Named models for wasi-nn #6134
Open
Named models for wasi-nn #6134
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This also updates wasmtime to use the latest version of the wasi-nn spec instead of older commit.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR implements named models for wasi-nn as discussed here WebAssembly/wasi-nn#36.
This PR adds a registry that tracks the byte level form of the models that can be populated from either the host or guest side. This will allow WASM host instances to reuse the same model multiple times without having to fully reload it. It currently imposes a hard-coded 10MB limit on the size of models that can be registered from a guest and extends the Backend wasi-nn API so that it can work directly with byte arrays, instead of passing through GuestPtr instances.
This implementation only caches the compiled graph for CPU execution targets as other targets may consume resources on GPU or TPU hardware, which is not possible to track in a straightforward fashion. In the future, it may become desirable to have visibility over resource usage so that hosts may make more sophisticated decisions about what models are fully cached for execution.
This PR also updates the WITX to the latest version of the spec as it was previously pointing to a much older commit.