Skip to content

weekly-tests

weekly-tests #46

Workflow file for this run

name: weekly-tests
on:
workflow_dispatch:
schedule:
- cron: '56 18 * * 5'
env:
WORKSPACE_PREFIX: $(echo $GITHUB_WORKSPACE |cut -d '/' -f 1-4)
SLURM_PARTITION: llm_s
jobs:
training_4GPU:
runs-on: [t_cluster]
timeout-minutes: 15
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_4GPU
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --exclusive --kill-on-bad-exit=1 --job-name=$jobname -n4 --ntasks-per-node=4 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_4GPU" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_4DP2TP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_4DP2TP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_4DP2TP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_4DP2TPSP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_4DP2TPSP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_4DP2TPSP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_4DP2TPSP_optimizer_v2:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_4DP2TPSP_optimizer_v2
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_4DP2TPSP_optimizer_v2" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_4DP2PP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_4DP2PP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_4DP2PP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_4DP2PP_optimizer_v2:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_4DP2PP_optimizer_v2
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_4DP2PP_optimizer_v2" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_4DP2PP_InterleavedOverlap:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_4DP2PP_InterleavedOverlap
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_4DP2PP_InterleavedOverlap" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_16GPU_4DP2TP2PP_MTP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_16GPU_4DP2TP2PP_MTP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n16 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_16GPU_4DP2TP2PP_MTP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_16GPU_4DP2TP2PP_MSP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_16GPU_4DP2TP2PP_MSP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n16 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_16GPU_4DP2TP2PP_MSP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_16GPU_4DP2TP2PP_MSP_optimizer_v2:
runs-on: [t_cluster]
timeout-minutes: 15
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_16GPU_4DP2TP2PP_MSP_optimizer_v2
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n16 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_16GPU_4DP2TP2PP_MSP_optimizer_v2" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_16GPU_4DP2TP2PP_FSP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_16GPU_4DP2TP2PP_FSP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n16 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_16GPU_4DP2TP2PP_FSP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_ISP:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_ISP
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_ISP" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
training_8GPU_ISP_CKPT:
runs-on: [t_cluster]
timeout-minutes: 20
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: training_8GPU_ISP_CKPT
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_ISP_SAVE_CKPT" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
jobname=LOAD_${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -n8 --ntasks-per-node=8 --cpus-per-task=4 --gpus-per-task=1 pytest -s -v --color=yes -m "training_8GPU_ISP_LOAD_CKPT" ./tests/test_training/test_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
unit_test_optimizer:
runs-on: [t_cluster]
timeout-minutes: 35
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: test_optimizer
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -N 1 -n 1 --gres=gpu:8 python -m pytest -s ./tests/test_solver/test_optimizer.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
load_ckpt_then_assert_loss:
runs-on: [t_cluster]
timeout-minutes: 10
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- uses: actions/checkout@v3
with:
ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
- name: test_ckpt_loss
run: |
source activate ${evo_env_torch21_flash2}
jobname=${GITHUB_RUN_ID}-${GITHUB_JOB}-${GITHUB_RUN_ATTEMPT}
srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=$jobname -N 1 -n 1 --gres=gpu:8 python -m pytest -s ./tests/test_training/test_load_ckpt_loss.py
exit_code=$?
sh ./ci_scripts/common/check_slurm_cancled.sh $exit_code $jobname
notify_to_feishu:
if: ${{ always() && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'develop' || github.ref_name == 'main') }}
needs: [
training_4GPU,
training_8GPU_4DP2TP,
training_8GPU_4DP2TPSP,
training_8GPU_4DP2PP,
training_8GPU_4DP2PP_InterleavedOverlap,
training_16GPU_4DP2TP2PP_MTP,
training_16GPU_4DP2TP2PP_MSP,
training_16GPU_4DP2TP2PP_FSP,
training_8GPU_ISP,
training_8GPU_ISP_CKPT,
unit_test_optimizer,
load_ckpt_then_assert_loss
]
runs-on: [t_cluster]
steps:
- name: mask env
run: |
echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
echo "::add-mask::$path_prefix"
- name: notify
run: |
curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"'${{ github.repository }}' GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}' ${{ secrets.WEBHOOK_URL }}