A new nightly version has been released! ![]()
See the quickstart guide for installation instructions: Quickstart | Modular
MAX changelog updates:
- Constrained decoding (structured output) now unpacks the grammar bitmask on
the GPU. The packedint32bitmask is transferred to device as-is and
unpacked and applied to the logits in a single fused kernel
(apply_packed_bitmask), instead of unpacking to abooltensor on the CPU.
Mojo changelog updates:
Raw MAX diff: https://github.com/modular/modular/compare/998f2002d8006f6ce37b0c247affb865913a51ef...b54959854f1bc47e2c3ae86f8697dbea8b181f57)>
Current Mojo changelog: https://github.com/modular/modular/blob/main/mojo/docs/nightly-changelog.md