JetStream icon indicating copy to clipboard operation
JetStream copied to clipboard

JetStream is a throughput and memory optimized engine for LLM inference on XLA devices, starting with TPUs (and GPUs in future -- PRs welcome).

Results 35 JetStream issues
Sort by recently updated
recently updated
newest added

Supporting openai-api-compatible endpoints like the `/v1/chat/completions` and `/v1/completions` APIs would have the following benefits: * Allow Jetstream to be used as a drop-in replacement for the vLLM server * Make...

Migrate config path for llama2-70b inference.

pull ready

Bumps the pip group with 1 update in the /experimental/jax directory: [torch](https://github.com/pytorch/pytorch). Updates `torch` from 2.3.0+cpu to 2.6.0 Release notes Sourced from torch's releases. PyTorch 2.6.0 Release Highlights Tracked Regressions...

dependencies
python

I have been successfully running JetStream with the MaxText engine on v4-8. However, all my attempts at running them on v5 seem to fail. I am starting the server with...

Set logger to emit all logs. Previously was set to `INFO` logs only then `WARNING` logs only. Not setting the logger will emit all logs. This reenables viewing the `INFO`...

It seems lexically, the variable metrics_server_config is not needed. Hence this propose to remove it. Please let me know if my coding reading effort went wrong. Thanks!