Stream: git-wasmtime

Topic: wasmtime / issue #3977 Adding the TensorFlow backend to w...


view this post on Zulip Wasmtime GitHub notifications bot (Mar 31 2022 at 01:58):

brianjjones commented on issue #3977:

Depends on #3978

view this post on Zulip Wasmtime GitHub notifications bot (Mar 31 2022 at 02:00):

brianjjones commented on issue #3977:

@abrown This is the latest Tensorflow backend for WASI-NN using runtime linking.

view this post on Zulip Wasmtime GitHub notifications bot (Mar 31 2022 at 02:03):

brianjjones commented on issue #3977:

This is still a work in progress, mostly because it adds considerable size to Wasmtime. For now I'd like to see this as an added optional feature of the wasi-nn crate, that users can choose to use. It works as intended, but as I said its quite large and slimming it down will take some time. Suggestions, concerns, etc welcomed.

view this post on Zulip Wasmtime GitHub notifications bot (Sep 01 2022 at 18:19):

abrown commented on issue #3977:

@brianjjones, I refactored the tf.js file in this commit. This should address some of the comments I made above. There is more refactoring that could/should be done in that area (fix the custom parsing, evaluate the use of Arcs) but hopefully that commit helps!

view this post on Zulip Wasmtime GitHub notifications bot (May 03 2023 at 19:01):

geekbeast commented on issue #3977:

TensorFlow's security model makes it challenging to support inside of wasi-nn. https://chromium.googlesource.com/external/github.com/tensorflow/tensorflow/+/HEAD/SECURITY.md

In addition, the SavedModel format assumes direct file system access and this is baked pretty deeply into the underlying Keras framework. It should be possible to read/write h5 format for byte stream, but it may break some of the functionality of more complicated models.

One last note-- in Python it should be possible to pass a virtualized filesystem, but this is not exposed cleanly through existing C interfaces.


Last updated: Dec 23 2024 at 12:05 UTC