Manage GPU compute jobs on the Qizhi platform via CLI.
npx skills add https://github.com/Mang30/myskills --skill qzcli-mang30
Please help me install this Agent Skill. Skill: qzcli Source: https://github.com/Mang30/myskills/tree/main/skills/qzcli Command: npx skills add https://github.com/Mang30/myskills --skill qzcli-mang30
qzcli solves the problem of manually managing GPU compute jobs on the Qizhi (启智) platform by providing a kubectl-style workflow for logging in, discovering resources, submitting jobs, and monitoring or stopping them.
Use qzcli to submit and then watch your training job by running: "Log in, discover resources, create a job named 'my-training-job' that runs 'bash train.sh' on workspace 'My Workspace' using compute group 'My Compute Group' with image 'your.registry/team/image:tag' and 4 instances, then list and watch the job until it finishes."
High-intent search queries and answers about installing and using this skill.
To manage distributed training jobs on the Qizhi GPU platform, you can use a kubectl-style workflow to authenticate, discover compute groups, submit batch workloads, and monitor execution status until completion.
Yes, you can submit batch GPU compute jobs by providing a JSON configuration file. This allows you to run checkpoint or step sweep experiments across selected workspaces and compute groups deterministically.
You can discover available compute capacity by refreshing and querying workspaces and compute groups, applying filters like free-node count to find suitable nodes before submitting your job.
Yes, Qizhi job management supports configuration via CLI arguments, stdin, environment variables, .env files, and local config files to authenticate and manage sessions reliably.
The best way to stop or audit running GPU jobs across multiple workspaces is using deterministic CLI commands to list job statuses, watch execution, and stop jobs across one or multiple workspaces.
duypham2801Manage GPU compute jobs on the Qizhi platform via a kubectl-style CLI.
zhuyingqinManage GPU compute jobs on the Qizhi platform via kubectl-style qzcli commands.
googleAutomate batch and HPC workload deployment on Google Kubernetes Engine.
niznik-devOrchestrate SLURM fine-tuning jobs with torchtune and evaluate with inspect-ai.
