laya-apple prebuilt ANE artifacts

Prebuilt Core ML (Apple Neural Engine) artifacts for laya-apple. They are downloaded by laya-apple artifacts fetch MODEL and are not meant to be loaded directly.

What is in this repository

index.json: one entry per archive, with the platform profile it was built and validated on (SoC, macOS major version, coremltools version) and the archive SHA-256.
<model>/<revision[:12]>/<profile>/<model>-L<bucket>.tar.gz: laya-apple artifacts export archives, one per model and sequence-length bucket.

A receiving machine re-validates every archive before use: file hashes, the FP16 parity gate against upstream Laya and the ANE placement probe. An archive that fails is rejected. device="ane" never falls back silently. Archives only match machines of the same platform profile; other machines build locally as before.

Source models and license

The archives contain converted (FP16 Core ML) weights of these checkpoints, at these pinned revisions:

Model	Source revision
convaiinnovations/laya	`c5d78730f3493e4fe16d61507ef4b78eef7318cf`
convaiinnovations/laya-multilingual	`052592a15d198d9ad47da779604259b10b47b7aa`
convaiinnovations/laya-typed-decisions	`f9ab0b228f0fc0f14d873dbc99038f135c2da1b2`

The source checkpoints are licensed under the Apache License 2.0, and so are these converted artifacts. The weights were converted to Core ML and reshaped for the Neural Engine; they were not retrained or fine-tuned. The conversion code is in laya-apple (Apache-2.0).

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for tc3oliver/laya-apple-artifacts

Base model

convaiinnovations/laya

Finetuned

(107)

this model