| COMMAND=`echo "$*" | sed 's/\s\+/__SPACE__/g'` | |
| nebulactl run mdl \ | |
| --job_name="train_reranker_sft" \ | |
| --user_id=524947 \ | |
| --queue="pailitao_algo_high_2" \ | |
| --entry="scripts/entry/train_qwen35_entry.py" \ | |
| --nebula_project=pailitao_v100 \ | |
| --algo_name=pytorch280 \ | |
| --worker_count=1 \ | |
| --file.cluster_file="./cluster.json" \ | |
| --user_params="-DCOMMAND=$COMMAND" \ | |
| --nas_file_system_id=98d384a407-okl3.cn-zhangjiakou.nas.aliyuncs.com \ | |
| --oss_access_id=LTAI5tMZb5HHH6HHgAExAyiL \ | |
| --oss_access_key=6MVP8JnTE01t4TLUHhPoFV0swZNTrb \ | |
| --oss_bucket=pailitao-ai \ | |
| --oss_endpoint=oss-cn-zhangjiakou.aliyuncs.com \ | |
| --requirements_file_name=requirements_qwen35.txt | |
| # queue是队列名，algo_name由队列所决定 | |
| # nebula_project是集群名 | |
| # entry是要运行的主文件 | |
| # worker_count 是要用的GPU数量 | |
| # nas_file_system_id 和 oss是存储文件的地址 | |
| # cluster.json中 需要改的是 cpu核心+memory内存，gpu不改 | |