Skip to content

fix: various follow-up bug fixes #302

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
cquil11 merged 4 commits into from
Dec 7, 2025
Merged

fix: various follow-up bug fixes #302

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
0e2cb1e
various bug fixes
cquil11 Dec 7, 2025
d601ffa
various bug fixes pt 2
cquil11 Dec 7, 2025
2a16ce5
various bug fixes pt 3
cquil11 Dec 7, 2025
7b869d8
various bug fixes pt 4 -- use bash expansion in cw
cquil11 Dec 7, 2025
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
4 changes: 2 additions & 2 deletions .github/workflows/full-sweep-1k1k-scheduler.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@ jobs:
config: ${{ fromJson(needs.get-dsr1-configs.outputs.multi-node-search-space-config) }}
secrets: inherit
with:
exp-name: "dsr1_1k1k"
isl: 1024
osl: 1024
max-model-len: 2248
Expand All @@ -60,7 +61,6 @@ jobs:
model-prefix: ${{ matrix.config.model-prefix }}
framework: ${{ matrix.config.framework }}
precision: ${{ matrix.config.precision }}
exp-name: "dsr1_1k1k"
conc-list: ${{ toJson(matrix.config.conc) }}
spec-decoding: ${{ matrix.config.spec-decoding }}
disagg: ${{ matrix.config.disagg }}
Expand Down Expand Up @@ -116,6 +116,7 @@ jobs:
config: ${{ fromJson(needs.get-gptoss-configs.outputs.multi-node-search-space-config) }}
secrets: inherit
with:
exp-name: "gptoss_1k1k"
isl: 1024
osl: 1024
max-model-len: 2248
Expand All @@ -125,7 +126,6 @@ jobs:
model-prefix: ${{ matrix.config.model-prefix }}
framework: ${{ matrix.config.framework }}
precision: ${{ matrix.config.precision }}
exp-name: "dsr1_1k1k"
conc-list: ${{ toJson(matrix.config.conc) }}
spec-decoding: ${{ matrix.config.spec-decoding }}
disagg: ${{ matrix.config.disagg }}
Expand Down
4 changes: 2 additions & 2 deletions .github/workflows/full-sweep-1k8k-scheduler.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@ jobs:
config: ${{ fromJson(needs.get-dsr1-configs.outputs.multi-node-search-space-config) }}
secrets: inherit
with:
exp-name: "dsr1_1k8k"
isl: 1024
osl: 8192
max-model-len: 9416
Expand All @@ -60,7 +61,6 @@ jobs:
model-prefix: ${{ matrix.config.model-prefix }}
framework: ${{ matrix.config.framework }}
precision: ${{ matrix.config.precision }}
exp-name: "dsr1_1k8k"
conc-list: ${{ toJson(matrix.config.conc) }}
spec-decoding: ${{ matrix.config.spec-decoding }}
disagg: ${{ matrix.config.disagg }}
Expand Down Expand Up @@ -116,6 +116,7 @@ jobs:
config: ${{ fromJson(needs.get-gptoss-configs.outputs.multi-node-search-space-config) }}
secrets: inherit
with:
exp-name: "gptoss_1k8k"
isl: 1024
osl: 8192
max-model-len: 9416
Expand All @@ -125,7 +126,6 @@ jobs:
model-prefix: ${{ matrix.config.model-prefix }}
framework: ${{ matrix.config.framework }}
precision: ${{ matrix.config.precision }}
exp-name: "dsr1_1k8k"
conc-list: ${{ toJson(matrix.config.conc) }}
spec-decoding: ${{ matrix.config.spec-decoding }}
disagg: ${{ matrix.config.disagg }}
Expand Down
4 changes: 2 additions & 2 deletions .github/workflows/full-sweep-8k1k-scheduler.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@ jobs:
config: ${{ fromJson(needs.get-dsr1-configs.outputs.multi-node-search-space-config) }}
secrets: inherit
with:
exp-name: "dsr1_8k1k"
isl: 8192
osl: 1024
max-model-len: 9416
Expand All @@ -60,7 +61,6 @@ jobs:
model-prefix: ${{ matrix.config.model-prefix }}
framework: ${{ matrix.config.framework }}
precision: ${{ matrix.config.precision }}
exp-name: "dsr1_8k1k"
conc-list: ${{ toJson(matrix.config.conc) }}
spec-decoding: ${{ matrix.config.spec-decoding }}
disagg: ${{ matrix.config.disagg }}
Expand Down Expand Up @@ -116,6 +116,7 @@ jobs:
config: ${{ fromJson(needs.get-gptoss-configs.outputs.multi-node-search-space-config) }}
secrets: inherit
with:
exp-name: "gptoss_8k1k"
isl: 8192
osl: 1024
max-model-len: 9416
Expand All @@ -125,7 +126,6 @@ jobs:
model-prefix: ${{ matrix.config.model-prefix }}
framework: ${{ matrix.config.framework }}
precision: ${{ matrix.config.precision }}
exp-name: "dsr1_8k1k"
conc-list: ${{ toJson(matrix.config.conc) }}
spec-decoding: ${{ matrix.config.spec-decoding }}
disagg: ${{ matrix.config.disagg }}
Expand Down
2 changes: 1 addition & 1 deletion runners/launch_b200-nb.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ HF_HUB_CACHE_MOUNT="/mnt/data/hf-hub-cache-${USER: -1}/"
PARTITION="main"
FRAMEWORK_SUFFIX=$([[ "$FRAMEWORK" == "trt" ]] && printf '_trt' || printf '')

UCX_NET_DEVICES=mlx5_4:1,mlx5_5:1,mlx5_6:1,mlx5_7:1,mlx5_8:1,mlx5_9:1,mlx5_10:1,mlx5_11:1
UCX_NET_DEVICES=eth0

# Cleanup any stale enroot locks from previous runs
find /var/cache/enroot-container-images/$UID -type f -name "*.lock" | xargs rm
Expand Down
4 changes: 3 additions & 1 deletion runners/launch_h200-cw.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,12 +21,14 @@ else
CONTAINER_IMAGE=$(realpath $SQUASH_FILE)
fi

# The 'rm -rf /dev/shm/sagemaker_sessions' is to clean up shared memory used by sagemaker sessions inside the container
# This seems to have been introduced in vLLM 0.11.2, but the issue is specific to CoreWeave runners.
srun --jobid=$JOB_ID \
--container-image=$CONTAINER_IMAGE \
--container-mounts=$GITHUB_WORKSPACE:/workspace/,$HF_HUB_CACHE_MOUNT:$HF_HUB_CACHE \
--container-mount-home \
--container-workdir=/workspace/ \
--no-container-entrypoint --export=ALL \
bash -c 'bash benchmarks/'"${EXP_NAME%%_*}_${PRECISION}"'_h200_slurm.sh; rm -rf /dev/shm/sagemaker_sessions'
bash -c "bash benchmarks/${MODEL_CODE}_${PRECISION}_h200${FRAMEWORK_SUFFIX}_slurm.sh; rm -rf /dev/shm/sagemaker_sessions"

scancel $JOB_ID
Loading