| #!/bin/bash |
| |
| |
|
|
| rm -rf ms-swift/ms_swift.egg-info |
|
|
| CLUSTER_CONFIG='./cluster.json' |
| cat <<EOT > ${CLUSTER_CONFIG} |
| { |
| "worker": { |
| "gpu": 100, |
| "cpu": 1200, |
| "memory": 50000 |
| } |
| } |
| EOT |
|
|
| WORKER_COUNT=1 |
|
|
| |
| args="--config_file=./args/gsm_codi_pmps_math_test.yaml" |
|
|
| nebulactl run mdl --engine=xdl \ |
| --refresh \ |
| --ignore=data/*,tools/pycocoevalcap/*,tools/pycocotools/*,.git/*,Coconut/*,CODI/*,CODI_CT/* \ |
| --queue=pailitao_algo_high_2 \ |
| --worker_count=$WORKER_COUNT \ |
| --odps_project=pailitao_search_offline \ |
| --entry="./test_math.py" \ |
| --oss_access_id=LTAI5tMZb5HHH6HHgAExAyiL \ |
| --oss_access_key=6MVP8JnTE01t4TLUHhPoFV0swZNTrb \ |
| --oss_bucket=pailitao-ai \ |
| --oss_endpoint=oss-cn-zhangjiakou-internal.aliyuncs.com \ |
| --algo_name=pytorch280 \ |
| --user_params="$args" \ |
| --job_name=llama3b_math_k32_t0.1_mt0.1_fac0.3_l8_lr5e4_btz256_nopos_ep300-eval \ |
| --file.cluster_file=$CLUSTER_CONFIG \ |
| --force \ |
| --max_failover_times=2 \ |
| --requirements_file_name="./requirements.txt" |
|
|
| rm ${CLUSTER_CONFIG} |
|
|
