GH Repository · dstackai
dstack
Vendor-agnostic orchestration for training, inference and agentic workloads across NVIDIA, AMD, TPU, and Tenstorrent on clouds, Kubernetes, and bare metal.
- stars
- 2,247
- 30-day movement
- +62/day
- Related entries
- 63
- Connections
- 2
skill-packPythonagentic-orchestrationagent-skillsmachine-learningdockergpuslurmcloudtrainingorchestrationcontainersllmsinferencekubernetesfine-tuningpythonk8snvidiaamd
dstack is a vendor-agnostic orchestration tool for training, inference, and agentic workloads across NVIDIA, AMD, TPU, and Tenstorrent hardware on clouds, Kubernetes, and bare metal. This entry is packaged as a skill pack and ships an AGENTS.md file.
Reach for it when you need to schedule ML or agent workloads across heterogeneous accelerators and infrastructure without vendor lock-in.
Use it to
- Orchestrate GPU training jobs across clouds and Kubernetes
- Run inference workloads on NVIDIA, AMD, TPU, or Tenstorrent hardware
- Fine-tune LLMs on managed or bare-metal clusters
- Use the shipped AGENTS.md skill pack in agentic workflows
- Manage containerized workloads with Slurm or Docker backends
For ML engineers and platform teams running multi-vendor compute
- Role
- skill-pack
- Language
- Python
- Licence
- MPL-2.0
- Forks
- 261
- Open issues
- 62
- Last push
- 2026-09-15
- Latest release
- 0.2 · 2023-03-09
- Skills shipped
- 3
topicsorchestrationgpumachine-learningkubernetesllmsagentic