Skip to content

chore(fix): Deployment parallelism #2189

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
ko3n1g merged 5 commits into from
Feb 17, 2026
Merged

chore(fix): Deployment parallelism #2189

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
fdbb81d
chore(fix): Deployment parallelism
ko3n1g Feb 3, 2026
e824e12
fix
ko3n1g Feb 3, 2026
cb0e835
Merge branch 'main' into ko3n1g/ci/fix-evaluation
ko3n1g Feb 17, 2026
b79b08c
expert_model_parallel_size
ko3n1g Feb 17, 2026
ccf5e9e
revert
ko3n1g Feb 17, 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
9 changes: 6 additions & 3 deletions examples/evaluation/deploy.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,9 @@ for i in $(env | grep ^PMIX_ | cut -d"=" -f 1); do unset -v $i; done
MEGATRON_CHECKPOINT=$1
NUM_REPLICAS=$2
NUM_GPUS=$3
TP=$4
PP=$5
CP=$6
python \
/opt/Export-Deploy/scripts/deploy/nlp/deploy_ray_inframework.py \
--megatron_checkpoint "$MEGATRON_CHECKPOINT" \
Expand All @@ -14,6 +17,6 @@ python \
--port 8000 \
--num_gpus "$NUM_GPUS" \
--num_replicas "$NUM_REPLICAS" \
--tensor_model_parallel_size 1 \
--pipeline_model_parallel_size 1 \
--context_parallel_size 1
--tensor_model_parallel_size "$TP" \
--pipeline_model_parallel_size "$PP" \
--context_parallel_size "$CP"
14 changes: 13 additions & 1 deletion examples/evaluation/launch_evaluation_pipeline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -111,7 +111,19 @@ def main(args):
executor=executor,
)
job.start(
command=f"bash /opt/Megatron-Bridge/examples/evaluation/deploy.sh {args.megatron_checkpoint} {args.num_replicas} {args.num_gpus} | tee -a deploy.log & sleep 120; bash /opt/Megatron-Bridge/examples/evaluation/eval.sh {args.output_dir} {args.parallelism} | tee -a eval.log",
command=f"""
bash /opt/Megatron-Bridge/examples/evaluation/deploy.sh \
{args.megatron_checkpoint} \
{args.num_replicas} \
{args.num_gpus} \
{args.tensor_model_parallel_size} \
{args.pipeline_model_parallel_size} \
{args.context_model_parallel_size} | tee -a deploy.log & \
sleep 120; \
bash /opt/Megatron-Bridge/examples/evaluation/eval.sh \
{args.output_dir} \
{args.parallelism} | tee -a eval.log
""",
workdir=None,
)

Expand Down
2 changes: 1 addition & 1 deletion scripts/performance/setup_experiment.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -403,7 +403,7 @@ def main(
error_msg = None
n_attempts = 0
exp_name = (
exp_name[:37] if dgxc_cluster is not None else exp_name
exp_name[:33] if dgxc_cluster is not None else exp_name
) # Some k8s clusters have a limit on the length of the experiment name.
wandb_run_id = None
while n_attempts <= max_retries:
Expand Down
Loading