neuralmagic / nm-vllm

A high-throughput and memory-efficient inference and serving engine for LLMs
https://nm-vllm.readthedocs.io
Other
251 stars 10 forks source link

upload RELEASE wheel to pypi.org #361

Closed derekk-nm closed 4 months ago

derekk-nm commented 4 months ago

Besides uploading all generated assets to our NeuralMagic pypi server, we also want to push generated wheels to the external pypi.org server for RELEASE versions only.

The structure of the workflows doesn't allow pre-testing of these changes, since they should only apply when wf_category is "RELEASE". Discussing w/ @andy-neuma , we'll just have to see if it fails/succeeds when we're ready to release.

derekk-nm commented 4 months ago

@andy-neuma , I'm not sure that the name of the wheel as passed to nm-upload-assets-to-gcp.yml will work in there. Does it need some path, or some other context to find the file?