Skip to content

Updates experiment script #1641

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
finbarrtimbers merged 5 commits into from
Apr 27, 2026
+6 −5
Merged

Updates experiment script #1641

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
91b2474
Updates main experiment script.
finbarrtimbers Apr 27, 2026
444d628
updated script
finbarrtimbers Apr 27, 2026
90a8373
Pass BEAKER_IMAGE via env var instead of positional arg. Co-Authored-…
finbarrtimbers Apr 27, 2026
1484667
Simplify script invocation in build_image_and_launch.sh; BEAKER_IMAGE…
finbarrtimbers Apr 27, 2026
46679a5
Revert BEAKER_IMAGE env-var change; align qwen3_4b_dapo_math.sh with …
finbarrtimbers Apr 27, 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
11 changes: 6 additions & 5 deletions scripts/train/qwen/qwen3_4b_dapo_math.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,16 +4,17 @@ EXP_NAME="${EXP_NAME:-qwen3_4b_base_dapo}"
RUN_NAME="${RUN_NAME:-${EXP_NAME}_$(date +%Y%m%d_%H%M%S)}"

NUM_GPUS="${NUM_GPUS:-8}"
BEAKER_IMAGE="${BEAKER_IMAGE:-nathanl/open_instruct_auto}"
BEAKER_IMAGE="${1:-nathanl/open_instruct_auto}"

CLUSTER="${CLUSTER:-ai2/jupiter ai2/ceres}"
PRIORITY="${PRIORITY:-high}"
CLUSTER="${CLUSTER:-ai2/jupiter}"
PRIORITY="${PRIORITY:-urgent}"
WORKSPACE="${WORKSPACE:-ai2/olmo-instruct}"

uv run mason.py \
--task_name ${EXP_NAME} \
--description "${RUN_NAME}" \
--cluster ${CLUSTER} \
--workspace ai2/oe-adapt-code \
--workspace ${WORKSPACE} \
--priority ${PRIORITY} \
--pure_docker_mode \
--no_auto_dataset_cache \
Expand Down Expand Up @@ -43,7 +44,7 @@ uv run open_instruct/grpo_fast.py \
--per_device_train_batch_size 1 \
--dataset_mixer_list hamishivi/DAPO-Math-17k-Processed_filtered 1.0 \
--dataset_mixer_list_splits "train" \
--dataset_mixer_eval_list mnoukhov/aime_2025_openinstruct 1.0 mnoukhov/brumo_2025_openinstruct 1.0 \
--dataset_mixer_eval_list allenai/aime_2025_openinstruct 1.0 allenai/brumo_2025_openinstruct 1.0 \
--dataset_mixer_eval_list_splits "train" \
--max_prompt_token_length 2048 \
--response_length 8192 \
Expand Down
Loading