Results for “ssm”
2 skillsmcore-run-on-slurm
Launch distributed Megatron-LM training jobs on a SLURM cluster with a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules, container conventions, monitoring, and per-rank failure diagnosis.
2.2k · bundle
ssh
通过 paramiko 库建立 SSH 连接,在远程服务器上执行命令、安装软件、查看日志,并支持交互式会话。
1 · bundle