Home
| GitHub | Blog | User Forum | Developer Slack |
About¶
vLLM TorchTPU is the integration of TorchTPU and vLLM. It is a vLLM platform plugin packaged as vllm_torchtpu, with TPU kernels and runtime code for TorchTPU, so PyTorch model definitions run performantly on Cloud TPU with the same vLLM user experience, telemetry, and interface.
What are you trying to do today?¶
-
I'm New
Provision a Cloud TPU, install
vllm-torchtpu, and send your first request. -
I Want to Serve & Scale
Explore supported models, 6D parallelism, P/D disaggregation, and TPU recipes.
-
I Want to Contribute
Contribute code, dive into inference examples, and follow the submission flow.
-
I Want to Learn More
Learn the architecture, run benchmarks, and profile TPU kernels.
Contribute¶
We're always looking for ways to partner with the community to accelerate vLLM TorchTPU development. If you're interested in contributing, check out the Contributing guide, the Roadmap for what's planned next, and Issues to start.

