatible with gradient accumulation fusion; disabling fusion (GRADIENT_ACCUMULATION_FUSION=0)."
GRADIENT_ACCUMULATION_FUSION=0
fi
if [ "$TP" -gt 1 ]; then
echo "It is not recommended to use FSDP and TP together. Disabling TP."
TP=1
Expand Down Expand Up @@ -466,8 +473,18 @@ else
run_cmd="$run_cmd |& tee $TRAIN_LOG"
fi

if [ "$NO_TRAINING" -eq 0 ]; then
TRAIN_RC=0
if [ "$NO_TRAINING" -eq 0 ]; then
# pipefail so the run's `... |& tee $TRAIN_LOG` (TEE_OUTPUT=1) returns torchrun's status, not tee's
# (~always 0). Without it a training crash is masked and the script would exit 0. ($? after `eval`
# of the pipeline, since `eval` collapses PIPESTATUS to a single element = tee's status.)
set -o pipefail
eval $run_cmd
TRAIN_RC=$?
set +o pipefail
if [ "$TRAIN_RC" -ne 0 ]; then
echo "ERROR: training (torchrun pretrain_gpt.py) failed with exit code $TRAIN_RC" |& tee -a $TRAIN_LOG
fi
fi


Expand Down Expand Up @@ -509,3 +526,6 @@ MEMUSAGE=$(python3 mean_log_value.py tmp.txt)
echo "mem usages: $MEMUSAGE" |& tee -a "$TRAIN_LOG"
rm tmp.txt

# Propagate the training exit status (the perf-parse above must not mask a torchrun failure).
exit $TRAIN_RC

22 changes: 21 additions & 1 deletion examples/llama/train_llama3.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,6 +137,13 @@ fi

if [ "$FSDP" -eq 1 ] || [ "$MEGATRON_FSDP" -eq 1 ]; then
unset CUDA_DEVICE_MAX_CONNECTIONS
# Gradient accumulation fusion is incompatible with FSDP: torch-FSDP2 hard-asserts against it
# in arguments.py, and Megatron-FSDP crashes at runtime with an fsdp_grads "No buffer found for
# bucket_id" assertion. Force it off so the default-on toggle does not break the FSDP suites.
if [ "$GRADIENT_ACCUMULATION_FUSION" -eq 1 ]; then
echo "FSDP is incompatible with gradient accumulation fusion; disabling fusion (GRADIENT_ACCUMULATION_FUSION=0)."
GRADIENT_ACCUMULATION_FUSION=0
fi
if [ "$TP" -gt 1 ]; then
echo "It is not recommended to use FSDP and TP together. Disabling TP."
TP=1
Expand Down Expand Up @@ -498,8 +505,18 @@ else
run_cmd="$run_cmd |& tee $TRAIN_LOG"
fi

if [ "$NO_TRAINING" -eq 0 ]; then
TRAIN_RC=0
if [ "$NO_TRAINING" -eq 0 ]; then
# pipefail so the run's `... |& tee $TRAIN_LOG` (TEE_OUTPUT=1) returns torchrun's status, not tee's
# (~always 0). Without it a training crash is masked and the script would exit 0. ($? after `eval`
# of the pipeline, since `eval` collapses PIPESTATUS to a single element = tee's status.)
set -o pipefail
eval $run_cmd
TRAIN_RC=$?
set +o pipefail
if [ "$TRAIN_RC" -ne 0 ]; then
echo "ERROR: training (torchrun pretrain_gpt.py) failed with exit code $TRAIN_RC" |& tee -a $TRAIN_LOG
fi
fi


Expand Down Expand Up @@ -540,3 +557,6 @@ grep -Eo 'mem usages: [^|]*' "$TRAIN_LOG" | sed -E 's/.*mem usages: ([0-9\.]+).*
MEMUSAGE=$(python3 mean_log_value.py tmp.txt)
echo "mem usages: $MEMUSAGE" |& tee -a "$TRAIN_LOG"
rm tmp.txt

# Propagate the training exit status (the perf-parse above must not mask a torchrun failure).
exit $TRAIN_RC
Loading